The Cryosphere 7 823ndash839 2013wwwthe-cryospherenet78232013doi105194tc-7-823-2013copy Author(s) 2013 CC Attribution 30 License

EGU Journal Logos (RGB)

Advances in Geosciences

Open A

ccess

Natural Hazards and Earth System

Sciences

Open A

ccess

Annales Geophysicae

Open A

ccess

Nonlinear Processes in Geophysics

Open A

ccess

Atmospheric Chemistry

and Physics

Open A

ccess

Atmospheric Chemistry

and Physics

Open A

ccess

Discussions

Atmospheric Measurement

Techniques

Open A

ccess

Atmospheric Measurement

Techniques

Open A

ccess

Discussions

Biogeosciences

Open A

ccess

Open A

ccess

BiogeosciencesDiscussions

Climate of the Past

Open A

ccess

Open A

ccess

Climate of the Past

Discussions

Earth System Dynamics

Open A

ccess

Open A

ccess

Earth System Dynamics

Discussions

GeoscientificInstrumentation

Methods andData Systems

Open A

ccess

GeoscientificInstrumentation

Methods andData Systems

Open A

ccess

Discussions

GeoscientificModel Development

Open A

ccess

Open A

ccess

GeoscientificModel Development

Discussions

Hydrology and Earth System

Sciences

Open A

ccess

Hydrology and Earth System

SciencesO

pen Access

Discussions

Ocean Science

Open A

ccess

Open A

ccess

Ocean ScienceDiscussions

Solid Earth

Open A

ccess

Open A

ccess

Solid EarthDiscussions

The Cryosphere

Open A

ccess

Open A

ccess

The CryosphereDiscussions

Natural Hazards and Earth System

Sciences

Open A

ccess

Discussions

Satellite-derived volume loss rates and glacier speeds for theCordillera Darwin Icefield Chile

A K Melkonian 1 M J Willis 1 M E Pritchard 1 A Rivera23 F Bown2 and S A Bernstein4

1Department of Earth and Atmospheric Sciences Cornell University Ithaca New York USA2Centro de Estudios Cientıficos (CECs) Valdivia Chile3Departamento de Geografıa Universidad de Chile Santiago Chile4St Timothyrsquos School 8400 Greenspring Ave Stevenson MD 21153 USA

Correspondence toA K Melkonian (akm26cornelledu)

Received 12 July 2012 ndash Published in The Cryosphere Discuss 31 August 2012Revised 30 March 2013 ndash Accepted 17 April 2013 ndash Published 14 May 2013

Abstract We produce the first icefield-wide volume changerate and glacier velocity estimates for the Cordillera Dar-win Icefield (CDI) a 2605 km2 temperate icefield in southernChile (696 W 546 S) Velocities are measured from opti-cal and radar imagery between 2001ndash2011 Thirty-six digitalelevation models (DEMs) from ASTER and the SRTM DEMare stacked and a weighted linear regression is applied to el-evations on a pixel-by-pixel basis to estimate volume changerates

The CDI lost mass at an average rate ofminus39plusmn 15 Gt yrminus1

between 2000 and 2011 equivalent to a sea level rise (SLR)of 001plusmn 0004 mm yrminus1 and an area-averaged thinning rateof minus15plusmn 06 m we(water equivalent) yrminus1

Thinning is widespread with concentrations near the frontof two northern glaciers (Marinelli Darwin) and one western(CDI-08) glacier Thickening is apparent in the south mostnotably over the advancing Garibaldi Glacier The north-eastern part of the CDI has an average thinning rate ofminus19plusmn 07 m we yrminus1 while the southwestern part has anaverage thinning rate ofminus10plusmn 04 m we yrminus1

Velocities are obtained over many of the CDI glaciersfor the first time We provide a repeat speed time series atthe Marinelli Glacier There we measure maximum frontspeeds of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in2003 and 10plusmn 03 m dayminus1 in 2011 The maintenance of highfront speeds from 2001 to 2011 supports the hypothesis thatMarinelli is in the retreat phase of the tidewater cycle withdynamic thinning governed by the fjord bathymetry

1 Introduction

We focus on the Cordillera Darwin Icefield (CDI) the thirdlargest temperate icefield in the Southern Hemisphere (Bownet al 2013) which along with the Northern Patagonian Ice-field (NPI) and Southern Patagonian Icefield (SPI) has ex-perienced a rapid reduction in ice-covered area (Rivera et al2007 Masiokas et al 2009 Lopez et al 2010 Davies andGlasser 2012 Willis et al 2012b) The CDI is located inthe southernmost Andes (Fig1) in Tierra del Fuego It iscoalesced around two main mountain peaks Mount Darwin(2469 m aslKoppes et al 2009) and the nearby MountSarmiento (2300 m aslStrelin et al 2008) The icefieldcovers 2605 km2 measured from ice outlines derived fromsatellite imagery acquired from 2001 to 2004 It extendsroughly 200 km westndasheast from 718 W to 685 W androughly 50 km southndashnorth from 549 S to 542 S board-ered to the north by the Almirantazgo Fjord and the Bea-gle Channel in the south Precipitation during the wintercomes predominantly from the southsouthwest (Holmlundand Fuenzalida 1995) and the EndashW orientation of the CDIleads to an orographic effect with greater snowfall on south-ern and western glaciers and drier warmer conditions onnorthern and eastern glaciers (Holmlund and Fuenzalida1995 Strelin and Iturraspe 2007 Koppes et al 2009 Lopezet al 2010) Previous length-change measurements of CDIglaciers show retreat at northern and eastern glaciers and sta-bleadvancing fronts at southern and western glaciers (Holm-lund and Fuenzalida 1995) The ldquosouthernrdquo part or side ofthe CDI refers to southern and western glaciers and the

Published by Copernicus Publications on behalf of the European Geosciences Union

824 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 1 Map of dhdt

for the CDI (area indicated by red box in inset) Three glaciers Marinelli Darwin and CDI-08 stand out with extensivethinning towards their fronts The purple line indicates the divide between the ldquonorthernrdquo and ldquosouthernrdquo regions of the CDI

ldquonorthernrdquo part or side of the CDI refers to northern andeastern glaciers The purple line in Fig1 shows the dividebetween the northern (1322 km2) and southern (1283 km2)sides

There are few studies on the CDI compared to other tem-perate ice fields (Masiokas et al 2009 Lopez et al 2010)such as the Alaskan ice fields the NPI and the SPI (egArendt et al 2002 Rignot et al 2003 Berthier et al 2010Glasser et al 2011 Ivins et al 2011 Willis et al 2012a)Climate and mass balance studies are scarce for southernhemispheric ice bodies outside of Antarctica (Holmlund andFuenzalida 1995 Lopez et al 2010) due to the difficult ac-cess and weather

Studies of observational and reanalysis data indicate theloss of ice at the CDI can be attributed to climatic changesthat include 20th century regional decreases in precipitation(Quintana 2004) coupled with atmospheric warming (Holm-lund and Fuenzalida 1995 Lopez et al 2010) and dynamicinstability at the largest glacier on the icefield Marinelli(eg Holmlund and Fuenzalida 1995) On a local scalechanges in wind patterns have increased precipitation onthe southern side of the CDI (Holmlund and Fuenzalida1995 Strelin and Iturraspe 2007) while decreasing precipi-tation on the northern side (Holmlund and Fuenzalida 1995Koppes et al 2009)

Temperate ice fields are disproportionately large contribu-tors to SLR (egArendt et al 2002 Rignot et al 2003)Rignot et al(2003) claim this is particularly true of thePatagonian glaciers which they say account for 9 of thenonpolar contribution to SLR The CDI along with the NPIand SPI provides an opportunity to examine the response ofdifferent glaciers (eg calving vs noncalving) in differentclimates (maritime on the southern side versus more conti-nental on the northern side) to regional changes in climate

(Holmlund and Fuenzalida 1995) and unlike the NPI andSPI the contribution of the CDI to SLR has not yet been es-timated (Lopez et al 2010)

The CDI is the closest icefield to the Antarctic Peninsulaa region that has also experienced significant warming Massloss at the CDI might be contaminating GRACE measure-ments of the Antarctic Peninsula NPI and SPI (Ivins et al2011) so our constraints on the mass loss rate occurring atthe CDI will help isolate this signal Thinning and accelera-tion have been observed on glaciers in the Antarctic Penin-sula and the NPI (Pritchard and Vaughan 2007 Willis et al2012a) we assess whether this is the case for any glaciers onthe CDI

In this study we calculate both the elevation change rates(dh

dt) over the entire CDI and measure glacier velocities using

pixel-tracking applied to pairs of optical and radar imagesWith dh

dtand an assumed density of material lostgained we

can estimate the mass change rate allowing us to quantifyits SLR contribution and compare it with other ice fields Wealso use the surface elevation change rates to identify whichglaciers are providing the largest contribution to SLR andshould be the focus of further study Additionally measuringglacier velocities allows an estimate of mass flux out of theglacier if the thickness is known Increased speed and massflux through the front of the glacier can cause ldquodynamic thin-ningrdquo if it is not balanced by increased mass input Our re-sults will provide a baseline measurement over many glaciersand areas of the icefield for which ice velocities have notbeen measured

The date format used throughout this paper isDDMMYYYY

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 825

2 Methods

21 Data Preparation

The Advanced Spaceborne Thermal Emission and ReflectionRadiometer (ASTER) has a stereo-imaging capability en-abling DEMs to be generated on-demand by NASArsquos LandProcesses Distributed Active Archive Center (LP DAAC)(Fujisada et al 2005) In this study ASTER DEMs (prod-uct 14) are used to calculatedh

dt while band 3N images

(product 1B) are used for pixel-tracking Clouds are notmasked during data preparation rather they are largely re-moved by excluding elevations based on deviation from thethe SRTM DEM (Sect22) NASArsquos Automatic Registrationand Orthorectification Package (AROPGao et al 2009) isused to co-register ASTER images and DEMs to a LandsatGLS image (available from the Global Land Cover Facility)and orthorectify the ASTER L1B images using the Shut-tle Radar Topography Mission (SRTM) DEM (acquired in2000) Landsat GLS images are orthorectified to the SRTMDEM (Tucker et al 2004) therefore co-registering theASTER imagery to the Landsat GLS image effectively co-registers them to the SRTM DEM (seeWillis et al 2012afordetails) The SRTM DEM is our default DEM for orthorec-tifying the ASTER imagery it is not as adversely affectedby clouds snow and other features that cause errors in op-tically derived (eg ASTER) DEMs (Scherler et al 2008)All ASTER images used for pixel-tracking are orthorectifiedto the SRTM DEM except for the 07092001 to 25092001and 06092003 to 13092003 pairs covering the front ofMarinelli Glacier where pronounced thinning necessitatesthe use of the corresponding ASTER DEMs These pairsare orthorectified to the 25092001 and 13092003 ASTERDEMs respectively

22 Elevation change rates

Horizontally co-registered ASTER DEMs are subsequentlyvertically co-registered and a weighted linear regression isapplied to calculatedh

dtfor each pixel In all 36 ASTER

DEMs (derived from imagery acquired from 2001 to 2011)and the SRTM DEM (acquired in February 2000) are pro-cessed with an average of 4ndash5 elevations per pixel incor-porated into the regression (Fig2 shows elevations and re-gression lines for several points on Marinelli Glacier) Eachelevation is weighted by the inverse of the standard devia-tion of the bedrock elevation differences between its ASTERDEM and the SRTM DEM This is a common measure ofthe uncertainty associated with ASTER DEM elevations (Fu-jisada et al 2005 Howat et al 2008) Horizontal misalign-ment will appear as off-ice (ldquobedrockrdquo) elevation differenceswhen comparing ASTER DEMs to the SRTM DEM and istherefore included in the standard deviation of the bedrockelevation differences (ie the greater the misalignment thegreater the standard deviation which is the measure of un-

certainty that we use for each ASTER DEM) We typicallyfind values from 8 m to 20 m as our uncertainty for ASTERDEMs similar to other studies (Kaab 2002 Fujisada et al2005 San and Suzen 2005 Rivera et al 2007 Howat et al2008 Berthier et al 2010)

In the ablation zone we exclude elevations from the re-gression that deviate more than +5minus30 m yrminus1 from thefirst elevation in the accumulation zone the cutoff is+5minus10 m yrminus1 These cutoffs largely remove the areas oferroneously high elevations due to clouds as well as out-liers caused by shadow snow and other sources which eitherobscure the ice or are largely featureless at ASTER resolu-tion It is important to note that although these cutoffs arebased on expecteddh

dt they are imposed on each individual

map point (see Fig2) before applying the weighted linearregression rather than to already calculateddh

dt which would

yield drastically different results The first elevation in ourtime series is SRTM for 94 of the pixels for the remain-ing areas we use the first available elevation from ASTERThe average end date is 17052008 The high percentageof points with SRTM elevations makes our cutoff strategyfeasible The radar-derived SRTM DEM is not influencedby clouds like ASTER making SRTM elevations a reliablestarting point

We limit the maximum negative deviation tominus30 m yrminus1

in the ablation zone to just capture the maximum thinningand exclude spurious elevations from the regression calcula-tions Changing this tominus60 m yrminus1 has no affect on the zonesof maximum thinning Less thinning is expected in the accu-mulation zone than in the ablation zone therefore we allowa maximum negative deviation of onlyminus10 m yrminus1 above theELA Permitting a greater negative deviation in the accumu-lation zone only results in incoherent areas of extreme andunrealistic thinning

In order to determine the ELA and distinguish where toimpose our different constraints we examine late season(January and February) ASTER images We record the al-titude of the transition between bare ice and snow and takethis as a rough proxy for the ELA (egBamber and Rivera2007) We are able to measure the ELA for 11 larger glaciersfor the rest of the icefield we assume a regional ELA of1090 m (Strelin and Iturraspe 2007)

The volume change rate at each pixel is thedhdt

for the pixelmultiplied by the area of the pixel Summing together the vol-ume change rate of every pixel yields a volume change rate(dV

dt) for the entire icefielddV

dtis multiplied by the density

of glacier ice set to 900 kg mminus3 (egCuffey and Paterson2010) to produce an estimate of the mass change rate Futureground-based measurements of densities on the CDI will beneeded to find more accurate and precise values

We have a sufficient number of ASTER DEMs to providedhdt

rates for 96 of the total area of the icefield Each pixelin the remaining small gaps is filled with the mediandh

dtvalue

within 1 km This local method is more accurate than using

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

826 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

0

100

200

300

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

1 dhdt=minus2648 myr

Excluded

dhdt

300

400

500

600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

2 dhdt=minus1389 myr

400

500

600

700

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

3 dhdt=minus1163 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

4 dhdt=minus937 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

5 dhdt=minus763 myr

900

1000

1100

1200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

6 dhdt=minus083 myr

1300

1400

1500

1600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

7 dhdt=065 myr

1600

1700

1800

1900

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

8 dhdt=426 myr

1900

2000

2100

2200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

9 dhdt=minus023 myr

2100

2200

2300

2400

2500

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

10 dhdt=248 myr

69˚38W

54˚38S

54˚30S

minus20

minus15

minus10

minus5

0

5dhdt

1

2

3

4

5

6

78

9

10

Fig 2 Elevation values anddhdt

for randomly selected pixels over Marinelli Glacier The left-most elevation in each graph is the SRTM

elevation at that pixel Blue lines indicate thedhdt

calculated for each pixel elevation points bolded red are excluded fromdhdt

calculation The

bottom right panel shows thedhdt

map with numbered circles indicating the location corresponding to each graph

the averagedhdt

of an elevation bin to fill gaps within that binor filling in gaps with the mean of alldh

dtcalculated (egRig-

not et al 2003) For example the average low-elevationdhdt

and overall averagedhdt

are both negative these rates wouldnot be suitable for filling a gap near the front of GaribaldiGlacier which is known to be advancing and where we mea-sure positivedh

dtat low elevations

Penetration of C-band radar into ice and (particularly)snow (egRignot et al 2001) is a potential problem whenusing the SRTM DEM We use a technique pioneered byGardelle et al(2012) and applied byWillis et al (2012b)to compensate for penetration effectsWillis et al (2012b)compared X-band SRTM elevations (which have negligi-ble penetration) with C-band SRTM elevations and found

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 827

approximately 2 m of C-band penetration over the SPI at allelevations Due to a lack of X-band SRTM coverage overthe CDI we cannot provide a similar analysis here Insteadwe assume the penetration depth is similar to that at SPI andadd 2 m to each SRTM elevation over ice which increasesour mass loss rate by about 13 We note the CDI is colderthan the SPI which could lead to drier conditions and greaterpenetration (Rignot et al 2001) but we do not have adequateX-band data to quantify the difference

An additional source of mass change that we consider issub-aqueous mass lossgain (which does not contribute to sealevel rise) We cannot directly measure sub-aqueous ice gainsor losses but make a rough estimate by measuring changesin the areal extent at the front of several glaciers that have un-dergone relatively large advance or retreat (from image pairs)and multiplying these area changes by the approximate depthbelow water of the ice We assume an average depth belowwater of 150 m for Marinelli Glacier (seeKoppes et al 2009Fig 4a and b) and 60 m for the other glaciers with an uncer-tainty of plusmn50 m (the uncertainty on the change in area ofeach glacier is negligible) Dividing the sub-aqueous volumechange by the time interval separating the images used tofind area change gives a rate This calculation shows that sub-aqueous mass loss while not well constrained in this studyis an order of magnitude lower than the overall mass changerate calculated fromdh

dt(lt 5 )

221 Sources of uncertainty

Below we consider sources of uncertainty on our masschange rate These include the uncertainty on the elevationsincoporated into the regression uncertainty on the ELA theeffect of varying the maximum deviation allowed from thefirst elevation different density scenarios and uncertainty onthe penetration depth of the C-band SRTM DEM

The uncertainty associated with thedhdt

for each pixel iscalculated from the model covariance matrix (egAsteret al 2005) which accounts for the uncertainties on the el-evations incorporated into the regression The 95 confi-dence interval for the volume change rate uncertainty is cal-culated using the formula 196times U

radicN

U is the total ldquovolumerdquo

of uncertainty calculated by taking the uncertainty at eachpixel multiplying it by the area of the pixel (to determinea ldquovolumerdquo of uncertainty for that pixel) and then addingtogether ldquovolumerdquo of uncertainty for each pixel where adh

dtis calculatedN is the number of independent pixels (egHowat et al 2008) which we determine by dividing the to-tal area by the area over which off-icedh

dtare correlated (eg

Rolstad et al 2009) We estimate the scale at which thedhdt

are independent by finding the area at which the variance ofthe off-icedh

dtbegins to ldquoflattenrdquo (seeRolstad et al 2009and

Willis et al 2012afor details on the method) which we es-timate to be 1800 m by 1800 m (Fig4) This is analogousto the ldquocornerrdquo point on anL curve (egAster et al 2005p 91 Fig 52) and indicates the lengthscale past which the

dhdt

are no longer correlated The total contribution from theuncertainty on individual elevations issim 035 Gt yrminus1

The regional ELA is poorly known In order to investigatethe impact of changing the regional ELA on the mass lossrates we lower our regional ELA from 1090 to 650 m anELA that has been found for several glaciers on the southernand western regions of the CDI (Bown et al 2013) We alsoassume an uncertainty ofplusmn200 m on the 11 glaciers wherewe have estimated the ELA from optical imagery The dif-ference between the two ELA scenarios is incorporated intoour uncertainties (sim 037 Gt yrminus1)

The deviation allowed from the first elevation has a largeimpact on the mass loss rate (egWillis et al 2012b) Therate produced by allowing +10 m yrminus1 is contaminated bylow clouds but is given as a rough minimum estimate ofthe mass loss rate The unsymmetric cutoff (+5 m yrminus1 ver-susminus10minus30 m yrminus1) may bias our results towards thinningbut we argue that an unsymmetric cutoff is more physicallyjustified than a symmetric cutoff Rapid retreat has been in-dependently observed at Marinelli Glacier (Holmlund andFuenzalida 1995 Koppes et al 2009) from which we knowthat a large amount of thinning must be occurring The max-imum allowed negative deviation from the first elevation ofminus30 m yrminus1 just captures the maximum thinning at MarinelliGlacier (see Fig2 point 1)

The cutoff of +5 m yrminus1 from the first elevation is basedon the limited precipitation data available for this region Theupper limit on precipitation in Patagonia is about 10 m yrminus1

(egHolmlund and Fuenzalida 1995 Rignot et al 2003Rasmussen et al 2007 Koppes et al 2011) and PRECISmodel results show maximum precipitation in the 4000 to6000 mm yrminus1 range for the CDI (Fernandez et al 2011)One year of accumulated precipitation undergoes densifica-tion into firn and eventually ice and provides much less than+10 m yrminus1 of elevation change through time so +5 m yrminus1

is chosen as a reasonable upper limit on the maximum thick-ening expected over the large areas covered in this studyWe note that a point measurement may yield adh

dthigher

than +5 m yrminus1 but we would not expect a single point to berepresentative of sustained thickening rates over square kilo-meters Figure2 highlights the importance of applying thiscutoff cloud-influenced elevations would seriously degradethe quality of our regressions for each pixel Tests using ahigher positive cutoff produced discontinuous and incoher-ent ldquosplotchesrdquo of extreme positivedh

dtthat are unrealistic

The only exception is the lower ablation zone of GaribaldiGlacier where we allow a maximum positive deviation of+10 m yrminus1 to accommodate the known advance of Garibaldi(Fig 3)

The sub-panels in Fig2 illustrate the problem of using apositive cutoff of +30 m yrminus1 which would lead to the inclu-sion of ASTER elevations that are obvious outliers ASTERelevations in the accumulation zone are generally less reli-able due to greater cloud and snow cover (lack of contrast)

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

828 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 3 Frontal variation history of Garibaldi Glacier from Land-sat TM ASTER and aerial photographs The background is a15012011 ASTER image

Errors due to clouds tend to be positive rather than nega-tive this is confirmed by comparing the average off-icedh

dtof increasing symmetric maximum allowed deviations Al-lowing plusmn5 plusmn10 andplusmn30 m yrminus1 results in averagedh

dtof

minus004 m yrminus1 +010 m yrminus1 and +082 m yrminus1 Allowing adeviation ofminus30 m yrminus1 is necessary to capture maximumthinning but 330 m of thickening (+30 m yrminus1 from 2000 to2011) is unreasonable in both the ablation and accumulationzone for this region (Koppes et al 2009 Lopez et al 2010)

In situ measurements of accumulation rate on the CDI arerequired to refine our estimates further the cutoffs we useare the best available based on the literature (egFernan-dez et al 2011) To assess the uncertainty due to the choiceof maximum deviation allowed we find the mode (most fre-quent occurrence ie peak) of the distribution of elevationdifferences between all ASTER DEMs and the SRTM DEM(normalized by dividing by the time interval between eachASTER DEM and the SRTM DEM) The mode is not af-fected by the choice of allowed deviation and so is an inde-pendent measure that we can compare our regression-derivedrates to

We do not present the ldquomode raterdquo as representing thebest estimate of the averagedh

dt It is approximately what the

rate would be if the distribution were Gaussian around thepeak dh

dt This implicitly assumes that rates that are equidis-

tant from the mode rate are equally likely (eg if the moderate isminus15 m yrminus1) then rates of +185 m yrminus1 are approx-imately as common as rates ofminus205 m yrminus1 We consider

1800 m

Data (variance of o-ice dhdt atcorresponding lengthscale on x-axis)

Fit

Fig 4 A plot of pixel size (listed as side length of square pixel)vs variance of bedrockdh

dt(calculated using aminus10+10 m yrminus1 cut-

off) plotted as red dots A linear fit to log(x)y shows the trend ofthe data (blue line) The black circle indicates the point at whichthe curve is ldquoflatteningrdquo ie the lengthscale past which the vari-ance no longer changes significantly indicating that thedh

dtare no

longer correlated 1800 m is selected as a conservative estimate ofthe decorrelation length for thedh

dt

this unlikely given the coherent thinning of approximatelyminus25 m yrminus1 estimated at the front of several glaciers and thelack of evidence for a similar degree of coherent thickeninganywhere on the icefield Therefore the mode rate is proba-bly lower than the the actual thinning rate and the differencebetween the mode rate and our regression-deriveddh

dtis an

overestimate of the maximum ldquobiasrdquo due to our choice ofdhdt

We add the difference (plusmn122 Gt yrminus1) that results from

using the same density and penetration assumptions as ourregression rate to our uncertainties to fully account for anypossible bias

We assume all volume change is at a density of900 kg mminus3 consistent with previous studies (egRignotet al 2003 Berthier et al 2004 2010) However the densityof the lost material is likely variable We consider two dif-ferent density scenarios Assuming that firn (600 kg mminus3) islost in the accumulation zone and ice (900 kg mminus3) is lost inthe ablation zone (considered as a possible scenario byKaabet al 2012) reduces the mass change rate by 03 Gt yrminus1Assuming an ice density of 900plusmn 25 kg mminus3 (egGardneret al 2012) addsplusmn01 Gt yrminus1 to our uncertainty The ef-fect of density changes are small (sim 10 ) as most of themass loss occurs from ice in the ablation zone We incorpo-rate the 600900 kg m3 scenario into our uncertainties as itadds a furtherplusmn03 Gt yrminus1 and is thus a more conservativeestimate of uncertainty

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

824 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 1 Map of dhdt

for the CDI (area indicated by red box in inset) Three glaciers Marinelli Darwin and CDI-08 stand out with extensivethinning towards their fronts The purple line indicates the divide between the ldquonorthernrdquo and ldquosouthernrdquo regions of the CDI

ldquonorthernrdquo part or side of the CDI refers to northern andeastern glaciers The purple line in Fig1 shows the dividebetween the northern (1322 km2) and southern (1283 km2)sides

There are few studies on the CDI compared to other tem-perate ice fields (Masiokas et al 2009 Lopez et al 2010)such as the Alaskan ice fields the NPI and the SPI (egArendt et al 2002 Rignot et al 2003 Berthier et al 2010Glasser et al 2011 Ivins et al 2011 Willis et al 2012a)Climate and mass balance studies are scarce for southernhemispheric ice bodies outside of Antarctica (Holmlund andFuenzalida 1995 Lopez et al 2010) due to the difficult ac-cess and weather

Studies of observational and reanalysis data indicate theloss of ice at the CDI can be attributed to climatic changesthat include 20th century regional decreases in precipitation(Quintana 2004) coupled with atmospheric warming (Holm-lund and Fuenzalida 1995 Lopez et al 2010) and dynamicinstability at the largest glacier on the icefield Marinelli(eg Holmlund and Fuenzalida 1995) On a local scalechanges in wind patterns have increased precipitation onthe southern side of the CDI (Holmlund and Fuenzalida1995 Strelin and Iturraspe 2007) while decreasing precipi-tation on the northern side (Holmlund and Fuenzalida 1995Koppes et al 2009)

Temperate ice fields are disproportionately large contribu-tors to SLR (egArendt et al 2002 Rignot et al 2003)Rignot et al(2003) claim this is particularly true of thePatagonian glaciers which they say account for 9 of thenonpolar contribution to SLR The CDI along with the NPIand SPI provides an opportunity to examine the response ofdifferent glaciers (eg calving vs noncalving) in differentclimates (maritime on the southern side versus more conti-nental on the northern side) to regional changes in climate

(Holmlund and Fuenzalida 1995) and unlike the NPI andSPI the contribution of the CDI to SLR has not yet been es-timated (Lopez et al 2010)

The CDI is the closest icefield to the Antarctic Peninsulaa region that has also experienced significant warming Massloss at the CDI might be contaminating GRACE measure-ments of the Antarctic Peninsula NPI and SPI (Ivins et al2011) so our constraints on the mass loss rate occurring atthe CDI will help isolate this signal Thinning and accelera-tion have been observed on glaciers in the Antarctic Penin-sula and the NPI (Pritchard and Vaughan 2007 Willis et al2012a) we assess whether this is the case for any glaciers onthe CDI

In this study we calculate both the elevation change rates(dh

dt) over the entire CDI and measure glacier velocities using

pixel-tracking applied to pairs of optical and radar imagesWith dh

dtand an assumed density of material lostgained we

can estimate the mass change rate allowing us to quantifyits SLR contribution and compare it with other ice fields Wealso use the surface elevation change rates to identify whichglaciers are providing the largest contribution to SLR andshould be the focus of further study Additionally measuringglacier velocities allows an estimate of mass flux out of theglacier if the thickness is known Increased speed and massflux through the front of the glacier can cause ldquodynamic thin-ningrdquo if it is not balanced by increased mass input Our re-sults will provide a baseline measurement over many glaciersand areas of the icefield for which ice velocities have notbeen measured

The date format used throughout this paper isDDMMYYYY

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 825

2 Methods

21 Data Preparation

The Advanced Spaceborne Thermal Emission and ReflectionRadiometer (ASTER) has a stereo-imaging capability en-abling DEMs to be generated on-demand by NASArsquos LandProcesses Distributed Active Archive Center (LP DAAC)(Fujisada et al 2005) In this study ASTER DEMs (prod-uct 14) are used to calculatedh

dt while band 3N images

(product 1B) are used for pixel-tracking Clouds are notmasked during data preparation rather they are largely re-moved by excluding elevations based on deviation from thethe SRTM DEM (Sect22) NASArsquos Automatic Registrationand Orthorectification Package (AROPGao et al 2009) isused to co-register ASTER images and DEMs to a LandsatGLS image (available from the Global Land Cover Facility)and orthorectify the ASTER L1B images using the Shut-tle Radar Topography Mission (SRTM) DEM (acquired in2000) Landsat GLS images are orthorectified to the SRTMDEM (Tucker et al 2004) therefore co-registering theASTER imagery to the Landsat GLS image effectively co-registers them to the SRTM DEM (seeWillis et al 2012afordetails) The SRTM DEM is our default DEM for orthorec-tifying the ASTER imagery it is not as adversely affectedby clouds snow and other features that cause errors in op-tically derived (eg ASTER) DEMs (Scherler et al 2008)All ASTER images used for pixel-tracking are orthorectifiedto the SRTM DEM except for the 07092001 to 25092001and 06092003 to 13092003 pairs covering the front ofMarinelli Glacier where pronounced thinning necessitatesthe use of the corresponding ASTER DEMs These pairsare orthorectified to the 25092001 and 13092003 ASTERDEMs respectively

22 Elevation change rates

Horizontally co-registered ASTER DEMs are subsequentlyvertically co-registered and a weighted linear regression isapplied to calculatedh

dtfor each pixel In all 36 ASTER

DEMs (derived from imagery acquired from 2001 to 2011)and the SRTM DEM (acquired in February 2000) are pro-cessed with an average of 4ndash5 elevations per pixel incor-porated into the regression (Fig2 shows elevations and re-gression lines for several points on Marinelli Glacier) Eachelevation is weighted by the inverse of the standard devia-tion of the bedrock elevation differences between its ASTERDEM and the SRTM DEM This is a common measure ofthe uncertainty associated with ASTER DEM elevations (Fu-jisada et al 2005 Howat et al 2008) Horizontal misalign-ment will appear as off-ice (ldquobedrockrdquo) elevation differenceswhen comparing ASTER DEMs to the SRTM DEM and istherefore included in the standard deviation of the bedrockelevation differences (ie the greater the misalignment thegreater the standard deviation which is the measure of un-

certainty that we use for each ASTER DEM) We typicallyfind values from 8 m to 20 m as our uncertainty for ASTERDEMs similar to other studies (Kaab 2002 Fujisada et al2005 San and Suzen 2005 Rivera et al 2007 Howat et al2008 Berthier et al 2010)

In the ablation zone we exclude elevations from the re-gression that deviate more than +5minus30 m yrminus1 from thefirst elevation in the accumulation zone the cutoff is+5minus10 m yrminus1 These cutoffs largely remove the areas oferroneously high elevations due to clouds as well as out-liers caused by shadow snow and other sources which eitherobscure the ice or are largely featureless at ASTER resolu-tion It is important to note that although these cutoffs arebased on expecteddh

dt they are imposed on each individual

map point (see Fig2) before applying the weighted linearregression rather than to already calculateddh

dt which would

yield drastically different results The first elevation in ourtime series is SRTM for 94 of the pixels for the remain-ing areas we use the first available elevation from ASTERThe average end date is 17052008 The high percentageof points with SRTM elevations makes our cutoff strategyfeasible The radar-derived SRTM DEM is not influencedby clouds like ASTER making SRTM elevations a reliablestarting point

We limit the maximum negative deviation tominus30 m yrminus1

in the ablation zone to just capture the maximum thinningand exclude spurious elevations from the regression calcula-tions Changing this tominus60 m yrminus1 has no affect on the zonesof maximum thinning Less thinning is expected in the accu-mulation zone than in the ablation zone therefore we allowa maximum negative deviation of onlyminus10 m yrminus1 above theELA Permitting a greater negative deviation in the accumu-lation zone only results in incoherent areas of extreme andunrealistic thinning

In order to determine the ELA and distinguish where toimpose our different constraints we examine late season(January and February) ASTER images We record the al-titude of the transition between bare ice and snow and takethis as a rough proxy for the ELA (egBamber and Rivera2007) We are able to measure the ELA for 11 larger glaciersfor the rest of the icefield we assume a regional ELA of1090 m (Strelin and Iturraspe 2007)

The volume change rate at each pixel is thedhdt

for the pixelmultiplied by the area of the pixel Summing together the vol-ume change rate of every pixel yields a volume change rate(dV

dt) for the entire icefielddV

dtis multiplied by the density

of glacier ice set to 900 kg mminus3 (egCuffey and Paterson2010) to produce an estimate of the mass change rate Futureground-based measurements of densities on the CDI will beneeded to find more accurate and precise values

We have a sufficient number of ASTER DEMs to providedhdt

rates for 96 of the total area of the icefield Each pixelin the remaining small gaps is filled with the mediandh

dtvalue

within 1 km This local method is more accurate than using

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

826 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

0

100

200

300

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

1 dhdt=minus2648 myr

Excluded

dhdt

300

400

500

600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

2 dhdt=minus1389 myr

400

500

600

700

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

3 dhdt=minus1163 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

4 dhdt=minus937 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

5 dhdt=minus763 myr

900

1000

1100

1200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

6 dhdt=minus083 myr

1300

1400

1500

1600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

7 dhdt=065 myr

1600

1700

1800

1900

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

8 dhdt=426 myr

1900

2000

2100

2200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

9 dhdt=minus023 myr

2100

2200

2300

2400

2500

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

10 dhdt=248 myr

69˚38W

54˚38S

54˚30S

minus20

minus15

minus10

minus5

0

5dhdt

1

2

3

4

5

6

78

9

10

Fig 2 Elevation values anddhdt

for randomly selected pixels over Marinelli Glacier The left-most elevation in each graph is the SRTM

elevation at that pixel Blue lines indicate thedhdt

calculated for each pixel elevation points bolded red are excluded fromdhdt

calculation The

bottom right panel shows thedhdt

map with numbered circles indicating the location corresponding to each graph

the averagedhdt

of an elevation bin to fill gaps within that binor filling in gaps with the mean of alldh

dtcalculated (egRig-

not et al 2003) For example the average low-elevationdhdt

and overall averagedhdt

are both negative these rates wouldnot be suitable for filling a gap near the front of GaribaldiGlacier which is known to be advancing and where we mea-sure positivedh

dtat low elevations

Penetration of C-band radar into ice and (particularly)snow (egRignot et al 2001) is a potential problem whenusing the SRTM DEM We use a technique pioneered byGardelle et al(2012) and applied byWillis et al (2012b)to compensate for penetration effectsWillis et al (2012b)compared X-band SRTM elevations (which have negligi-ble penetration) with C-band SRTM elevations and found

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 827

approximately 2 m of C-band penetration over the SPI at allelevations Due to a lack of X-band SRTM coverage overthe CDI we cannot provide a similar analysis here Insteadwe assume the penetration depth is similar to that at SPI andadd 2 m to each SRTM elevation over ice which increasesour mass loss rate by about 13 We note the CDI is colderthan the SPI which could lead to drier conditions and greaterpenetration (Rignot et al 2001) but we do not have adequateX-band data to quantify the difference

An additional source of mass change that we consider issub-aqueous mass lossgain (which does not contribute to sealevel rise) We cannot directly measure sub-aqueous ice gainsor losses but make a rough estimate by measuring changesin the areal extent at the front of several glaciers that have un-dergone relatively large advance or retreat (from image pairs)and multiplying these area changes by the approximate depthbelow water of the ice We assume an average depth belowwater of 150 m for Marinelli Glacier (seeKoppes et al 2009Fig 4a and b) and 60 m for the other glaciers with an uncer-tainty of plusmn50 m (the uncertainty on the change in area ofeach glacier is negligible) Dividing the sub-aqueous volumechange by the time interval separating the images used tofind area change gives a rate This calculation shows that sub-aqueous mass loss while not well constrained in this studyis an order of magnitude lower than the overall mass changerate calculated fromdh

dt(lt 5 )

221 Sources of uncertainty

Below we consider sources of uncertainty on our masschange rate These include the uncertainty on the elevationsincoporated into the regression uncertainty on the ELA theeffect of varying the maximum deviation allowed from thefirst elevation different density scenarios and uncertainty onthe penetration depth of the C-band SRTM DEM

The uncertainty associated with thedhdt

for each pixel iscalculated from the model covariance matrix (egAsteret al 2005) which accounts for the uncertainties on the el-evations incorporated into the regression The 95 confi-dence interval for the volume change rate uncertainty is cal-culated using the formula 196times U

radicN

U is the total ldquovolumerdquo

of uncertainty calculated by taking the uncertainty at eachpixel multiplying it by the area of the pixel (to determinea ldquovolumerdquo of uncertainty for that pixel) and then addingtogether ldquovolumerdquo of uncertainty for each pixel where adh

dtis calculatedN is the number of independent pixels (egHowat et al 2008) which we determine by dividing the to-tal area by the area over which off-icedh

dtare correlated (eg

Rolstad et al 2009) We estimate the scale at which thedhdt

are independent by finding the area at which the variance ofthe off-icedh

dtbegins to ldquoflattenrdquo (seeRolstad et al 2009and

Willis et al 2012afor details on the method) which we es-timate to be 1800 m by 1800 m (Fig4) This is analogousto the ldquocornerrdquo point on anL curve (egAster et al 2005p 91 Fig 52) and indicates the lengthscale past which the

dhdt

are no longer correlated The total contribution from theuncertainty on individual elevations issim 035 Gt yrminus1

The regional ELA is poorly known In order to investigatethe impact of changing the regional ELA on the mass lossrates we lower our regional ELA from 1090 to 650 m anELA that has been found for several glaciers on the southernand western regions of the CDI (Bown et al 2013) We alsoassume an uncertainty ofplusmn200 m on the 11 glaciers wherewe have estimated the ELA from optical imagery The dif-ference between the two ELA scenarios is incorporated intoour uncertainties (sim 037 Gt yrminus1)

The deviation allowed from the first elevation has a largeimpact on the mass loss rate (egWillis et al 2012b) Therate produced by allowing +10 m yrminus1 is contaminated bylow clouds but is given as a rough minimum estimate ofthe mass loss rate The unsymmetric cutoff (+5 m yrminus1 ver-susminus10minus30 m yrminus1) may bias our results towards thinningbut we argue that an unsymmetric cutoff is more physicallyjustified than a symmetric cutoff Rapid retreat has been in-dependently observed at Marinelli Glacier (Holmlund andFuenzalida 1995 Koppes et al 2009) from which we knowthat a large amount of thinning must be occurring The max-imum allowed negative deviation from the first elevation ofminus30 m yrminus1 just captures the maximum thinning at MarinelliGlacier (see Fig2 point 1)

The cutoff of +5 m yrminus1 from the first elevation is basedon the limited precipitation data available for this region Theupper limit on precipitation in Patagonia is about 10 m yrminus1

(egHolmlund and Fuenzalida 1995 Rignot et al 2003Rasmussen et al 2007 Koppes et al 2011) and PRECISmodel results show maximum precipitation in the 4000 to6000 mm yrminus1 range for the CDI (Fernandez et al 2011)One year of accumulated precipitation undergoes densifica-tion into firn and eventually ice and provides much less than+10 m yrminus1 of elevation change through time so +5 m yrminus1

is chosen as a reasonable upper limit on the maximum thick-ening expected over the large areas covered in this studyWe note that a point measurement may yield adh

dthigher

than +5 m yrminus1 but we would not expect a single point to berepresentative of sustained thickening rates over square kilo-meters Figure2 highlights the importance of applying thiscutoff cloud-influenced elevations would seriously degradethe quality of our regressions for each pixel Tests using ahigher positive cutoff produced discontinuous and incoher-ent ldquosplotchesrdquo of extreme positivedh

dtthat are unrealistic

The only exception is the lower ablation zone of GaribaldiGlacier where we allow a maximum positive deviation of+10 m yrminus1 to accommodate the known advance of Garibaldi(Fig 3)

The sub-panels in Fig2 illustrate the problem of using apositive cutoff of +30 m yrminus1 which would lead to the inclu-sion of ASTER elevations that are obvious outliers ASTERelevations in the accumulation zone are generally less reli-able due to greater cloud and snow cover (lack of contrast)

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

828 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 3 Frontal variation history of Garibaldi Glacier from Land-sat TM ASTER and aerial photographs The background is a15012011 ASTER image

Errors due to clouds tend to be positive rather than nega-tive this is confirmed by comparing the average off-icedh

dtof increasing symmetric maximum allowed deviations Al-lowing plusmn5 plusmn10 andplusmn30 m yrminus1 results in averagedh

dtof

minus004 m yrminus1 +010 m yrminus1 and +082 m yrminus1 Allowing adeviation ofminus30 m yrminus1 is necessary to capture maximumthinning but 330 m of thickening (+30 m yrminus1 from 2000 to2011) is unreasonable in both the ablation and accumulationzone for this region (Koppes et al 2009 Lopez et al 2010)

In situ measurements of accumulation rate on the CDI arerequired to refine our estimates further the cutoffs we useare the best available based on the literature (egFernan-dez et al 2011) To assess the uncertainty due to the choiceof maximum deviation allowed we find the mode (most fre-quent occurrence ie peak) of the distribution of elevationdifferences between all ASTER DEMs and the SRTM DEM(normalized by dividing by the time interval between eachASTER DEM and the SRTM DEM) The mode is not af-fected by the choice of allowed deviation and so is an inde-pendent measure that we can compare our regression-derivedrates to

We do not present the ldquomode raterdquo as representing thebest estimate of the averagedh

dt It is approximately what the

rate would be if the distribution were Gaussian around thepeak dh

dt This implicitly assumes that rates that are equidis-

tant from the mode rate are equally likely (eg if the moderate isminus15 m yrminus1) then rates of +185 m yrminus1 are approx-imately as common as rates ofminus205 m yrminus1 We consider

1800 m

Data (variance of o-ice dhdt atcorresponding lengthscale on x-axis)

Fit

Fig 4 A plot of pixel size (listed as side length of square pixel)vs variance of bedrockdh

dt(calculated using aminus10+10 m yrminus1 cut-

off) plotted as red dots A linear fit to log(x)y shows the trend ofthe data (blue line) The black circle indicates the point at whichthe curve is ldquoflatteningrdquo ie the lengthscale past which the vari-ance no longer changes significantly indicating that thedh

dtare no

longer correlated 1800 m is selected as a conservative estimate ofthe decorrelation length for thedh

dt

this unlikely given the coherent thinning of approximatelyminus25 m yrminus1 estimated at the front of several glaciers and thelack of evidence for a similar degree of coherent thickeninganywhere on the icefield Therefore the mode rate is proba-bly lower than the the actual thinning rate and the differencebetween the mode rate and our regression-deriveddh

dtis an

overestimate of the maximum ldquobiasrdquo due to our choice ofdhdt

We add the difference (plusmn122 Gt yrminus1) that results from

using the same density and penetration assumptions as ourregression rate to our uncertainties to fully account for anypossible bias

We assume all volume change is at a density of900 kg mminus3 consistent with previous studies (egRignotet al 2003 Berthier et al 2004 2010) However the densityof the lost material is likely variable We consider two dif-ferent density scenarios Assuming that firn (600 kg mminus3) islost in the accumulation zone and ice (900 kg mminus3) is lost inthe ablation zone (considered as a possible scenario byKaabet al 2012) reduces the mass change rate by 03 Gt yrminus1Assuming an ice density of 900plusmn 25 kg mminus3 (egGardneret al 2012) addsplusmn01 Gt yrminus1 to our uncertainty The ef-fect of density changes are small (sim 10 ) as most of themass loss occurs from ice in the ablation zone We incorpo-rate the 600900 kg m3 scenario into our uncertainties as itadds a furtherplusmn03 Gt yrminus1 and is thus a more conservativeestimate of uncertainty

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 825

2 Methods

21 Data Preparation

The Advanced Spaceborne Thermal Emission and ReflectionRadiometer (ASTER) has a stereo-imaging capability en-abling DEMs to be generated on-demand by NASArsquos LandProcesses Distributed Active Archive Center (LP DAAC)(Fujisada et al 2005) In this study ASTER DEMs (prod-uct 14) are used to calculatedh

dt while band 3N images

(product 1B) are used for pixel-tracking Clouds are notmasked during data preparation rather they are largely re-moved by excluding elevations based on deviation from thethe SRTM DEM (Sect22) NASArsquos Automatic Registrationand Orthorectification Package (AROPGao et al 2009) isused to co-register ASTER images and DEMs to a LandsatGLS image (available from the Global Land Cover Facility)and orthorectify the ASTER L1B images using the Shut-tle Radar Topography Mission (SRTM) DEM (acquired in2000) Landsat GLS images are orthorectified to the SRTMDEM (Tucker et al 2004) therefore co-registering theASTER imagery to the Landsat GLS image effectively co-registers them to the SRTM DEM (seeWillis et al 2012afordetails) The SRTM DEM is our default DEM for orthorec-tifying the ASTER imagery it is not as adversely affectedby clouds snow and other features that cause errors in op-tically derived (eg ASTER) DEMs (Scherler et al 2008)All ASTER images used for pixel-tracking are orthorectifiedto the SRTM DEM except for the 07092001 to 25092001and 06092003 to 13092003 pairs covering the front ofMarinelli Glacier where pronounced thinning necessitatesthe use of the corresponding ASTER DEMs These pairsare orthorectified to the 25092001 and 13092003 ASTERDEMs respectively

22 Elevation change rates

Horizontally co-registered ASTER DEMs are subsequentlyvertically co-registered and a weighted linear regression isapplied to calculatedh

dtfor each pixel In all 36 ASTER

DEMs (derived from imagery acquired from 2001 to 2011)and the SRTM DEM (acquired in February 2000) are pro-cessed with an average of 4ndash5 elevations per pixel incor-porated into the regression (Fig2 shows elevations and re-gression lines for several points on Marinelli Glacier) Eachelevation is weighted by the inverse of the standard devia-tion of the bedrock elevation differences between its ASTERDEM and the SRTM DEM This is a common measure ofthe uncertainty associated with ASTER DEM elevations (Fu-jisada et al 2005 Howat et al 2008) Horizontal misalign-ment will appear as off-ice (ldquobedrockrdquo) elevation differenceswhen comparing ASTER DEMs to the SRTM DEM and istherefore included in the standard deviation of the bedrockelevation differences (ie the greater the misalignment thegreater the standard deviation which is the measure of un-

certainty that we use for each ASTER DEM) We typicallyfind values from 8 m to 20 m as our uncertainty for ASTERDEMs similar to other studies (Kaab 2002 Fujisada et al2005 San and Suzen 2005 Rivera et al 2007 Howat et al2008 Berthier et al 2010)

In the ablation zone we exclude elevations from the re-gression that deviate more than +5minus30 m yrminus1 from thefirst elevation in the accumulation zone the cutoff is+5minus10 m yrminus1 These cutoffs largely remove the areas oferroneously high elevations due to clouds as well as out-liers caused by shadow snow and other sources which eitherobscure the ice or are largely featureless at ASTER resolu-tion It is important to note that although these cutoffs arebased on expecteddh

dt they are imposed on each individual

map point (see Fig2) before applying the weighted linearregression rather than to already calculateddh

dt which would

yield drastically different results The first elevation in ourtime series is SRTM for 94 of the pixels for the remain-ing areas we use the first available elevation from ASTERThe average end date is 17052008 The high percentageof points with SRTM elevations makes our cutoff strategyfeasible The radar-derived SRTM DEM is not influencedby clouds like ASTER making SRTM elevations a reliablestarting point

We limit the maximum negative deviation tominus30 m yrminus1

in the ablation zone to just capture the maximum thinningand exclude spurious elevations from the regression calcula-tions Changing this tominus60 m yrminus1 has no affect on the zonesof maximum thinning Less thinning is expected in the accu-mulation zone than in the ablation zone therefore we allowa maximum negative deviation of onlyminus10 m yrminus1 above theELA Permitting a greater negative deviation in the accumu-lation zone only results in incoherent areas of extreme andunrealistic thinning

In order to determine the ELA and distinguish where toimpose our different constraints we examine late season(January and February) ASTER images We record the al-titude of the transition between bare ice and snow and takethis as a rough proxy for the ELA (egBamber and Rivera2007) We are able to measure the ELA for 11 larger glaciersfor the rest of the icefield we assume a regional ELA of1090 m (Strelin and Iturraspe 2007)

The volume change rate at each pixel is thedhdt

for the pixelmultiplied by the area of the pixel Summing together the vol-ume change rate of every pixel yields a volume change rate(dV

dt) for the entire icefielddV

dtis multiplied by the density

of glacier ice set to 900 kg mminus3 (egCuffey and Paterson2010) to produce an estimate of the mass change rate Futureground-based measurements of densities on the CDI will beneeded to find more accurate and precise values

We have a sufficient number of ASTER DEMs to providedhdt

rates for 96 of the total area of the icefield Each pixelin the remaining small gaps is filled with the mediandh

dtvalue

within 1 km This local method is more accurate than using

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

826 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

0

100

200

300

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

1 dhdt=minus2648 myr

Excluded

dhdt

300

400

500

600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

2 dhdt=minus1389 myr

400

500

600

700

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

3 dhdt=minus1163 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

4 dhdt=minus937 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

5 dhdt=minus763 myr

900

1000

1100

1200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

6 dhdt=minus083 myr

1300

1400

1500

1600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

7 dhdt=065 myr

1600

1700

1800

1900

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

8 dhdt=426 myr

1900

2000

2100

2200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

9 dhdt=minus023 myr

2100

2200

2300

2400

2500

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

10 dhdt=248 myr

69˚38W

54˚38S

54˚30S

minus20

minus15

minus10

minus5

0

5dhdt

1

2

3

4

5

6

78

9

10

Fig 2 Elevation values anddhdt

for randomly selected pixels over Marinelli Glacier The left-most elevation in each graph is the SRTM

elevation at that pixel Blue lines indicate thedhdt

calculated for each pixel elevation points bolded red are excluded fromdhdt

calculation The

bottom right panel shows thedhdt

map with numbered circles indicating the location corresponding to each graph

the averagedhdt

of an elevation bin to fill gaps within that binor filling in gaps with the mean of alldh

dtcalculated (egRig-

not et al 2003) For example the average low-elevationdhdt

and overall averagedhdt

are both negative these rates wouldnot be suitable for filling a gap near the front of GaribaldiGlacier which is known to be advancing and where we mea-sure positivedh

dtat low elevations

Penetration of C-band radar into ice and (particularly)snow (egRignot et al 2001) is a potential problem whenusing the SRTM DEM We use a technique pioneered byGardelle et al(2012) and applied byWillis et al (2012b)to compensate for penetration effectsWillis et al (2012b)compared X-band SRTM elevations (which have negligi-ble penetration) with C-band SRTM elevations and found

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 827

approximately 2 m of C-band penetration over the SPI at allelevations Due to a lack of X-band SRTM coverage overthe CDI we cannot provide a similar analysis here Insteadwe assume the penetration depth is similar to that at SPI andadd 2 m to each SRTM elevation over ice which increasesour mass loss rate by about 13 We note the CDI is colderthan the SPI which could lead to drier conditions and greaterpenetration (Rignot et al 2001) but we do not have adequateX-band data to quantify the difference

An additional source of mass change that we consider issub-aqueous mass lossgain (which does not contribute to sealevel rise) We cannot directly measure sub-aqueous ice gainsor losses but make a rough estimate by measuring changesin the areal extent at the front of several glaciers that have un-dergone relatively large advance or retreat (from image pairs)and multiplying these area changes by the approximate depthbelow water of the ice We assume an average depth belowwater of 150 m for Marinelli Glacier (seeKoppes et al 2009Fig 4a and b) and 60 m for the other glaciers with an uncer-tainty of plusmn50 m (the uncertainty on the change in area ofeach glacier is negligible) Dividing the sub-aqueous volumechange by the time interval separating the images used tofind area change gives a rate This calculation shows that sub-aqueous mass loss while not well constrained in this studyis an order of magnitude lower than the overall mass changerate calculated fromdh

dt(lt 5 )

221 Sources of uncertainty

Below we consider sources of uncertainty on our masschange rate These include the uncertainty on the elevationsincoporated into the regression uncertainty on the ELA theeffect of varying the maximum deviation allowed from thefirst elevation different density scenarios and uncertainty onthe penetration depth of the C-band SRTM DEM

The uncertainty associated with thedhdt

for each pixel iscalculated from the model covariance matrix (egAsteret al 2005) which accounts for the uncertainties on the el-evations incorporated into the regression The 95 confi-dence interval for the volume change rate uncertainty is cal-culated using the formula 196times U

radicN

U is the total ldquovolumerdquo

of uncertainty calculated by taking the uncertainty at eachpixel multiplying it by the area of the pixel (to determinea ldquovolumerdquo of uncertainty for that pixel) and then addingtogether ldquovolumerdquo of uncertainty for each pixel where adh

dtis calculatedN is the number of independent pixels (egHowat et al 2008) which we determine by dividing the to-tal area by the area over which off-icedh

dtare correlated (eg

Rolstad et al 2009) We estimate the scale at which thedhdt

are independent by finding the area at which the variance ofthe off-icedh

dtbegins to ldquoflattenrdquo (seeRolstad et al 2009and

Willis et al 2012afor details on the method) which we es-timate to be 1800 m by 1800 m (Fig4) This is analogousto the ldquocornerrdquo point on anL curve (egAster et al 2005p 91 Fig 52) and indicates the lengthscale past which the

dhdt

are no longer correlated The total contribution from theuncertainty on individual elevations issim 035 Gt yrminus1

The regional ELA is poorly known In order to investigatethe impact of changing the regional ELA on the mass lossrates we lower our regional ELA from 1090 to 650 m anELA that has been found for several glaciers on the southernand western regions of the CDI (Bown et al 2013) We alsoassume an uncertainty ofplusmn200 m on the 11 glaciers wherewe have estimated the ELA from optical imagery The dif-ference between the two ELA scenarios is incorporated intoour uncertainties (sim 037 Gt yrminus1)

The deviation allowed from the first elevation has a largeimpact on the mass loss rate (egWillis et al 2012b) Therate produced by allowing +10 m yrminus1 is contaminated bylow clouds but is given as a rough minimum estimate ofthe mass loss rate The unsymmetric cutoff (+5 m yrminus1 ver-susminus10minus30 m yrminus1) may bias our results towards thinningbut we argue that an unsymmetric cutoff is more physicallyjustified than a symmetric cutoff Rapid retreat has been in-dependently observed at Marinelli Glacier (Holmlund andFuenzalida 1995 Koppes et al 2009) from which we knowthat a large amount of thinning must be occurring The max-imum allowed negative deviation from the first elevation ofminus30 m yrminus1 just captures the maximum thinning at MarinelliGlacier (see Fig2 point 1)

The cutoff of +5 m yrminus1 from the first elevation is basedon the limited precipitation data available for this region Theupper limit on precipitation in Patagonia is about 10 m yrminus1

(egHolmlund and Fuenzalida 1995 Rignot et al 2003Rasmussen et al 2007 Koppes et al 2011) and PRECISmodel results show maximum precipitation in the 4000 to6000 mm yrminus1 range for the CDI (Fernandez et al 2011)One year of accumulated precipitation undergoes densifica-tion into firn and eventually ice and provides much less than+10 m yrminus1 of elevation change through time so +5 m yrminus1

is chosen as a reasonable upper limit on the maximum thick-ening expected over the large areas covered in this studyWe note that a point measurement may yield adh

dthigher

than +5 m yrminus1 but we would not expect a single point to berepresentative of sustained thickening rates over square kilo-meters Figure2 highlights the importance of applying thiscutoff cloud-influenced elevations would seriously degradethe quality of our regressions for each pixel Tests using ahigher positive cutoff produced discontinuous and incoher-ent ldquosplotchesrdquo of extreme positivedh

dtthat are unrealistic

The only exception is the lower ablation zone of GaribaldiGlacier where we allow a maximum positive deviation of+10 m yrminus1 to accommodate the known advance of Garibaldi(Fig 3)

The sub-panels in Fig2 illustrate the problem of using apositive cutoff of +30 m yrminus1 which would lead to the inclu-sion of ASTER elevations that are obvious outliers ASTERelevations in the accumulation zone are generally less reli-able due to greater cloud and snow cover (lack of contrast)

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

828 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 3 Frontal variation history of Garibaldi Glacier from Land-sat TM ASTER and aerial photographs The background is a15012011 ASTER image

Errors due to clouds tend to be positive rather than nega-tive this is confirmed by comparing the average off-icedh

dtof increasing symmetric maximum allowed deviations Al-lowing plusmn5 plusmn10 andplusmn30 m yrminus1 results in averagedh

dtof

minus004 m yrminus1 +010 m yrminus1 and +082 m yrminus1 Allowing adeviation ofminus30 m yrminus1 is necessary to capture maximumthinning but 330 m of thickening (+30 m yrminus1 from 2000 to2011) is unreasonable in both the ablation and accumulationzone for this region (Koppes et al 2009 Lopez et al 2010)

In situ measurements of accumulation rate on the CDI arerequired to refine our estimates further the cutoffs we useare the best available based on the literature (egFernan-dez et al 2011) To assess the uncertainty due to the choiceof maximum deviation allowed we find the mode (most fre-quent occurrence ie peak) of the distribution of elevationdifferences between all ASTER DEMs and the SRTM DEM(normalized by dividing by the time interval between eachASTER DEM and the SRTM DEM) The mode is not af-fected by the choice of allowed deviation and so is an inde-pendent measure that we can compare our regression-derivedrates to

We do not present the ldquomode raterdquo as representing thebest estimate of the averagedh

dt It is approximately what the

rate would be if the distribution were Gaussian around thepeak dh

dt This implicitly assumes that rates that are equidis-

tant from the mode rate are equally likely (eg if the moderate isminus15 m yrminus1) then rates of +185 m yrminus1 are approx-imately as common as rates ofminus205 m yrminus1 We consider

1800 m

Data (variance of o-ice dhdt atcorresponding lengthscale on x-axis)

Fit

Fig 4 A plot of pixel size (listed as side length of square pixel)vs variance of bedrockdh

dt(calculated using aminus10+10 m yrminus1 cut-

off) plotted as red dots A linear fit to log(x)y shows the trend ofthe data (blue line) The black circle indicates the point at whichthe curve is ldquoflatteningrdquo ie the lengthscale past which the vari-ance no longer changes significantly indicating that thedh

dtare no

longer correlated 1800 m is selected as a conservative estimate ofthe decorrelation length for thedh

dt

this unlikely given the coherent thinning of approximatelyminus25 m yrminus1 estimated at the front of several glaciers and thelack of evidence for a similar degree of coherent thickeninganywhere on the icefield Therefore the mode rate is proba-bly lower than the the actual thinning rate and the differencebetween the mode rate and our regression-deriveddh

dtis an

overestimate of the maximum ldquobiasrdquo due to our choice ofdhdt

We add the difference (plusmn122 Gt yrminus1) that results from

using the same density and penetration assumptions as ourregression rate to our uncertainties to fully account for anypossible bias

We assume all volume change is at a density of900 kg mminus3 consistent with previous studies (egRignotet al 2003 Berthier et al 2004 2010) However the densityof the lost material is likely variable We consider two dif-ferent density scenarios Assuming that firn (600 kg mminus3) islost in the accumulation zone and ice (900 kg mminus3) is lost inthe ablation zone (considered as a possible scenario byKaabet al 2012) reduces the mass change rate by 03 Gt yrminus1Assuming an ice density of 900plusmn 25 kg mminus3 (egGardneret al 2012) addsplusmn01 Gt yrminus1 to our uncertainty The ef-fect of density changes are small (sim 10 ) as most of themass loss occurs from ice in the ablation zone We incorpo-rate the 600900 kg m3 scenario into our uncertainties as itadds a furtherplusmn03 Gt yrminus1 and is thus a more conservativeestimate of uncertainty

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

826 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

0

100

200

300

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

1 dhdt=minus2648 myr

Excluded

dhdt

300

400

500

600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

2 dhdt=minus1389 myr

400

500

600

700

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

3 dhdt=minus1163 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

4 dhdt=minus937 myr

500

600

700

800

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

5 dhdt=minus763 myr

900

1000

1100

1200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

6 dhdt=minus083 myr

1300

1400

1500

1600

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

7 dhdt=065 myr

1600

1700

1800

1900

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

8 dhdt=426 myr

1900

2000

2100

2200

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

9 dhdt=minus023 myr

2100

2200

2300

2400

2500

Ele

va

tio

n (

m)

00 02 04 06 08 10 12

Year

10 dhdt=248 myr

69˚38W

54˚38S

54˚30S

minus20

minus15

minus10

minus5

0

5dhdt

1

2

3

4

5

6

78

9

10

Fig 2 Elevation values anddhdt

for randomly selected pixels over Marinelli Glacier The left-most elevation in each graph is the SRTM

elevation at that pixel Blue lines indicate thedhdt

calculated for each pixel elevation points bolded red are excluded fromdhdt

calculation The

bottom right panel shows thedhdt

map with numbered circles indicating the location corresponding to each graph

the averagedhdt

of an elevation bin to fill gaps within that binor filling in gaps with the mean of alldh

dtcalculated (egRig-

not et al 2003) For example the average low-elevationdhdt

and overall averagedhdt

are both negative these rates wouldnot be suitable for filling a gap near the front of GaribaldiGlacier which is known to be advancing and where we mea-sure positivedh

dtat low elevations

Penetration of C-band radar into ice and (particularly)snow (egRignot et al 2001) is a potential problem whenusing the SRTM DEM We use a technique pioneered byGardelle et al(2012) and applied byWillis et al (2012b)to compensate for penetration effectsWillis et al (2012b)compared X-band SRTM elevations (which have negligi-ble penetration) with C-band SRTM elevations and found

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 827

approximately 2 m of C-band penetration over the SPI at allelevations Due to a lack of X-band SRTM coverage overthe CDI we cannot provide a similar analysis here Insteadwe assume the penetration depth is similar to that at SPI andadd 2 m to each SRTM elevation over ice which increasesour mass loss rate by about 13 We note the CDI is colderthan the SPI which could lead to drier conditions and greaterpenetration (Rignot et al 2001) but we do not have adequateX-band data to quantify the difference

An additional source of mass change that we consider issub-aqueous mass lossgain (which does not contribute to sealevel rise) We cannot directly measure sub-aqueous ice gainsor losses but make a rough estimate by measuring changesin the areal extent at the front of several glaciers that have un-dergone relatively large advance or retreat (from image pairs)and multiplying these area changes by the approximate depthbelow water of the ice We assume an average depth belowwater of 150 m for Marinelli Glacier (seeKoppes et al 2009Fig 4a and b) and 60 m for the other glaciers with an uncer-tainty of plusmn50 m (the uncertainty on the change in area ofeach glacier is negligible) Dividing the sub-aqueous volumechange by the time interval separating the images used tofind area change gives a rate This calculation shows that sub-aqueous mass loss while not well constrained in this studyis an order of magnitude lower than the overall mass changerate calculated fromdh

dt(lt 5 )

221 Sources of uncertainty

Below we consider sources of uncertainty on our masschange rate These include the uncertainty on the elevationsincoporated into the regression uncertainty on the ELA theeffect of varying the maximum deviation allowed from thefirst elevation different density scenarios and uncertainty onthe penetration depth of the C-band SRTM DEM

The uncertainty associated with thedhdt

for each pixel iscalculated from the model covariance matrix (egAsteret al 2005) which accounts for the uncertainties on the el-evations incorporated into the regression The 95 confi-dence interval for the volume change rate uncertainty is cal-culated using the formula 196times U

radicN

U is the total ldquovolumerdquo

of uncertainty calculated by taking the uncertainty at eachpixel multiplying it by the area of the pixel (to determinea ldquovolumerdquo of uncertainty for that pixel) and then addingtogether ldquovolumerdquo of uncertainty for each pixel where adh

dtis calculatedN is the number of independent pixels (egHowat et al 2008) which we determine by dividing the to-tal area by the area over which off-icedh

dtare correlated (eg

Rolstad et al 2009) We estimate the scale at which thedhdt

are independent by finding the area at which the variance ofthe off-icedh

dtbegins to ldquoflattenrdquo (seeRolstad et al 2009and

Willis et al 2012afor details on the method) which we es-timate to be 1800 m by 1800 m (Fig4) This is analogousto the ldquocornerrdquo point on anL curve (egAster et al 2005p 91 Fig 52) and indicates the lengthscale past which the

dhdt

are no longer correlated The total contribution from theuncertainty on individual elevations issim 035 Gt yrminus1

The regional ELA is poorly known In order to investigatethe impact of changing the regional ELA on the mass lossrates we lower our regional ELA from 1090 to 650 m anELA that has been found for several glaciers on the southernand western regions of the CDI (Bown et al 2013) We alsoassume an uncertainty ofplusmn200 m on the 11 glaciers wherewe have estimated the ELA from optical imagery The dif-ference between the two ELA scenarios is incorporated intoour uncertainties (sim 037 Gt yrminus1)

The deviation allowed from the first elevation has a largeimpact on the mass loss rate (egWillis et al 2012b) Therate produced by allowing +10 m yrminus1 is contaminated bylow clouds but is given as a rough minimum estimate ofthe mass loss rate The unsymmetric cutoff (+5 m yrminus1 ver-susminus10minus30 m yrminus1) may bias our results towards thinningbut we argue that an unsymmetric cutoff is more physicallyjustified than a symmetric cutoff Rapid retreat has been in-dependently observed at Marinelli Glacier (Holmlund andFuenzalida 1995 Koppes et al 2009) from which we knowthat a large amount of thinning must be occurring The max-imum allowed negative deviation from the first elevation ofminus30 m yrminus1 just captures the maximum thinning at MarinelliGlacier (see Fig2 point 1)

The cutoff of +5 m yrminus1 from the first elevation is basedon the limited precipitation data available for this region Theupper limit on precipitation in Patagonia is about 10 m yrminus1

(egHolmlund and Fuenzalida 1995 Rignot et al 2003Rasmussen et al 2007 Koppes et al 2011) and PRECISmodel results show maximum precipitation in the 4000 to6000 mm yrminus1 range for the CDI (Fernandez et al 2011)One year of accumulated precipitation undergoes densifica-tion into firn and eventually ice and provides much less than+10 m yrminus1 of elevation change through time so +5 m yrminus1

is chosen as a reasonable upper limit on the maximum thick-ening expected over the large areas covered in this studyWe note that a point measurement may yield adh

dthigher

than +5 m yrminus1 but we would not expect a single point to berepresentative of sustained thickening rates over square kilo-meters Figure2 highlights the importance of applying thiscutoff cloud-influenced elevations would seriously degradethe quality of our regressions for each pixel Tests using ahigher positive cutoff produced discontinuous and incoher-ent ldquosplotchesrdquo of extreme positivedh

dtthat are unrealistic

The only exception is the lower ablation zone of GaribaldiGlacier where we allow a maximum positive deviation of+10 m yrminus1 to accommodate the known advance of Garibaldi(Fig 3)

The sub-panels in Fig2 illustrate the problem of using apositive cutoff of +30 m yrminus1 which would lead to the inclu-sion of ASTER elevations that are obvious outliers ASTERelevations in the accumulation zone are generally less reli-able due to greater cloud and snow cover (lack of contrast)

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

828 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 3 Frontal variation history of Garibaldi Glacier from Land-sat TM ASTER and aerial photographs The background is a15012011 ASTER image

Errors due to clouds tend to be positive rather than nega-tive this is confirmed by comparing the average off-icedh

dtof increasing symmetric maximum allowed deviations Al-lowing plusmn5 plusmn10 andplusmn30 m yrminus1 results in averagedh

dtof

minus004 m yrminus1 +010 m yrminus1 and +082 m yrminus1 Allowing adeviation ofminus30 m yrminus1 is necessary to capture maximumthinning but 330 m of thickening (+30 m yrminus1 from 2000 to2011) is unreasonable in both the ablation and accumulationzone for this region (Koppes et al 2009 Lopez et al 2010)

In situ measurements of accumulation rate on the CDI arerequired to refine our estimates further the cutoffs we useare the best available based on the literature (egFernan-dez et al 2011) To assess the uncertainty due to the choiceof maximum deviation allowed we find the mode (most fre-quent occurrence ie peak) of the distribution of elevationdifferences between all ASTER DEMs and the SRTM DEM(normalized by dividing by the time interval between eachASTER DEM and the SRTM DEM) The mode is not af-fected by the choice of allowed deviation and so is an inde-pendent measure that we can compare our regression-derivedrates to

We do not present the ldquomode raterdquo as representing thebest estimate of the averagedh

dt It is approximately what the

rate would be if the distribution were Gaussian around thepeak dh

dt This implicitly assumes that rates that are equidis-

tant from the mode rate are equally likely (eg if the moderate isminus15 m yrminus1) then rates of +185 m yrminus1 are approx-imately as common as rates ofminus205 m yrminus1 We consider

1800 m

Data (variance of o-ice dhdt atcorresponding lengthscale on x-axis)

Fit

Fig 4 A plot of pixel size (listed as side length of square pixel)vs variance of bedrockdh

dt(calculated using aminus10+10 m yrminus1 cut-

off) plotted as red dots A linear fit to log(x)y shows the trend ofthe data (blue line) The black circle indicates the point at whichthe curve is ldquoflatteningrdquo ie the lengthscale past which the vari-ance no longer changes significantly indicating that thedh

dtare no

longer correlated 1800 m is selected as a conservative estimate ofthe decorrelation length for thedh

dt

this unlikely given the coherent thinning of approximatelyminus25 m yrminus1 estimated at the front of several glaciers and thelack of evidence for a similar degree of coherent thickeninganywhere on the icefield Therefore the mode rate is proba-bly lower than the the actual thinning rate and the differencebetween the mode rate and our regression-deriveddh

dtis an

overestimate of the maximum ldquobiasrdquo due to our choice ofdhdt

We add the difference (plusmn122 Gt yrminus1) that results from

using the same density and penetration assumptions as ourregression rate to our uncertainties to fully account for anypossible bias

We assume all volume change is at a density of900 kg mminus3 consistent with previous studies (egRignotet al 2003 Berthier et al 2004 2010) However the densityof the lost material is likely variable We consider two dif-ferent density scenarios Assuming that firn (600 kg mminus3) islost in the accumulation zone and ice (900 kg mminus3) is lost inthe ablation zone (considered as a possible scenario byKaabet al 2012) reduces the mass change rate by 03 Gt yrminus1Assuming an ice density of 900plusmn 25 kg mminus3 (egGardneret al 2012) addsplusmn01 Gt yrminus1 to our uncertainty The ef-fect of density changes are small (sim 10 ) as most of themass loss occurs from ice in the ablation zone We incorpo-rate the 600900 kg m3 scenario into our uncertainties as itadds a furtherplusmn03 Gt yrminus1 and is thus a more conservativeestimate of uncertainty

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 827

approximately 2 m of C-band penetration over the SPI at allelevations Due to a lack of X-band SRTM coverage overthe CDI we cannot provide a similar analysis here Insteadwe assume the penetration depth is similar to that at SPI andadd 2 m to each SRTM elevation over ice which increasesour mass loss rate by about 13 We note the CDI is colderthan the SPI which could lead to drier conditions and greaterpenetration (Rignot et al 2001) but we do not have adequateX-band data to quantify the difference

An additional source of mass change that we consider issub-aqueous mass lossgain (which does not contribute to sealevel rise) We cannot directly measure sub-aqueous ice gainsor losses but make a rough estimate by measuring changesin the areal extent at the front of several glaciers that have un-dergone relatively large advance or retreat (from image pairs)and multiplying these area changes by the approximate depthbelow water of the ice We assume an average depth belowwater of 150 m for Marinelli Glacier (seeKoppes et al 2009Fig 4a and b) and 60 m for the other glaciers with an uncer-tainty of plusmn50 m (the uncertainty on the change in area ofeach glacier is negligible) Dividing the sub-aqueous volumechange by the time interval separating the images used tofind area change gives a rate This calculation shows that sub-aqueous mass loss while not well constrained in this studyis an order of magnitude lower than the overall mass changerate calculated fromdh

dt(lt 5 )

221 Sources of uncertainty

Below we consider sources of uncertainty on our masschange rate These include the uncertainty on the elevationsincoporated into the regression uncertainty on the ELA theeffect of varying the maximum deviation allowed from thefirst elevation different density scenarios and uncertainty onthe penetration depth of the C-band SRTM DEM

The uncertainty associated with thedhdt

for each pixel iscalculated from the model covariance matrix (egAsteret al 2005) which accounts for the uncertainties on the el-evations incorporated into the regression The 95 confi-dence interval for the volume change rate uncertainty is cal-culated using the formula 196times U

radicN

U is the total ldquovolumerdquo

of uncertainty calculated by taking the uncertainty at eachpixel multiplying it by the area of the pixel (to determinea ldquovolumerdquo of uncertainty for that pixel) and then addingtogether ldquovolumerdquo of uncertainty for each pixel where adh

dtis calculatedN is the number of independent pixels (egHowat et al 2008) which we determine by dividing the to-tal area by the area over which off-icedh

dtare correlated (eg

Rolstad et al 2009) We estimate the scale at which thedhdt

are independent by finding the area at which the variance ofthe off-icedh

dtbegins to ldquoflattenrdquo (seeRolstad et al 2009and

Willis et al 2012afor details on the method) which we es-timate to be 1800 m by 1800 m (Fig4) This is analogousto the ldquocornerrdquo point on anL curve (egAster et al 2005p 91 Fig 52) and indicates the lengthscale past which the

dhdt

are no longer correlated The total contribution from theuncertainty on individual elevations issim 035 Gt yrminus1

The regional ELA is poorly known In order to investigatethe impact of changing the regional ELA on the mass lossrates we lower our regional ELA from 1090 to 650 m anELA that has been found for several glaciers on the southernand western regions of the CDI (Bown et al 2013) We alsoassume an uncertainty ofplusmn200 m on the 11 glaciers wherewe have estimated the ELA from optical imagery The dif-ference between the two ELA scenarios is incorporated intoour uncertainties (sim 037 Gt yrminus1)

The deviation allowed from the first elevation has a largeimpact on the mass loss rate (egWillis et al 2012b) Therate produced by allowing +10 m yrminus1 is contaminated bylow clouds but is given as a rough minimum estimate ofthe mass loss rate The unsymmetric cutoff (+5 m yrminus1 ver-susminus10minus30 m yrminus1) may bias our results towards thinningbut we argue that an unsymmetric cutoff is more physicallyjustified than a symmetric cutoff Rapid retreat has been in-dependently observed at Marinelli Glacier (Holmlund andFuenzalida 1995 Koppes et al 2009) from which we knowthat a large amount of thinning must be occurring The max-imum allowed negative deviation from the first elevation ofminus30 m yrminus1 just captures the maximum thinning at MarinelliGlacier (see Fig2 point 1)

The cutoff of +5 m yrminus1 from the first elevation is basedon the limited precipitation data available for this region Theupper limit on precipitation in Patagonia is about 10 m yrminus1

(egHolmlund and Fuenzalida 1995 Rignot et al 2003Rasmussen et al 2007 Koppes et al 2011) and PRECISmodel results show maximum precipitation in the 4000 to6000 mm yrminus1 range for the CDI (Fernandez et al 2011)One year of accumulated precipitation undergoes densifica-tion into firn and eventually ice and provides much less than+10 m yrminus1 of elevation change through time so +5 m yrminus1

is chosen as a reasonable upper limit on the maximum thick-ening expected over the large areas covered in this studyWe note that a point measurement may yield adh

dthigher

than +5 m yrminus1 but we would not expect a single point to berepresentative of sustained thickening rates over square kilo-meters Figure2 highlights the importance of applying thiscutoff cloud-influenced elevations would seriously degradethe quality of our regressions for each pixel Tests using ahigher positive cutoff produced discontinuous and incoher-ent ldquosplotchesrdquo of extreme positivedh

dtthat are unrealistic

The only exception is the lower ablation zone of GaribaldiGlacier where we allow a maximum positive deviation of+10 m yrminus1 to accommodate the known advance of Garibaldi(Fig 3)

The sub-panels in Fig2 illustrate the problem of using apositive cutoff of +30 m yrminus1 which would lead to the inclu-sion of ASTER elevations that are obvious outliers ASTERelevations in the accumulation zone are generally less reli-able due to greater cloud and snow cover (lack of contrast)

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

828 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 3 Frontal variation history of Garibaldi Glacier from Land-sat TM ASTER and aerial photographs The background is a15012011 ASTER image

Errors due to clouds tend to be positive rather than nega-tive this is confirmed by comparing the average off-icedh

dtof increasing symmetric maximum allowed deviations Al-lowing plusmn5 plusmn10 andplusmn30 m yrminus1 results in averagedh

dtof

minus004 m yrminus1 +010 m yrminus1 and +082 m yrminus1 Allowing adeviation ofminus30 m yrminus1 is necessary to capture maximumthinning but 330 m of thickening (+30 m yrminus1 from 2000 to2011) is unreasonable in both the ablation and accumulationzone for this region (Koppes et al 2009 Lopez et al 2010)

In situ measurements of accumulation rate on the CDI arerequired to refine our estimates further the cutoffs we useare the best available based on the literature (egFernan-dez et al 2011) To assess the uncertainty due to the choiceof maximum deviation allowed we find the mode (most fre-quent occurrence ie peak) of the distribution of elevationdifferences between all ASTER DEMs and the SRTM DEM(normalized by dividing by the time interval between eachASTER DEM and the SRTM DEM) The mode is not af-fected by the choice of allowed deviation and so is an inde-pendent measure that we can compare our regression-derivedrates to

We do not present the ldquomode raterdquo as representing thebest estimate of the averagedh

dt It is approximately what the

rate would be if the distribution were Gaussian around thepeak dh

dt This implicitly assumes that rates that are equidis-

tant from the mode rate are equally likely (eg if the moderate isminus15 m yrminus1) then rates of +185 m yrminus1 are approx-imately as common as rates ofminus205 m yrminus1 We consider

1800 m

Data (variance of o-ice dhdt atcorresponding lengthscale on x-axis)

Fit

Fig 4 A plot of pixel size (listed as side length of square pixel)vs variance of bedrockdh

dt(calculated using aminus10+10 m yrminus1 cut-

off) plotted as red dots A linear fit to log(x)y shows the trend ofthe data (blue line) The black circle indicates the point at whichthe curve is ldquoflatteningrdquo ie the lengthscale past which the vari-ance no longer changes significantly indicating that thedh

dtare no

longer correlated 1800 m is selected as a conservative estimate ofthe decorrelation length for thedh

dt

this unlikely given the coherent thinning of approximatelyminus25 m yrminus1 estimated at the front of several glaciers and thelack of evidence for a similar degree of coherent thickeninganywhere on the icefield Therefore the mode rate is proba-bly lower than the the actual thinning rate and the differencebetween the mode rate and our regression-deriveddh

dtis an

overestimate of the maximum ldquobiasrdquo due to our choice ofdhdt

We add the difference (plusmn122 Gt yrminus1) that results from

using the same density and penetration assumptions as ourregression rate to our uncertainties to fully account for anypossible bias

We assume all volume change is at a density of900 kg mminus3 consistent with previous studies (egRignotet al 2003 Berthier et al 2004 2010) However the densityof the lost material is likely variable We consider two dif-ferent density scenarios Assuming that firn (600 kg mminus3) islost in the accumulation zone and ice (900 kg mminus3) is lost inthe ablation zone (considered as a possible scenario byKaabet al 2012) reduces the mass change rate by 03 Gt yrminus1Assuming an ice density of 900plusmn 25 kg mminus3 (egGardneret al 2012) addsplusmn01 Gt yrminus1 to our uncertainty The ef-fect of density changes are small (sim 10 ) as most of themass loss occurs from ice in the ablation zone We incorpo-rate the 600900 kg m3 scenario into our uncertainties as itadds a furtherplusmn03 Gt yrminus1 and is thus a more conservativeestimate of uncertainty

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

828 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 3 Frontal variation history of Garibaldi Glacier from Land-sat TM ASTER and aerial photographs The background is a15012011 ASTER image

Errors due to clouds tend to be positive rather than nega-tive this is confirmed by comparing the average off-icedh

dtof increasing symmetric maximum allowed deviations Al-lowing plusmn5 plusmn10 andplusmn30 m yrminus1 results in averagedh

dtof

minus004 m yrminus1 +010 m yrminus1 and +082 m yrminus1 Allowing adeviation ofminus30 m yrminus1 is necessary to capture maximumthinning but 330 m of thickening (+30 m yrminus1 from 2000 to2011) is unreasonable in both the ablation and accumulationzone for this region (Koppes et al 2009 Lopez et al 2010)

In situ measurements of accumulation rate on the CDI arerequired to refine our estimates further the cutoffs we useare the best available based on the literature (egFernan-dez et al 2011) To assess the uncertainty due to the choiceof maximum deviation allowed we find the mode (most fre-quent occurrence ie peak) of the distribution of elevationdifferences between all ASTER DEMs and the SRTM DEM(normalized by dividing by the time interval between eachASTER DEM and the SRTM DEM) The mode is not af-fected by the choice of allowed deviation and so is an inde-pendent measure that we can compare our regression-derivedrates to

We do not present the ldquomode raterdquo as representing thebest estimate of the averagedh

dt It is approximately what the

rate would be if the distribution were Gaussian around thepeak dh

dt This implicitly assumes that rates that are equidis-

tant from the mode rate are equally likely (eg if the moderate isminus15 m yrminus1) then rates of +185 m yrminus1 are approx-imately as common as rates ofminus205 m yrminus1 We consider

1800 m

Data (variance of o-ice dhdt atcorresponding lengthscale on x-axis)

Fit

Fig 4 A plot of pixel size (listed as side length of square pixel)vs variance of bedrockdh

dt(calculated using aminus10+10 m yrminus1 cut-

off) plotted as red dots A linear fit to log(x)y shows the trend ofthe data (blue line) The black circle indicates the point at whichthe curve is ldquoflatteningrdquo ie the lengthscale past which the vari-ance no longer changes significantly indicating that thedh

dtare no

longer correlated 1800 m is selected as a conservative estimate ofthe decorrelation length for thedh

dt

this unlikely given the coherent thinning of approximatelyminus25 m yrminus1 estimated at the front of several glaciers and thelack of evidence for a similar degree of coherent thickeninganywhere on the icefield Therefore the mode rate is proba-bly lower than the the actual thinning rate and the differencebetween the mode rate and our regression-deriveddh

dtis an

overestimate of the maximum ldquobiasrdquo due to our choice ofdhdt

We add the difference (plusmn122 Gt yrminus1) that results from

using the same density and penetration assumptions as ourregression rate to our uncertainties to fully account for anypossible bias

We assume all volume change is at a density of900 kg mminus3 consistent with previous studies (egRignotet al 2003 Berthier et al 2004 2010) However the densityof the lost material is likely variable We consider two dif-ferent density scenarios Assuming that firn (600 kg mminus3) islost in the accumulation zone and ice (900 kg mminus3) is lost inthe ablation zone (considered as a possible scenario byKaabet al 2012) reduces the mass change rate by 03 Gt yrminus1Assuming an ice density of 900plusmn 25 kg mminus3 (egGardneret al 2012) addsplusmn01 Gt yrminus1 to our uncertainty The ef-fect of density changes are small (sim 10 ) as most of themass loss occurs from ice in the ablation zone We incorpo-rate the 600900 kg m3 scenario into our uncertainties as itadds a furtherplusmn03 Gt yrminus1 and is thus a more conservativeestimate of uncertainty

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 829

The penetration depth of the C-band SRTM into snowand ice is an additional source of uncertainty Assuming arange ofplusmn2 for the penetration depth yields a variation of+047 Gt yrminus1 for 0 m of penetration andminus048 Gt yrminus1 for4 m We addplusmn048 Gt yrminus1 to our uncertainties to accountfor penetration depth uncertainty

Taking the square root of the sum of the squared uncertain-ties gives us an overall uncertainty ofplusmn15 Gt yrminus1 Uncer-tainties beyond the statistical uncertainty from the regressionare added as an average rate to the uncertainties for individ-ual glacier basins in Table1

23 Velocities

In this section we describe how we use data from three differ-ent satellites with different resolutions to measure sub-pixeloffsets and convert them to glacier velocities

231 ASTER

Sub-pixel offsets between ASTER image pairs (pixel reso-lution of 15 mpixel) are measured via normalized amplitudecross correlation with a spatial resolution of 120 m (ie astep size between cross correlations of 8 pixels seeWilliset al 2012afor details) This technique known as ldquopixel-trackingrdquo has been used to track velocities on many glaciers(egScambos et al 1992)

AROP is used to co-register the more recent scene in apair of orthorectified ASTER images to the earlier sceneto minimize misfits The open source ROIPACrsquos ldquoamp-corrdquo routine (Rosen et al 2004) is used to calculate EndashW and NndashS offsets The results are post-filtered by exclud-ing offsets with a signal-to-noise ratio (SNR which is thepeak cross correlation value divided by the average) belowa manually selected threshold (Willis et al 2012a) A lin-ear elevation-dependent correction (determined from appar-ent ldquobedrockrdquo velocities) is applied to the velocities to cor-rect for the elevation-dependent bias due to imprecise co-registrationorthorectification (Nuth and Kaab 2011) Fig-ure 5 shows a typical trend for an ASTER pair We do notknow the physical origin of the biasAhn and Howat(2011)attribute the systematic elevation-dependent displacement er-rors over the ice surface to co-registration errors Applyingan elevation-dependent correction based on the displacementof off-ice areas largely removes this (Fig6)

Uncertainty for each pair is estimated from motion onice-adjacent ldquobedrockrdquo (seeWillis et al 2012afor details)which should be zero Horizontal misalignment leads to off-ice motions when calculating offsets so error due to mis-alignment is included in the uncertainty estimate

232 QuickBird 2

A high-resolution (1 mpixel) QB02 image pair from 2011(30072011ndash16082011) is run through largely the sameprocessing applied to ASTER image pairs to generate off-

Table 1Volume change rates and averagedhdt

for the 16 largest out-let glaciers on the CDI with ldquonorthernrdquo ldquosouthernrdquo and whole ice-field totals in the last three rows Includes estimates of sub-aqueousvolume change for Marinelli CDI-08 Garibaldi and Darwin

Glacier Area dVdt

Averagedhdt

(km2) (km3 yrminus1) (m we yrminus1)

Marinelli Glacier 13293 minus074plusmn 016 50plusmn 11CDI-08 Glacier 12742 minus043plusmn 013 minus30plusmn 09Roncagli Glacier 11606 minus018plusmn 015 minus14plusmn 12Stoppani Glacier 10204 minus02plusmn 012 minus18plusmn 11Garibaldi Glacier 6386 002plusmn 013 02plusmn 18CDI 157 Glacier 5742 minus011plusmn 008 minus17plusmn 13de la Vedova Glacier 5683 minus015plusmn 007 minus24plusmn 11BahiaBroken Glacier 5322 minus01plusmn 008 minus18plusmn 13CDI 150 Glacier 5231 minus002plusmn 007 minus03plusmn 12CDI 456 Glacier 5122 minus003plusmn 009 minus05plusmn 15Darwin Glacier 4561 minus018plusmn 007 minus36plusmn 13Cuevas Glacier 4527 minus006plusmn 008 minus12plusmn 16CDI 455 Glacier 3773 002plusmn 008 04plusmn 20Oblicuo Glacier 2981 002plusmn 008 06plusmn 24CDI 48 Glacier 2858 minus004plusmn 005 minus13plusmn 15CDI 142 Glacier 2679 minus005plusmn 007 minus18plusmn 23

ldquoNorthernrdquo side 1322 minus30plusmn 10 minus19plusmn 07ldquoSouthernrdquo side 1283 minus13plusmn 06 minus10plusmn 04Entire CDI 2605 minus43plusmn 13 minus15plusmn 06

sets at 30 m resolution The images are orthorectified to an13112007 ASTER DEM with Rational Polynomial Coef-ficients (RPCs) Orthorectifying to a more recent ASTERDEM rather than the SRTM DEM minimizes errors causedby pronounced thinning

233 Advanced Land Observing Satellite (ALOS)

We use L-band SAR pixel-tracking (egRignot 2008Strozzi et al 2008 Rignot et al 2011) from two46 day ascending ALOS pairs (28092008ndash13112008 and04012011ndash19022011) to produce ice velocities for sev-eral CDI glaciers The ALOS SAR images have an ini-tial pixel resolution of 33 m (azimuth) by 83 m (range)The offsets have an effective resolution of approximately150 m (azimuth) by 200 m (range) based on the step size of50 pixels (azimuth) by 25 pixels (range) The Synthetic Aper-ture RADAR (SAR) images cover a broader area than theoptical images and are not limited by cloud cover provid-ing velocities at many glaciers with no ASTER observationsSAR pixel-tracking also performs well in the snow-coveredhigh-altitude accumulation zone where optical images lacktrackable features The ALOS pairs fail to capture the highspeeds near the fronts of several glaciers due to decorrelationcaused by strain and possibly melting

Raw ALOS SAR data are processed using ROIPAC andoffsets are produced by ldquoampcorrdquo The results are SNR-filtered and run through the elevation-dependent correction

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

830 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 5Scatter plot of EndashW ldquobedrockrdquo velocities for the 07092001to 25092001 ASTER pair The black line shows the linear trendfitted to the blue points this is removed from the overall EndashW ve-locity results Red points are excluded when fitting the trend (theyare greater thanplusmnσ from the median value) This cutoff makes anegligible difference in the trend for this pair but for other pairsit can change the trend significantly (eggt 1 m dayminus1 per 1000 melevation)

routine ALOS interferometry does not yield velocities due tothe relatively large motions long separation between scenesand changing surface characteristics

234 Orthorectification errors from thinning

Pronounced thinning can cause orthorectification errors inoptical imagery that do not appear in the off-ice velocitiesThe longer the timespan between the acquisition date of im-agery used to measure offsets and the DEM used to orthorec-tify the imagery the greater the error Our standard process-ing uses the SRTM DEM acquired in February 2000 to or-thorectify the ASTER L1B images used for pixel-trackingExamining ourdh

dtresults (Fig1) reveals this to be a po-

tential problem over the three most rapidly thinning glaciersMarinelli CDI-08 and Darwin For Marinelli Glacier we mit-igate this effect by orthorectifying QuickBird 2 imagery from2011 to a 2007 ASTER DEM rather than the SRTM DEMFurthermore the difference in incidence angle between thetwo QuickBird 2 images is less than 02 degrees This meansthe baseheight (BH ) ratio (egFujisada et al 2005) is low(0003) and the pair is insensitive to DEM errors or elevationchange between the time of the DEM and the time of imageacquisition

The two ASTER pairs for which we obtain front speeds atMarinelli are from 2001 (07092001ndash25092001) and 2003

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

a0 1 2 3Speed (mday)

minus69˚45 minus69˚30 minus69˚15

minus54˚30

minus54˚15

b

Fig 6 Effect of elevation-dependent velocity correction for07092001 to 25092001 ASTER image pair covering MarinelliGlacier Glacier outlines in black bedrock in gray (correction is notapplied to water pixels)(a) Is a map of speeds with no correctionapplied(b) is a map of speeds with correction applied Correctionreduces bedrock motion the mean and standard deviation for theentire pair drops from 06plusmn 04 m dayminus1 for the uncorrected speedsto 04plusmn 03 m dayminus1 for the corrected speeds

(06092003ndash13092003) We co-register the earlier imagein each of these pairs to the later image and orthorectify us-ing the DEM of the later image Orthorectifying the ASTERimagery to the coincident ASTER DEM minimizes disparitydue to DEM errors (with error due to thinning almost en-tirely removed) and difference in incidence angle betweenthe ASTER images (egScherler et al 2008 equation 1)

Front speeds are captured for the Darwin Glacier froma 7 day ASTER image pair in 2001 (25092001 to02102001) For this pair we orthorectify to the SRTM DEMas the time between SRTM acquisition and the ASTER pairis fairly short The direction of the velocity vectors are con-sistent with glacier flowlines visible in the ASTER imageryproviding confidence in the result Velocities over CDI-08 arefrom radar pixel-tracking orthorectification errors are not aconcern because radar images are not orthorectified

235 Marinelli Glacier ndash Flux

We calculate flux along transects perpendicular to glacierflow (as close as possible to the front) for velocities from07092001 to 25092001 06092003 to 13092003 and30072011 to 16082011 The height of the glacier is de-termined from the 25092001 ASTER DEM for the 2001pair 13092003 ASTER DEM for the 2003 pair and the13112007 ASTER DEM for the 2001 pair (adjusted to2011 using ourdh

dt) The 25092001 and 13092003 ASTER

DEMs do not need to be adjusted because they are coinci-dent with one of the images in the velocity pairs they are be-ing used to orthorectify and the maximum interval is 18 days(for the 2001 pair) We assume an average glacier depth be-low water of 150 m (seeKoppes et al 2009 Fig 4a and b)Adding this to the height gives an approximate thickness Wemultiply the glacier thickness by the perpendicular velocity

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 831

along the transect to calculate flux Sources of uncertaintythat we include in the uncertainty for our flux estimates arethe uncertainties on the speed uncertainty on the depth belowwater (plusmn50 m) and uncertainty on the DEMs used to obtainelevations

3 Results

31 Elevation change rates

Figure1 provides a map ofdhdt

for the CDI Thinning is con-centrated at lower elevations at the front of several tidewaterglaciers Thickening is apparent on the southern side of someof the highest mountains and at the front of the advancingGaribaldi Glacier We calculate adV

dtof minus42plusmn 17 km3 yrminus1

for the CDI (extent shown in Fig1) This rate equates toan area averageddh

dtof minus16plusmn 07 m yrminus1 Our estimate of

the sub-aqueous volume loss rate (minus012plusmn 006 km3 yrminus1)provides an additional (sim 3 ) loss giving a final dV

dt

of minus43plusmn 17 km3 yrminus1 equivalent to a mass loss rate ofminus39plusmn 15 Gt yrminus1 assuming surface elevation changes arenot due to densification processes and assuming an ice den-sity of 900 kg m3 This corresponds to an overall averagedthinning rate ofminus15plusmn 06 m we yrminus1 The ablation zonewhich comprises 60 of the icefield accounts for 76 of the mass loss The ablation zone has an average thin-ning rate ofminus18plusmn 06 m we yrminus1 (including sub-aqueousmass change) the accumulation zone has an average rateof minus09plusmn 03 m we yrminus1 Three glaciers (Marinelli GlacierDarwin Glacier and CDI-08 Glacier) account for 31 of themass loss but cover only 12 of the icefield area

We estimate a thinning rate ofminus10plusmn 04 m we yrminus1 forthe southern side which is lower than the northern side(minus19plusmn 07 m we yrminus1) The contrast between north andsouth is most likely due to warming in the north and chang-ing weather patterns that have increased precipitation on thewindward side of the mountains and decreased precipitationon the lee side (egHolmlund and Fuenzalida 1995) andof course the combination of climate with dynamic instabil-ity at glaciers such as Marinelli (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Table1 gives thedV

dtfor the 16 largest glaciers with rates

for the entire basin the accumulation zone and the ablationzone

32 Velocities

Pixel-tracking provides useful velocities from twentyASTER image pairs one QB02 pair and two ALOS pairswith acquisition dates between August 2001 and August2011 (Fig7) ASTER pixel-tracking results are generallybetter over the northern half of the CDI as there is less cloudcover than to the south (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007) Composite speed results are

0

5

10

15

20

Pair

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Year

ASTER

QB02

ALOS

Fig 7Date intervals for 22 pairs that produce usable pixel-trackingresults The most recent is a QB02 pair (blue) the next most recentis an ALOS pair (green) and the remainder are ASTER pairs (red)

shown in Fig8 velocities for individual pairs over selectedglaciers are shown in Fig9

Average front speeds for the tidewater Marinelli Glacier(133 km2 ndash the largest glacier of the CDI egKoppes et al2009) are obtained between 07092001 and 25092001(18 days) and 06092003ndash13092003 (7 days) from opticalimage pairs acquired by ASTER In order to capture re-cent rapid motions at the front we process a 30072011to 16082011 QB02 pair Figure10 shows a map of the2011 velocities from the QB02 pair along with longitudi-nal speed profiles for the QB02 2001 ASTER 2003 ASTERand 04012011ndash19022011 ALOS pairs highlighting theconsistently high front speeds from 2001 to 2011 Speedsat the front reach a maximum ofsim 75plusmn 02 m dayminus1 in2001sim 95plusmn 06 m dayminus1 in 2003 andsim 10plusmn 03 m dayminus1

in 2011Speeds at the front of the tidewater Darwin Glacier

(46 km2) reach a maximum average of 97plusmn 08 m dayminus1 forthe period 25092001 to 02102001 No repeat measure-ments of motion from ASTER pairs are available for theDarwin Glacier CDI-08 (127 km2) the furthest west andsouth of the three most rapidly thinning glaciers reachesspeeds of 20plusmn 05 m dayminus1 within 1 km of its 15012011front (04012011 to 19022011 ALOS pair) unfortunatelyrepeat speeds are not available for this glacier

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

832 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 8Composite speed map for the CDI incorporating ASTER- and ALOS-derived ice velocities

4 Discussion

41 Elevation change rates

411 Individual glaciers

A longitudinal profile of elevation and thinning rates forMarinelli Glacier (Fig11) illustrates both terminus retreatand the dramatic thinning occurring at the glacier Thinningextends from the terminus to the highest parts of the accumu-lation zone The longitudinal profile of the glacier surface re-mains convex as the glacier thins and retreats (Fig11) indi-cating that basal stresses are relatively high and the front re-mains grounded (egKoppes et al 2009) as of 13112007

The Marinelli Glacier retreatedsim 4 km between 2001 and2011 (measured from ASTER and QB02 imagery) an aver-age rate of retreat of 04 km yrminus1 This is lower than the re-treat rate of 1 km yrminus1 measured byKoppes et al(2009) dur-ing the late 1990s Our rate is similar to their average rate ofsim 03 km yrminus1 from 1960 to 2005Koppes et al(2009) infervolume loss rates at Marinelli Glacier from observed retreatrates A higher retreat rate provides a higher volume loss es-timate They find that the volume change rate dropped fromminus07 km3 yrminus1 in 1997 tominus02 km3 yrminus1 by 2005 In con-trast our measurements show that volume loss at Marinelliis sustained at an average rate ofminus07plusmn 02 km3 yrminus1 untilat least 2007

The accumulation area ratio (AAR) of Marinelli Glacieris 038 at an ELA of 1100 m (Fig12b) Marinelli under-went a significant reduction in AAR as the ELA moved up toits current altitude likely contributing to the period of neg-

ative mass balance in the mid-1990s preceding the currentphase of rapid retreat (egKoppes et al 2009) The AARhas already moved through the part of the hypsometry curvewhere small changes in elevation lead to large changes in theAAR an increase of 200 m from the current ELA of 1100 mwould only reduce the AAR from 038 to 031 The nega-tive mass balance operates in conjunction with dynamic in-stability now that large changes in the ELA have relativelylittle affect on the AAR the continued recession of MarinelliGlacier will likely be even more dependent on the underlyingfjord bathymetry

The tidewater Darwin Glacier has a similar thinning pro-file to the Marinelli Glacier (Fig13) retreatingsim 2 km from22022000 to 15012011 The 15012011 profile is con-vex for Darwin Glacier suggesting that the front remainsgrounded (egBenn and Evans 1998 Koppes et al 2009)The retreat and thinning at Darwin Glacier as for MarinelliGlacier is characteristic of other fast-moving temperate tide-water glaciers (Koppes et al 2009) Examples include JorgeMontt HPS-12 and Upsala glaciers on the SPI (Willis et al2012b see Figs 1 and 2Sakakibara et al 2013) Upsalacalves into a lake but its pattern of thinning and retreat issimilar to Jorge Montt and HPS-12

CDI-08 the third glacier with a strong thinning signal ison the southern side of the icefield and faces west This po-sition and aspect favors increased snowfall in the prevailingclimate (Holmlund and Fuenzalida 1995) Given its locationand the direction it is facing CDI-08 is somewhat anomalousin that almost 2 km of retreat has occurred and the glacierhas thinned rapidly between 2001 and 2011 The average

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 833

Fig 9Velocity maps for individual glaciers (separate color scale for each map) highlighting the quality of the velocity results for individualimage pairs(a) Shows velocities from the 04012011ndash19022011 ALOS pair for CDI-08 Glacier with an ASTER image from 15012011in the background(b) Shows velocities from the 07092001ndash25092001 ASTER pair for Marinelli Glacier with an ASTER image from07092001 in the background(c) Shows velocities from the 04012011ndash19022011 ALOS pair for Roncagli Glacier with an ASTER imagefrom 02072002 in the background(d) Shows velocities from the 25092001ndash02102001 ASTER pair for Darwin Glacier with an ASTERimage from 25092001 in the background

thinning rate for CDI-08 isminus30plusmn 09 m we yrminus1 comparedto minus10plusmn 04 m we yrminus1 for the southern part of the CDI

Garibaldi Glacier on the southern side of the icefield witha southern aspect by contrast has relatively large areas ofpositive dh

dtin its ablation zone Ourdh

dthere are based on

sparse temporal coverage and there is only one ASTER im-age (13092003) that covers the entire front of GaribaldiGlacier Comparison of a WorldView-1 (optical) image from09272011 with a 13092003 ASTER image shows thatthe glacier has advanced by more than 1 km between 2003and 2011 Figure3 shows the frontal variation history ofGaribaldi over asim 65 yr period from Landsat TM and aerialimagery and also shows advance in the past decade Further

confirmation of advance in the past decade comes from fieldreports in the austral summer of 2007 when the glacier wasobserved destroying trees and frequently calving (Masiokaset al 2009) This gives us confidence in the positivedh

dtwe

observe for Garibaldi Glacier despite the limited data Themaximum recorded retreat was reached in 2001 followed byadvance to the 2011 front that is within 05 km of the 1945front (the maximum extent Garibaldi during this period) andno more than 2 to 3 km of retreatadvance in the past 65 yr

The contrast in current behavior between CDI-08 andGaribaldi may be due partly to the lower ELA of CDI-08 andthe higher sensitivity of its AAR to changes in climate com-pared with Garibaldi (Fig12a and c show the AAR versus

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

834 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Fig 10Speeds for Marinelli Glacier(a) Is a map of velocities from the QB02 (30072011ndash16082011) pair (background is the 07092001ASTER V3N image)(b) Shows profiles (indicated by the red line ina) from two ASTER pairs (07092001ndash25092001 06092003ndash13092003) the QB02 pair and an ALOS pair (04012011ndash19022011) The QB02 velocities extend to the 2011 front (shown by theyellow line in a) comparing them with the 2001 ASTER V3N image and the ASTER speed profiles highlights thesim 4 km of front retreatbetween 2001 and 2011 We measure maximum front speeds at Marinelli Glacier of 75plusmn 02 m dayminus1 in 2001 95plusmn 06 m dayminus1 in 2003andsim 10plusmn 03 m dayminus1 in 2011

Fig 11 dhdt

and elevations for a longitudinal profile (starting towards the front) on Marinelli Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 13112007 From the elevation profiles it is clear that the front has retreated between 2ndash3 km between 2000 and2007(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

elevation curves for Garibaldi and CDI-08) At our estimatedELA of 640 m CDI-08 has an AAR of 071 at an ELA of650 m Garibaldi has an AAR of 089 Shifting the ELA ofCDI-08 up to 840 m reduces the AAR to 046 (a reductionof 25 ) raising the ELA of Garibaldi by more than twiceas much to 1090 m would reduce the AAR to 067 (22 )Post et al(2011) note that an AAR above 08 typically fa-vors an advance in TWG assuming the ELA of Garibaldi islower than 900 m its AAR is greater than 08 The behavior

of Garibaldi Glacier with an oscillating quasi-stable frontsince at least 1945 is consistent with an AAR that hoversaround 08 and is relatively insensitive to rising temperaturesin the region (egHolmlund and Fuenzalida 1995 Strelinand Iturraspe 2007 Lopez et al 2010)

Our results (positivedhdt

and advance in the ablationzone versus relatively more negativedh

dtin the accumula-

tion zone) indicate the possibility of surge-like behavior (egRivera et al 1997) However a time series of velocities for

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 835

0

400

800

1200

1600

2000

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Garibaldi Glacier Hypsometry minus AAR vs Elevation

a

0

500

1000

1500

2000

2500

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

Marinelli Glacier Hypsometry minus AAR vs Elevation

b

0

400

800

1200

1600

Ele

va

tio

n (

m)

0 10 20 30 40 50 60 70 80 90 100

AAR ()

CDIminus08 Glacier Hypsometry minus AAR vs Elevation

c

Fig 12Hypsometry for Garibaldi(a) Marinelli (b) and CDI-08(c)glaciers from the SRTM DEM The curve indicates what the AARwould be for the corresponding elevation The black dot indicatesthe current ELA the red dot indicates a 200 m upward shft Thiswould reduce the AAR for Garibaldi Glacier from 089 to 082 forMarinelli Glacier form 039 to 031 and for CDI-08 Glacier theAAR would drop from 071 to 046

Garibaldi Glacier would be necessary to ascertain whetherthis is the case It is important to note that the underly-ing topography may be playing an important role in regu-lating Garibaldirsquos behavior as well eg buoyancy forces atthe front may be reduced by a submarine terminal moraineandor a shallower fjord (egPost et al 2011)

412 Comparison with other southern hemisphericglaciers

Retreat and thinning is also occurring on other glaciers atsimilar latitudes in the Southern Hemisphere Ice massesin the Kerguelen Islands (49 S 69E) for example ex-

perienced accelerating retreat rates and an averagedhdt

ofminus14 tominus17 m yrminus1 between 1963 and 2000 (Berthier et al2009) Glaciers on the subantarctic island of South Geor-gia (545 S 37 W) have undergone retreat in response towarming since the 1950s with recession being particularlypronounced from the 1980s to the mid-2000s (Gordon et al2008) Widespread retreat is also observed for the glaciers ofHeard Island (531 S 735E) in the southern Indian Ocean(Thost and Truffer 2008)

A similar amount of warming has occurred in all these lo-cations with an increase in average temperature of 04 to14C south of 46 S in Patagonia since the beginning ofthe 20th century (Rosenbluth et al 1995) 1C of warm-ing between 1964 and 1982 for the Kerguelen Islands (egBerthier et al 2009) approximately 1C of warming from1950 to mid-1990s on South Georgia (egGordon et al2008) and 09C of warming for Heard Island since the mid-20th century (egThost and Truffer 2008)

At the CDI and SPI the thinning and rapid retreat ofMarinelli and Jorge Montt glaciers from topographic pinningpoints (Koppes et al 2009 Rivera et al 2012) anticipatesthe predicted response of tidewater glaciers in other regionssuch as South Georgia if warming trends continue (egGor-don et al 2008 Sect 51)

42 Velocities ndash Marinelli Glacier

421 Marinelli Glacier ndash overview

Below we estimate flux for Marinelli Glacier using ourspeeds then compare our results withKoppes et al(2009)who estimate the terminus speed and flux of Marinelli fromthe retreat rate We find that our results do not agree withKoppes et al(2009) Whereas they infer a reduction in ter-minus speed for Marinelli from 2000 to 2005 we find thatthe front speed in 2003 and 2011 is at least as high as in2001 and consequently the flux in 2011 is approximatelythe same as the flux in 2001 We conclude that thinningat Marinelli Glacier is probably dynamic with bed geome-try likely governing velocity and retreat We then considerMarinelli Glacier as a tidewater-cycle glacier (TWG) in re-treat phase (egMeier and Post 1987 Motyka et al 2003Post et al 2011) and compare it with Jorge Montt Glacieron the SPI

422 Marinelli Glacier ndash flux

We estimate a flux of 05plusmn 02 km3 yrminus1 for the 2001 pair07plusmn 02 km3 yrminus1 for the 2003 pair and 05plusmn 02 km3 yrminus1

for the 2011 pair Flux is highest in 2003 (due to higherspeeds than 2001 and a larger front than 2011) but the im-portant point is that the 2011 flux has not dropped relativeto 2001 due to speeds at the 2011 front that are higher than2001 and as high as 2003

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

836 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

minus69˚30

minus54˚42

a

minus25

minus20

minus15

minus10

minus5

0

5

dh

dt

(my

r)

0 4000Distance (m)

c

minus20 minus10 0

dhdt (myr)

0

100

200

300

400

500

600

700

Ele

va

tio

n (

m)

0 4000Distance (m)

b

022000

092001

092001

092001

102001

022002

092003

082004

012007

112007

012011

Fig 13 dhdt

and elevations for a longitudinal profile (starting towards the front) on Darwin Glacier A map ofdhdt

is shown in(a) the trackused in(b) and (c) is plotted in green(b) Gives elevation profiles for different dates along the green track in(a) The colorscale for theelevation profiles indicates the relative time of acquisition Dark blue is the SRTM elevation (22022000) the red track extending to the frontis an ASTER DEM from 15012011(c) Shows thedh

dtprofile for the green track in(a) The colorscale for(a) is the same as(c)

Seasonal changes in conditions at Marinelli Glacier couldbe influencing our results we note that the two ASTER pairsfrom which we obtain front speeds are both from SeptemberHowever it is possible that the speeds in September 2001 and2003 are different due to inter-annual variations in the onsetof conditions affecting the glacier speed

The 2011 QuickBird 2 pair is from 30072011 to16082011 While this is not an entirely different seasonfrom the ASTER pairs it is a month earlier which is longenough for seasonal variations to possibly play a role in anyobserved speed differences We consider it unlikely howeverthat a seasonal component of motion is dominant in the 30 increase in the front speed at Marinelli between September2001 (75 m dayminus1) and August 2011 (10 m dayminus1) Unfortu-nately we do not have enough repeat measurements to quan-tify any seasonal effect on speeds especially given that ourALOS results over Marinelli from the austral summer do notreach the front

423 Marinelli Glacier ndash comparison with previousresults

Koppes et al(2009) infer speeds from the retreat rate withlower retreat rate leading to lower inferred speed Theyshow Marinelli glacier slowing from a terminus ice speed of8 m dayminus1 in 2001 (similar to our 75 m dayminus1 2001 speed)to 55 m dayminus1 in 2003 Our results however are evidencethe glacier has not slowed from 2001 (75 m dayminus1) to 2003(95 m dayminus1) We find that contrary to the conclusion ofKoppes et al(2009) the front speed and therefore the fluxare not decreasing from 2001 to 2003 Furthermore the frontspeed in 2011 is higher than 2001 so despite the front beingless extensive our estimate of the 2011 flux is as high as for2001

424 Marinelli Glacier ndash thinning gradient maintainssurface slope at the front

The maintenance of relatively high speeds at the front ofMarinelli Glacier between 2001 and 2011 can be attributedin part to the observed gradient indh

dt There is rapid thinning

and retreat at the front with slower thinning rates upstreamThis imbalance maintains surface slope near the glacier frontsustaining the driving stress according to the equationσ =

ρghsin(α) whereσ is the driving stressρ is the densitygis gravitational accelerationh is thickness andα is the slope(egCuffey and Paterson 2010 p 295)

425 Marinelli Glacier ndash tidewater cycle

The rapid retreat noted in this study and covered elsewhere(egKoppes et al 2009 Warren and Aniya 1999) cou-pled with the thinning we observe between 2000 and 2011suggests that Marinelli Glacier is a tidewater-cycle glacier(TWG) in retreat phase (egMeier and Post 1987 Motykaet al 2003 Post et al 2011) Front recession has openedthe fjord at Marinelli Around 1945 retreat began as theglacier receded into deeper water from the arcuate terminalmoraine it had been pinned at with particularly rapid retreatthroughout the 1990s and 2000s (Holmlund and Fuenzalida1995 Porter and Santana 2003 Koppes et al 2009) Fjordbathymetry (and possibly ocean temperatures) likely governsthe ice dynamics and calving rate (egBenn et al 2007Koppes et al 2009 Straneo et al 2010) which probablycontrol thinning at Marinelli (egKoppes et al 2009) Thevelocity results for Marinelli Glacier support this hypothesisThe 2011 speeds at Marinelli which are higher than 2001and as high as 2003 suggest that the glacier has not yet re-treated to fjord depths shallow enough to slow it down Alongwith an estimate of the 2011 flux that is as high as 2001this leads us to predict that Marinelli Glacier will continue

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 837

to retreat at least until it reaches a new topographic pinningpoint which would likely cause the terminus speed to falland calving rates to drop There is no evidence of accumu-lation in the recent history of Marinelli Glacier that wouldcompensate the mass loss due to calving NCEP-NCAR cli-mate model results show that precipitation has decreased atMarinelli Glacier and temperatures have risen (Koppes et al2009)

We do not see similar large thinning signals at adjacentglaciers with similar settings which is further evidence thatdynamic instability and not melt due to a regional increasein atmospheric temperatures is the main proximate cause ofthe current thinning at Marinelli However warming and de-creased precipitation has likely helped trigger the initial re-treat and thinning from the terminal moraine and contributedto the recent history of negative mass balance (Holmlund andFuenzalida 1995 Porter and Santana 2003 Koppes et al2009)

Jorge Montt Glacier on the SPI is a somewhat analogousTWG with a grounded front thinning occurring at approxi-mately the same rate and a rapidly receding terminus (Riveraet al 2012 Willis et al 2012b) As at Marinelli terminusretreat opened the fjord at Jorge Montt with subsequent rapidretreat as the glacier receded into deeper water (Rivera et al2012) At Marinelli the retreat is thought to have been ini-tiated by thinning linked to climate changes (egKoppeset al 2009) Though there is a high degree of variabilityfrom glacier to glacier on the Patagonian ice fields (Riveraet al 2012) the similar behavior of these two glaciers in thesame region but on different ice fields suggests that changingclimate plays a role in retreat and thinning at Jorge Montt aswell given that the retreat was inferred to have been triggeredby warming-induced thinning at the fronts of both glaciers

5 Conclusions

We provide the first icefield-wide thinning rates and glaciervelocities for the Cordillera Darwin Icefield We find thin-ning at an average rate ofminus15plusmn 06 m we yrminus1 for theCDI between 2000 and 2011 (equivalent to a mass lossrate of minus39plusmn 15 Gt yrminus1) with less thinning in the south(minus10plusmn 04 m we yrminus1) compared to the northern part of theicefield (minus19plusmn 07 m we yrminus1) This pattern of thinning isconsistent with climate records that show a warming trendin this region from the 1940s through the 1990s leading tohigher temperatures along with decreased precipitation onthe northern side of the icefield and an increase in precipi-tation on the southern side (Holmlund and Fuenzalida 1995Strelin and Iturraspe 2007 Koppes et al 2009)

Operating in conjunction with local glacier geometry ris-ing temperatures triggered a period of prolonged retreat atMarinelli Glacier (since 1945) (egHolmlund and Fuenza-lida 1995 Porter and Santana 2003 Koppes et al 2009)Recently Marinelli Glacier retreated approximately 4 km

from 2001 to 2011 reflected in ourdhdt

by the zone ofrapid thinning near its front (approximatelyminus25 m yrminus1)Our repeat velocity measurements support the hypothesisthat Marinelli is still in a retreat phase We predict the retreatof Marinelli Glacier until it reaches a new topographic pin-ning point Ourdh

dtmap shows a similar pattern and degree of

thinning at Darwin and CDI-08 glaciers but we do not havethe thickness or repeat speed measurements required to esti-mate flux and decipher whether thinning at these glaciers isdynamic

All three Patagonian ice fields are losing mass TheNPI with an averagedh

dtof minus10plusmn 01 m we yrminus1 (Willis

et al 2012b) is thinning at a slower rate than the CDI(minus15plusmn 06 m we yrminus1) which is thinning at approximatelythe same rate as the SPI (minus16plusmn 01 m we yrminus1) (Williset al 2012b) Other glaciers at the same latitude (eg theKerguelen Islands the island of South Georgia and Heard Is-land) are thinning and retreating and have undergone a simi-lar degree of warming to Patagonia (Rosenbluth et al 1995Gordon et al 2008 Thost and Truffer 2008 Berthier et al2009) The CDI contributed 001plusmn 0004 mm yrminus1 to sealevel between 2000 and 2011 The combined NPI SPI andCDI ice fields contributed 0078plusmn 0008 mm yrminus1 over thesame period This rate compares well with recent GRACEestimates (Jacob et al 2012) and could be further refined byincluding for example DEMs generated from stereo high-resolution imagery Accurate basin-by-basin knowledge ofaccumulation rates would also help refine the cutoffs usedto exclude spurious elevations producing betterdh

dt

AcknowledgementsASTER data was provided by the LPDAACWe thank the Alaska Satellite Facility (ASF) and the JapanAerospace Exploration Agency (JAXA) for the ALOS data Sup-ported by NASA grant NNX08AI87G issued through the ScienceMission Directorates Earth Science Division Andres Rivera andFrancisca Bown acknowledge the GLIMS project for providingfree satellite imagery in the full glacier inventory and surfacevariations coming in a separate book chapter and the FONDECYTproject 1080320 QuickBird 2 and WorldView 1 images providedby NGA Commercial Imagery Program Sasha A Bernstein waspartially supported by the Cornell Rawlings Presidential ResearchScholars Program

Edited by A Kaab

References

Ahn Y and Howat I M Efficient Automated Glacier Sur-face Velocity Measurement From Repeat Images Using Multi-ImageMultichip and Null Exclusion Feature Tracking IEEE TGeosci Remote 49 2838ndash2846 2011

Arendt A A Echelmeyer K A Harrison W D Lingle C Sand Valentine V B Rapid wastage of Alaska glaciers and theircontribution to rising sea level Science (New York NY) 297382ndash386doi101126science1072497 2002

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013

838 A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI

Aster R C Borchers B and Thurber C H Parameter Estimationand Inverse Problems Elsevier Inc Burlington MA 2005

Bamber J L and Rivera A A review of remote sensing methodsfor glacier mass balance determination Global Planet Change59 138ndash148 2007

Benn D I and Evans D J Glaciers amp Glaciology Oxford Uni-versity Press Inc New York 1998

Benn D I Hulton N R J and Mottram R H ldquoCalvinglawsrdquo ldquosliding lawsrdquo and the stability of tidewater glaciers AnnGlaciol 46 123ndash130 2007

Berthier E Arnaud Y Baratoux D Vincent C and Remy FRecent rapid thinning of the ldquoMer de Glacerdquo glacier derivedfrom satellite optical images Geophys Res Lett 31 L17401doi1010292004GL020706 2004

Berthier E Le Bris R Mabileau L Testut L and RemyF Ice wastage on the Kerguelen Islands (49 S 69 E)between 1963 and 2006 J Geophys Res 114 F03005doi1010292008JF001192 2009

Berthier E Schiefer E Clarke G K C Menounos Band Remy F Contribution of Alaskan glaciers to sea-levelrise derived from satellite imagery Nat Geosci 3 92ndash95doi101038ngeo737 2010

Bown F Rivera A Zenteno P Bravo C and Cawkwell F Firstglacier inventory and recent glacier variations of Isla Grandede Tierra del Fuego and adjacent islands in Southern Chile inGLIMS book in press 2013

Cuffey K M and Paterson W S B The Physics of GlaciersButterworth-Heinemann fourth Edn 2010

Davies B and Glasser N Accelerating shrinkage of Patagonianglaciers from the Little Ice Age (simAD 1870) to 2011 J Glaciol58 1063ndash1084doi1031892012JoG12J026 2012

Fernandez R A Anderson J B Wellner J S and Hallet BTimescale dependence of glacial erosion rates A case studyof Marinelli Glacier Cordillera Darwin southern Patagonia JGeophys Res 116 F01020doi1010292010JF001685 2011

Fujisada H Bailey G Kelly G Hara S and Abrams MASTER DEM performance IEEE T Geosci Remote 43 2707ndash2714doi101109TGRS2005847924 2005

Gao F Masek J and Wolfe R E Automated registra-tion and orthorectification package for Landsat and Landsat-like data processing J Appl Remote Sens 3 033515doi10111713104620 2009

Gardelle J Berthier E and Arnaud Y Impact of reso-lution and radar penetration on glacier elevation changescomputed from DEM differencing J Glaciol 58 419ndash422doi1031892012JoG11J175 2012

Gardner A S Moholdt G Arendt A and Wouters B Long-term contributions of Baffin and Bylot Island Glaciers to sealevel rise an integrated approach using airborne and satel-lite laser altimetry stereoscopic imagery and satellite gravime-try The Cryosphere Discuss 6 1563ndash1610doi105194tcd-6-1563-2012 2012

Glasser N F Harrison S Jansson K N Anderson K and Cow-ley A Global sea-level contribution from the Patagonian Ice-fields since the Little Ice Age maximum Nat Geosci 4 303ndash307doi101038ngeo1122 2011

Gordon J E Haynes V M and Hubbard A Recent glacierchanges and climate trends on South Georgia Global PlanetChange 60 72ndash84doi101016jgloplacha200607037 2008

Holmlund P and Fuenzalida H Anomalous glacier responses to20th century climatic changes in Darwin Cordillera southernChile J Glaciol 41 465ndash473 1995

Howat I M Smith B E Joughin I and Scambos T A Ratesof southeast Greenland ice volume loss from combined ICE-Sat and ASTER observations Geophys Res Lett 35 L17505doi1010292008GL034496 2008

Ivins E R Watkins M M Yuan D-N Dietrich R Casassa Gand Rulke A On-land ice loss and glacial isostatic adjustment atthe Drake Passage 2003ndash2009 J Geophys Res 116 B02403doi1010292010JB007607 2011

Jacob T Wahr J Pfeffer W T and Swenson S Recent con-tributions of glaciers and ice caps to sea level rise Nature 482514ndash518doi101038nature10847 2012

Kaab A Monitoring high-mountain terrain deformation from re-peated air- and spaceborne optical data examples using digitalaerial imagery and ASTER data ISPRS J Photogramm 57 39ndash52 2002

Kaab A Berthier E Nuth C Gardelle J and ArnaudY Contrasting patterns of early twenty-first-century glaciermass change in the Himalayas Nature 488 495ndash498doi101038nature11324 2012

Koppes M Hallet B and Anderson J Synchronous accel-eration of ice loss and glacial erosion Glaciar MarinelliChilean Tierra del Fuego J Glaciol 55 207ndash220doi103189002214309788608796 2009

Koppes M Conway H Rasmussen L A and Chernos M De-riving mass balance and calving variations from reanalysis dataand sparse observations Glaciar San Rafael northern Patagonia1950ndash2005 The Cryosphere 5 791ndash808doi105194tc-5-791-2011 2011

Lopez P Chevallier P Favier V Pouyaud B Ordenes F andOerlemans J A regional view of fluctuations in glacier lengthin southern South America Global Planet Change 71 85ndash108doi101016jgloplacha200912009 2010

Masiokas M H Rivera A Espizua L E Villalba R DelgadoS and Aravena J C Glacier fluctuations in extratropical SouthAmerica during the past 1000 years Palaeogeogr Palaeocl 281242ndash268doi101016jpalaeo200908006 2009

Meier M F and Post A Fast tidewater glaciers J Geophys Res92 9051ndash9058 1987

Motyka R J OrsquoNeel S Connor C L and Echelmeyer K ATwentieth century thinning of Mendenhall Glacier Alaskaand its relationship to climate lake calving and glacier run-off Global Planet Change 35 93ndash112doi101016S0921-8181(02)00138-8 2003

Nuth C and Kaab A Co-registration and bias corrections of satel-lite elevation data sets for quantifying glacier thickness changeThe Cryosphere 5 271ndash290doi105194tc-5-271-2011 2011

Porter C and Santana A Rapid 20th century retreat of Ventis-quero Marinelli in the Cordillera Darwin Icefield Anales del In-stituto de la Patagonia 31 17ndash26 2003

Post A OrsquoNeel S Motyka R J and Streveler G A ComplexRelationship Between Calving Glaciers and Climate EOS Trans-actions American Geophysical Union 92 2011

Pritchard H D and Vaughan D G Widespread acceleration oftidewater glaciers on the Antarctic Peninsula J Geophys Res112 F03S29doi1010292006JF000597 2007

The Cryosphere 7 823ndash839 2013 wwwthe-cryospherenet78232013

A K Melkonian et al Satellite-derived volume loss rates and glacier speeds for the CDI 839

Quintana J Estudio de los factores que explican la variabilidad dela precipitacion en Chile en escalas de tiempo interdecadal Mscthesis Universidad de Chile 2004

Rasmussen L A Conway H and Raymond C F Influence ofupper air conditions on the Patagonia Icefields Global PlanetChange 59 203ndash216doi101016jgloplacha2006110252007

Rignot E Changes in West Antarctic ice stream dynamics ob-served with ALOS PALSAR data Geophys Res Lett 35L12505doi1010292008GL033365 2008

Rignot E Echelmeyer K and Krabill W Penetrationdepth of interferometric synthetic-aperture radar signalsin snow and ice Geophys Res Lett 28 3501ndash3504doi1010292000GL012484 2001

Rignot E Rivera A and Casassa G Contribution of the Patag-onia Icefields of South America to sea level rise Science (NewYork NY) 302 434ndash437doi101126science1087393 2003

Rignot E Mouginot J and Scheuchl B Ice Flowof the Antarctic Ice Sheet Science 333 1427ndash1430doi101126science1208336 2011

Rivera A Aravena J and Casassa G Recent fluctuations ofGlaciar Pıo XI Patagonia discussion of a glacial surge hypothe-sis Mt Res Dev 17 309ndash322 1997

Rivera A Benham T Casassa G Bamber J and DowdeswellJ Ice elevation and areal changes of glaciers from the NorthernPatagonia Icefield Chile Global Planet Change 59 126ndash137doi101016jgloplacha200611037 2007

Rivera A Koppes M Bravo C and Aravena J C Little Ice Ageadvance and retreat of Glaciar Jorge Montt Chilean PatagoniaClim Past 8 403ndash414doi105194cp-8-403-2012 2012

Rolstad C Haug T and Denby B Spatially integrated geodeticglacier mass balance and its uncertainty based on geostatisticalanalysis application to the western Svartisen ice cap NorwayJ Glaciol 55 666ndash680doi1031890022143097894709502009

Rosen P A Henley S Peltzer G and Simons M Updated Re-peat Orbit Interferometry Package EOS Transactions AmericanGeophysical Union 85 47doi1010292004EO050004 2004

Rosenbluth B Casassa G and Fuenzalida H Recent climaticchanges in western Patagonia Bull Glaciol Res 13 127ndash1321995

Sakakibara D Sugiyama S Sawagaki T Marinsek S andSkvarca P Rapid retreat acceleration and thinning of GlaciarUpsala Southern Patagonia Icefield initiated in 2008 AnnGlaciol 54 131ndash138doi1031892013AoG63A236 2013

San B T and Suzen M L Digital elevation model (DEM) genera-tion and accuracy assessment from ASTER stereo data Int J Re-mote Sens 26 5013ndash5027doi101080014311605001776202005

Scambos T A Dutkiewicz M J Wilson J C and BindschadlerR A Application of Image Cross-Correlation to the Measure-ment of Glacier Velocity Using Satellite Image Data RemoteSens Environ 42 177ndash186 1992

Scherler D Leprince S and Strecker M Glacier-surface veloci-ties in alpine terrain from optical satellite imagery ndash Accuracyimprovement and quality assessment Remote Sens Environ112 3806ndash3819doi101016jrse200805018 2008

Straneo F Hamilton G S Sutherland D A Stearns L ADavidson F Hammill M O Stenson G B and Rosing-Asvid A Rapid circulation of warm subtropical waters in amajor glacial fjord in East Greenland Nat Geosci 3 182ndash186doi101038ngeo764 2010

Strelin J and Iturraspe R Recent evolution andmass balance of Cordon Martial glaciers CordilleraFueguina Oriental Global Planet Change 59 17ndash26doi101016jgloplacha200611019 2007

Strelin J Casassa G Rosqvist G and Holmlund P Holoceneglaciations in the Ema Glacier valley Monte Sarmiento Mas-sif Tierra del Fuego Palaeogeogr Palaeocl 260 299ndash314doi101016jpalaeo200712002 2008

Strozzi T Kouraev A Wiesmann A Wegmuller U Sharov Aand Werner C Estimation of Arctic glacier motion with satelliteL-band SAR data Remote Sens Environ 112 636ndash645 2008

Thost D E and Truffer M Glacier Recession on Heard IslandSouthern Indian Ocean Arct Antarct Alp Res 40 199ndash214doi1016571523-0430(06-084)[THOST]20CO2 2008

Tucker C J Grant D M and Dykstra J D NASA Global Or-thorectified Landsat Data Set Photogramm Eng Rem S 70313ndash322 2004

Warren C R and Aniya M The calving glaciers of southern SouthAmerica Global Planet Change 22 59ndash77 1999

Willis M J Melkonian A K Pritchard M E and RamageJ M Ice Loss Rates at the Northern Patagonian Icefield De-rived Using a Decade of Satellite Remote Sensing Remote SensEnviron 117 184ndash198doi101016jrse201109017 2012a

Willis M J Melkonian A K Pritchard M E and Rivera AIce loss from the Southern Patagonian Ice Field South Amer-ica between 2000 and 2012 Geophys Res Lett 39 L17501doi1010292012GL053136 2012b

wwwthe-cryospherenet78232013 The Cryosphere 7 823ndash839 2013