JOURNAL OF GEOPHYSICAL RESEARCH.VOL. 109. D17104. doi:10.1029/2003JD00438R.2004

Stable water isotope characterization of human and natural impactson land-atmosphere exchanges in the Amazon BasinK. McGuffie

A. Henderson-Sellers

Department of Applied Physics. University of Technology. Sydney. Broadway, New South Wales. Australia

Environment, Australian Nuclear Science and Technology Organisation, Lucas Heights. New South Wales. Australia

Received 25 November 2003: revised 13 June 2004: accepted 30 June 2004: published 14 September 2004.

[I] Stable water isotopes have been employed as a means of challenging, validating, andimproving numerical models of the Amazon Basin since the 1980s. This paper serves as anexemplar of how characterization of human and natural impacts on surface-atmospherewater exchanges could beneficially exploit stable water isotope data and simulations.Interpretations of Amazonian isotopic data and model simulations are found to be seriouslyhampered by (1) poor simulation of the gross water budget (e.g., lack of surface waterconservation in models): (2) considerable model differences in the fate of precipitation (i.e.,between reevaporation and runoff'); (3) wide ranging characterization of natural causes ofwater isotopic fluctuations (especially El Nino and La Nina events); (4) isotopic land-atmosphere flux sensitivity to the prescription of boundary layer atmospheric water vaporisotopic depletion; and (5) significantly different characterization by current land-surfaceschemes of the partition of evaporation between isotopically fractionating (from lakes andrivers) and nonfractionating (transpiration) processes. Despite these obstacles, we findfeatures in the recent isotopic record that might be derived from circulation and land-usechanges. ENSO events may cause decreased depletion in the dry season, because ofreduced convective precipitation, while increases in upper basin isotope depletions in thewet season may result from relatively less nonfractionating recycling because there arefewer trees. The promise for isotopic fingerprinting of near-surface continental water cyclechanges depends upon fixing shortcomings in current atmospheric and land-surfacemodels. !:VDEX TER.\fS: 1615Global Change: Biogeochemicalprocesses (4805); 1610Global Change:Atmosphere (0315. 0325); 1833Hydrology: Hydroclimatology;3322 Meteorology and AtmosphericDynamics: Land/atmosphere interactions;KEl1fDRDS: stable isotopes. Amazon. models

Citation: McGuffie. K.. and A. Henderson-Sellers (2004). Stable water isotope characterization of human and natural impacts onland-atmosphere exchanges in the Amazon Basin. J Geophvs. Res .. j09. D17104.doi:I0.1029i2003JD004388.

1. Isotopes in the Amazon1.1. Isotopic Measurement and Modeling in theAmazon

[2] Land use change in the Amazon Basin. the largest andmost biologically diverse river system in the world. has thepotential to cause significant disruption to hydrological.biogeochemical and human systems. Within the atmosphericcomponent of the hydrologic cycle. fractionation of watermolecules arises from the processes of evaporation andcondensation [e.g .. Dansgaard, 1964]. By examining thetwo most abundant "heavy" isotopes of water (HDO andH201 x). it is possible to diagnose the history of evaporativeand condensation processes [e.g .. Ingraham and Craig.1986] (Figure la). An early review of Amazonian isotopicdata, published by Salati and Vose [1984]. was int1uentialbecause its publication coincided with the first Global Cli-

Copyright 2004 by the American Geophysical Union.0148-0227 1041:'003J D0043 88509 .00

mate Model (GeM) of the impact of Amazonian deforesta-tion on climate [Henderson-Sellers and Gornitz, 1984].Together. these papers underlined and disseminated the factthat the Amazon Basin recycles about half its water. Specif-ically, the central Amazon has a water recycling time of about5.5 days. and during this period. about half the rainfall isreevaporated or transpired, and of this around 50% falls againas precipitation [e.g.• Matsui et al., 1983]. As a consequence.the average gradient in 61

Xo going inland is only 1.5');;0per1000 krn as compared with 2.0%0 typical of other continentalareas [Rozanski et al., 1993].

[3] In 1991 the results of two isotopic models of Ama-zonian precipitation and its implications for regionalhydrology and climate were published. Gat and Matsui[1991] employed a simple steady state model of the centralAmazon Basin to demonstrate that some of the waterrecycling is from fractionating sources. Using data fromthe Intemational Atomic Energy Agency/World Meteoro-logical Organization (lAEA/WMO) global station networkup to 1981 [IAEA. 2003]. they interpreted a +3%0 deviation

Dl7104 I of 25

D17104 \!lCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES

troposphere

b=+1(+6)

D17104

Ocean

liquid/solid

~~~apo~

b)

/

$ooCO

sool.()/

Figure 1

2 or 25

DI7104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES DI7104

from the Meteoric Water line as evidence that 20-40% ofthe recycled moisture within the basin is derived fromsources such as lakes, the river or standing water, whichfractionate isotopes of oxygen and hydrogen. Victoria et al.[1991] combined IAEAiWMO isotopic results from Belemand Manaus (Figure Ib) between 1972 and 1986 and a box/sector model of Dall 'Olio [1976] and Salati et al. [1979] toshow that wet season recycling is primarily by transpiration,while dry season recycling is mostly accomplished byreevaporation of precipitation intercepted on the canopy(see also Gat [2000]. who reviews and revises this work).Martinelli et aI. [1996] used stable isotopes to determine thesources of evaporated water in the Amazon Basin.

[4] Henderson-Sellers et al. [2002] detected statisticallysignificant temporal changes (1965 to 1990) in stable waterisotopic signatures in the Amazon, which they comparedwith the results of GCM simulations, revealing notabledifferences. Their analysis found no significant change indry season isotopic characteristics despite earlier predictionsthat land-use change signals would be found first in the dryseason. In the context of recent GCM simulations ofAmazonian deforestation, Henderson-Sellers et al. [2002]suggested that changes observed in the isotope characteristicsare more consistent with the predicted effects of greenhousewarming, possibly combined with forest removal, than withthe predicted effects of deforestation alone.

[5] Although isotope-enabled GCMs (lGCMs) date backto the mid-1980s [e.g., Joussaume et al., 1984; Jouzel et al.,1987], there has been little application of their results to theAmazon until recently [e.g., Hoffman et al., 2003). Vuille etal. [2003a] analyze two IGCMs in terms of the interannualvariability of 6180 in precipitation over the tropical Amer-icas including the Amazon Basin. Both Hoffman et al.[2003] and Vuille et al. [2003b] focus on the 6180 inprecipitation in the Andes and the way in which itsvariability affects interpretation of the isotope record inAndean ice cores. They conclude that the impact of ENSOevents on 6180 integrates many fractionating factors includ-ing precipitation amount, moisture source and temperature(see Figure Ia).

1.2. Isotopic Interpretations of Land-AtmosphereInteractions

[6] Stable water isotopes allow differentiation betweentranspiration (through plants) and evaporation (from watersurfaces or stores). Throughout this paper, we use the termevaporation to encompass all water loss from the surface.Nonfractionating evaporation comprises transpiration(which, assuming the plant has a steady state water balance,has no net effect on the isotopic balance of soil water) andfull evaporation of canopy-intercepted precipitation (whichcannot fractionate if it is complete) [Moreira et al., 1997].Fractionating evaporation is therefore evaporation fromwater bodies (lakes and rivers). We set aside partial canopyevaporation where the remaining, enriched. water never

evaporates on the grounds that this is a somewhat implau-sible fate [Leopofdo, 1981). Delta values are given relativeto the Vienna SMOW (VSMOW) standard and obtainedfrom mixing ratios for the two stable water isotopes as:

I' = (_R__ 1)*1000.Rs 110 II

(I)

where RS'lIO\\ is 0.0020052 tor IXO,160and 0.00015576 forD/H and the isotope ratio for particular water flux is thetotal isotope in the flux divided by the total water flux.Components of the surface water budgets are evaluatedisotopicallv bv determininz the ~ otfsets (depletions or. .. IX ~,ennchments) of 0 and D (-H).

[7] Stable water isotopes can provide information on thesources and sinks of atmospheric moisture in the Amazon[Hotfil1un et al., 2003] and the internal recycling of waterwhere model validation is of increasing importance [e.g..Townsend et al., 2002). The failure of the GCM simulationsreviewed by Henderson-Sellers et al. [2002] to correctlyrepresent the relative seasonal importance of transpiration(nonfractionating) as compared to the fractionating evapo-ration seemed likely to be traceable to the land-surfaceparameterizations. Indeed, the land-surface schemesemployed in the GCMs' simulations reviewed by Henderson-Sellers et al. [2002] either did not have the capability toincorporate open water bodies or, if such an option existed, itwas not used in any available Amazon simulations.

[x] Specific caveats on the conclusions of Henderson-Sellers et al. [2002] included the following: (I) Monthlyisotope data to 1990 only were available and hence ana-lyzed; (2) the statistically significant wet season changesreported might be related to, or even exaggerated by, EINino-Southern Oscillation (ENSO) events or other climaticvariations that modify the Walker circulation and Intertrop-ical Convergence lone (TTf.Z) position and hence affect themoisture climatology of the Amazon; (3) no information ontluxes from simulated open water as a surface type in theAmazon GCM experiments was considered; (4) the selectedmodel sets analyzed by Henderson-Sellers et al. [2002]were found to be failing to correctly simulate the relativecomponents of transpiration and reevaporated canopy inter-ception in the Amazon dry season; and (5) no isotopetracking in the Amazon deforestation simulations wasreviewed. because none was available at that time.

[9] This paper reviews all of these topics as follows:(I) The monthly data are updated to 2000 for availableGNIP stations, and 1980s' daily data are reanalyzed forBelem and Manaus; (2) ENSO and other climatic variabilityimpacts are assessed; (3) the different representation ofevaporation terms in two land-surface schemes (lSM[Bonan, 1996] and ClM [Dai et al., 2003]) (incorporatinglakes explicitly) allows detailed analysis of simulated com-ponents of water recycling; (4) while isotope tracking is stillnot available in any Amazon deforestation simulations, an

Figure 1. (a) Schematic of the global hydrological cycle showing approximate depletions in ~IXO(and ~D in parentheses)as a result of the various processes. Depletions increase as altitude and distance from the ocean increase and as temperaturedecreases. Inputs of heavy isotopes from continental nonfractionating processes (e.g., transpiration and total evaporation ofcanopy-intercepted water) are also shown. (b) Amazon Basin showing the location of the isotope measurements used. Seecolor version of this figure in the HTML.

3 of 25

DI7104 \\CGUf'FIE Al\D HE'\DERSO'\J-SELLERS: STABLE WATER ISOTOPES DI7104

off-line simulation using a land surface scheme that includesisotope fluxes is assessed; and (5) a third IGCM's simula-tion of large-scale Amazonian water cycling is added to thetwo models reviewed by f 'uille et al. [2003a. 2003b].

[III] Our goal is to test the hypothesis that stable waterisotopes can now. or might in the future. illuminate, andperhaps separate. the impacts of natural and human-inducedvariability. To do this. we evaluate rep0l1ed mismatchesbetween isotopic simulations and observations of surface-atmosphere hydrological exchanges. We choose this focus.of land-atmosphere interactions. because the surface is thelocus of human activity and hence responses to climaticshifts have impacts on people here. As stable water isotopicland-surface schemes are just coming to be included inincreasingly comprehensive IGCMs. this paper serves as anexemplar of how investigations of the possible causes ofchanging hydrology could beneficially exploit stable waterisotope data and simulations. We focus on an area whereisotopic characterization initiated GCM analysis of land-atmosphere interactions: the Amazon Basin. and follow astaged investigation. Section 2 evaluates isotopic character-ization of surtace-atmosphere water fluxes as represented bycurrent land-surface schemes applied to the Amazon. Sec-tion 3 compares observations and simulations of large-scalehydro-climatic variations and applies isotopic analysis toolsto aspects of these basin-scale evaluations. In section 4 wecombine these learnings in a tentative assessment of recentisotopic perturbations in the Amazon.

2. Land-Surface Representation of Water andIsotopic Fluxes

[II] When assessing simulations of the impacts of defor-estation in the Amazon, Henderson-Sellers et al. [2002]noted that all GCMs to that date had neglected the possibleeffects of the large areas of surface water there. Since then.further assessments have emphasized the extent and impor-tance of open water areas_ Richey et al. [2002] found that inMay around 20% of the main Amazon river area is flooded.The combined effect of tributaries south of the Amazonreaching peak stage in April or May and those from thenorth peaking in June/July is to generate the maximumflooding in May (350.000 krrr', about 20% of the quadrantthey studied). The annual mean flooded area was deter-mined to be 250,000 km2 [Rich~l' et al., 2002). Flooding inneutral years was estimated by Foley et al. [2002] to befrom 20,378 to 170.070 krrr'. Foley et al. [2002] alsoexamined the impact of E~SO forcing on the extremes ofthe flooded area finding that. while both El Ninos andLa Nifias enlarge the minimum area (by +-2,483 and~7,049 krn ', respectively). La Nifias also increase themaximum while EI Nines diminish this extent. Overall.estimates of the maximum flooded area range from~350.000 km2 to about half this. ~

[12] In this section we employ two land-surface parame-terization schemes: lSM [Bonall. 1996] and ClM [Dai etal .. 2003] in off-line simulations of the Amazon [cf. Pitmanet al., 1999). Bonan et al. [2003] report that ClM greatlydiffers in its representation of some components of thehydrological cycle over the Amazon. Specifically. theClM (circles) (Figure 2) computes much larger canopyevaporation (which is fractionating if it is complete) than the

Transpirationa)100...-------------...,

o-20 L-..L-..L-....L......L---I.---J~.L._..&_ .•.••..•....•...___I.----I

1 2 3 6 7 8 9 1011121

Canopy Evaporation

4 5

b)120...----------------,

C\I

Es

-20L..-..L...-..J.......L......L---1.--J_L.-..L-..L-....L......L--J1 2 3 4 5 6 7 8 9 1011121

Figure 2. Two components of the evaporative flux(W m-2

) in the Amazon computed by two land-surfaceschemes (ClM and lSM) where ClM is open circles.lSM l by open squares, and lSM2 by dotted lines for(a) transpiration and (b) canopy evaporation [after Bonan etal., 2003]_ The much higher canopy evaporation fromClM is combined with reduced transpiration (and groundevaporation. not shown). Overall Cl.M's lH is lower thanthose of the earlier lSMs.

two earlier versions of a closely related scheme (lSM.squares and dotted line). Here we run both ClM andlSM forced by NCEP meteorology [National Oceanicand Atmospheric Administration-Cooperative Institute torResearch ill Environmental Sciences (NOAA-ClRES). 2003]for the AMIP II period. The version of lSM we use(ISOlSM) includes stable water isotopes [Ri/~)' et al ..2002).

2.1. Isotopic Fractionation of the Surface WaterBudget

[13] Henderson-Sellers et al. [2002] found it was notpossible to examine the Gat and Matsui [1991] conclusionregarding the fraction of recycled moisture from lakesdirectly in terms of the available GCM results. However.they found that the results of GCM simulations appeared tobe at odds with the conclusions of Victoria et al. [1991].The latter claimed on the basis of isotope analysis that

.:I of 25

DI7104

a)

-080~0 70-o

6050

40

3020

10

0-10

-2

:YICGUFFIE AND HE\:DERSON-SELLERS: STABLE WATER ISOTOPES D17104

6. Leaf waterCanopy waterSurface runoffSubsurface runoff

"f>b180V = aMDv=O

f>b180v = -5%0MDv= -60%0

a 2 64

b))(r>.Jv~euterium

ct ,\) excess~ .. ~ i?)( ()

~ r<-''\, ' .,;'0

(7, .-o '.- \+----------~~,Lil-,.....:-,.-:'-, '!'l<J.\e

moisture ~i/)0'O'recycling ~enhanced

-50

p¥2G2J slope = 8

, "", ,.~:,.. ....!<J.\e,. '" , ' evaporation~"o,.~

-100 e,":"', ' ' , ' loss enhanced

-150 -r---..•.---r----r---l-----.-20 -15 -10 -5 o 5

6'80 (%0)

Figure 3. (a) Simulations by ISOlSM of stable waterisotopic (6180 and bD) offsets tor deep soil drainage(subsurface runoff). soil surface runoff, intercepted canopywater, and leaf structural water for every point and everymonth in two 5-year off-line simulations for the AmazonBasin. In one case the enrichment of the vapor is set to zero(i.e., equal to VSMOW). and in the second case the vapor isdepleted by fixed amounts on the basis of the observationsreported in Table 3b. (The depletions were calculated asrepresentative of the basin by averaging the observeddifferences at Manaus between the mean vapor and meanprecipitation depletions in the top and lowest deciles inTable 3b. These average differences are /)180 = -5.1 %0andbD = -60.1%0. Here we use bfRO = -5%0 and ED = -60%0.)(b) Schematic of relationships between Ef80 and bD forvarious water components of large basins in comparisonwith the Vienna-Standard Mean Ocean Water (V-SMOW).The slope of 8 of the global Meteoric Water line (MWl) iscontrasted with the smaller gradient of residual waterremaining in lakes and rivers following local evaporation:the local evaporation line. A range of deuterium excess dvalues which show air mass origins and history is alsoillustrated (redrawn after Froehlich et al. [2002]). See colorversion of this figure in the HTML.

transpiration is the major source of recycled water in the wetseason while Henderson-Sellers et al. [2002] found that theGCM they analyzed, NCAR CCM\, which used BATS[Dickinson et al .. 1986] as its land-surface scheme, simu-lated transpiration as being very much more significant inthe forest's dry season budget of recycled water. Theisotopic data analyzed by Victoria et al. [1991] show adeuterium excess of 14%0in the dry season (June- Novem-ber) requiring significant input of recycled water from oneor more fractionating sources such as lakes and rivers [e.g.,Craig, 1961]. The CCM I-Oz model [lv1cGuf.!ie et al., 1988]failed to reproduce this component of the Amazon's hydro-logical cycle. so these results, while not absolutely contra-dicting the Victoria et al. [1991] findings of transpirationdominance in the wet season. seem to be in contention withthem. CCMI-Oz was found to have the vast majority (73%)of the recycling being by transpiration in the dry season,while the isotopic data suggest this dominance actuallyoccurs in the wet season.

[f4] A second means of GCM assessment employed byHenderson-Sellers et al. [2002] was to utilize the results ofGat and Matsui [1991] regarding the relative amounts ofwater recycled in the Amazon from fractionating and non-fractionating sources. By comparing t'XO and ED isotopicobservations with results from their steady state model ofthe central Amazon Basin, Gat and Matsui [1991] deducedthat 10%- 20% of the input precipitation is reevaporatedfrom fractionating sources (i.e .. sources where waterremains after evaporation such as lakes and rivers). 30%-40% from nonfractionating sources (e.g .. transpiring plantsand complete reevaporation of canopy-intercepted water),with about half of the total hydrological budget going torunoff Henderson-Sellers et al. [2002] noted that thesevalues differ from those shown in the CCM1 simulationsthey analyzed especially in terms of the relative proportionsof water recycling and running off. This method could beused to evaluate the hydrologic budget components ofGCMs especially if agreement can be achieved betweenthe Gat and Matsui [1991] proportions and those of laterresearchers [e.g .. McGuffie et al., 1995: Costa and Folev.2000]. ~ .. .

[f5] In this section we explore the questions of transpi-ration, reevaporation of canopy-intercepted water and othervegetation-water budget interactions, using two more recentvariants of the NCAR land-surface scheme. These are anisotope version of the lSM [Bonan, 1996; Rilev et al ..2002] and ClM [Bonan et al., 2003: Dai et ai.. 2003],which have different vegetation water budgets (Figure 2).The stand-alone simulations here are for illustrativepurposes only and are thus for periods shorter than the17-year AM1P 11predictions analyzed subsequently. Bonanet al. [2003] report in detail on the differences betweenClM and two earlier versions of the lSM.

2.2. Isotopic Tracking \Vith LSl\l[16] Rilev et al. [2002] describe the addition of water and

carbon isotopes to a version of the NCAR land-surfacemodel known as "lSM" [Bonan, 1996]. Here we employlS0lSMv1.2.1 [Riley et al.. 2002] in a series of off-linesimulations of the Amazon Basin's water cycling. lS0lSMis forced with NCEP reanalysis meteorology every 6 hoursand the isotopic characteristics of precipitation and atmo-

5 of 25

D17104 jVlCGUFF1E AND HE?'-iDERSON-SELLERS: STABLE WATER ISOTOPES

.-o~---o

CXJ

a)12108642o

-2-4-6

b)

• Leaf water

o Sap water

D17104

14.012.010.0 0-

8.0 ~C)CD

6.0 C/)C/)--.4.0 ~0--2.0

Carioba - R. Piracicaba 0.0

-3.0,-------------------------------,

-3.5-4.0-4.5-5.0-5.5-6.0-6.5

-7.0-7.5

-8.0-8.5

--.o~--oco

____ 618

0

___ d-excess

-2.0c.... c.... ):> (j) 0 z 0 c.... "Tl s: ):> s: c.... c.... ):> (j) 0 z 0 c.... "Tl s: ):> s: c.... c.... ):> (j)c ~ c co o 0 co OJ co ~ "0 OJ C C C co ~ 0 co OJ co ~ "0 OJ C s, C co::l eO to "0~ < o ::l 0- b '< ::l ~ to "0 < C( ::l 0- j' '< ::l eO to "0eO eO eO eO eO eO eO eO eO eO eO eO eO eO eO eO eO co I eO eO eOen OJ co co -...Ic.n c.n c.n c.n c.n c.n OJ OJ OJ OJ OJ OJ OJ OJ OJ OJ OJ

-...I -...I -...I -...I -...I -...I -...I -...I

Figure 4. Selected isotopic observations from plant and river water in and near the Amazon Basin.(a) bl~O values of bulk leaf water (striped) and sap water (shaded) of plants collected in the shore of LakeGrande in September of 2002. (b) bl~O and d excess observations from Piracicaba River for the periodJune 1995 to September 1997 (both after Martinelli [2003]) (see Figure 1b for locations).

spheric water vapor are prescribed as VSMOW. For conve-nience, we analyze the terrestrial results relative to (i.e.. as adifference from) these prescribed precipitation t I~O and bDvalues. The success with which the major controls on thevariability of 180 and deuterium are captured by ISOLSM isevaluated in the context of the Amazon.

[I"] Results presented here are based on a 5-year simu-lation using ISOLSM for the NCEP years 1998-2002. Thet offsets are reviewed for soil drainage (subsurface andrunoff), soil surface runoff. intercepted canopy water andleaf structural water. The first two of these were chosen toallow consideration of simulation of the isotopic character-ization of groundwater recharge and discharge into rivers.The latter two reservoirs are representative of the isotopic

exchanges occurring at the leaf scale as represented by a"big leaf' land surface parameterization scheme and, assuch, may provide insight into simulation of fractionatingand non fractionating components of the vegetation-derivedsurface water budgets. Two experiments were conducted: inone case, the enrichment of the \apor is set equal toVSMOW and, in the second case, the vapor is depletedby fixed amounts on the basis of the observations (see, e.g.,Table 3b).

[IS] For the two runoff components the basin-scale valuesOHI80 and tD in precipitation (Figure 3a) are little atfected(offsets are approximately zero). while for the two leafwater parameters there is isotopic enrichment by the land-surface scheme (i.e .. the calculated offsets are all positive).

6 of 25

017104 :VICGUFFIE AND HENDERSON-SEllERS: STABLE WATER ISOTOPES D17104

Table la. ClM Annually Averaged Partitioning of Surface Water Budget Components With the "Standard"Surface Types and With 20% Additional Lakes Around the Amazon River

Standard 200'0 1\lor~ Lak~s

Absolute, nun d I Percent of Precipitation Absolute. mm d-' Percent of Precipitation

Precipitation 1.71 1.71Evaporation 0.99 67 0.92 63Runoff 0.73 37 0.59 26ITotal) ( 104) (89)Canopy evaporation lUI 45 0.47 30Transpiration 0.07 6 0.05 4Soi I evaporation 0.21 \5 0.14 \0Open water evaporation 0.26 18

The former result is expected because surface runoff andsoil drainage share the isotopic characteristics of the pre-cipitation (here VSMOW) unless there is a mechanism forenrichment or depletion. The latter result reflects the frac-tionation during evaporation. The vapor sensitivity study.using VSMOW and relative depletions (appropriate for theAmazon Basin) of .ll'180 of -5%0 and .lbD of -60%0.reveals offsets (enhancements in os) which differ quitesignificantly. Figure 3a shows that the simulated deltadistributions of exO against bD for the Amazon for struc-rural leaf water and canopy-intercepted water are extremelysensitive to the prescribed values of water vapor isotopicdepletion. While enrichment in heavy water isotopes isclearly seen in the two vegetation components, the runoffparameters largely reflect the characteristics of the precip-itation (i.e .. offsets are near zero) for both vapor depletions.The two canopy reservoirs also show a near-linear relation-ship with slopes close to those of precipitation (the MWL)for vapor values set at VSMOW. However, the verydifferent gradient for the case of isotopically depleted vaporis similar to enrichment in an evaporative environment suchas a lake (Figure 3b).

[19] Figure 3a shows a tendency for the gradient to besteeper for interior (or structural) leaf water than for canopy-intercepted moisture for both vapor prescriptions. Thedeparture of the slope of the offset leaf water isotopic values(mean gradient for VSMOW vapor is ~12.5) from theMWL (and thus the precipitation imposed cs) with a slopeof ~8 indicates the importance of nonequilibrium processesin the simulated transpiration for this atmospheric watervapor prescription. Figure 3 underlines the fact that feed-back between the LSS and the surrounding atmosphere iscritical. Consequently. off-line experiments are very sensi-tive to the prescribed isotopic forcing, tor which no reliabledata exist on a basin or global scale. These results focusattention on the need for valid simulation of boundary layerisotopic enrichments in global atmospheric models if theyare to be coupled with isotope-capable land surface schemesfor meaningful investigation of vegetation effects on theisotope cycle.

[co] The site-specific and very spotty nature of isotopicdata render comparisons with basin-scale "big leaf' modelssuch as ISOLSM almost meaningless. However, it isimportant to demonstrate that, at the least. the depletionsin Figure 3a are reasonable. Figure 4 shows plant and riverwater isotopic data from Brazil. Figure 4a is of blXOdepletions and enrichments in sap and leaf water, respec-tively, collected from the Lake George shore in a short

recent campaign [Martinelli, 2003]. The enrichments in theleaf water range up to 10%0 in good agreement with thesimulations of ISOLSM. Figure 4b shows b180 and bD inriver water from the Piracicaba River in Sao Paulo state wellsouth of the Amazon Basin (Figure la) between 1995 and1997. These samples might be compared with soil waterrunoff values simulated by ISOLSM. In Figure 3a these areseen to be ",-,0%0which are not seriously different from theriver-based observations when it is recognized that the latterwill have been affected by evaporative fractionation fromthe river itself [e.g., Stolle et al .. 2003].

[zt] ISOLSM does not include (in the implementationsused here) either interaction with the host atmosphere orthe explicit effect of open water. As a consequence of thelatter. simulated river water. which would comprise thesurface and subsurface runoff in Figure 3a, must reflectvery closely the isotopic characteristics of the precipitation.In reality. however, lake and river evaporation playa rolein determining the isotopic enrichment of these reservoirs.The ISOLSM mechanisms of computation of the isotopicenrichment would result in similar (and rather small)modifications to precipitation characteristics for groundwa-ter and river input but much larger changes for vegetation-modified water components. In contrast, the very largesensitivity we have found to the prescribed vapor depletionsshows that simulations of isotopic enrichment in the canopy-intercepted water and leaf structural water can differ verygreatly from the prescribed precipitation. This degree of

Table lb. ClM Partitioning of Surface Evaporative BudgetComponents Into Fractionating and Nonfractionating With the"Standard" Surface Types and With 200;' Additional lakesAround the Amazon River for the Annual Average, the CentralBasin Wet Season (January- March). and the Dry Season (April toAugust)"

Standard 20% More Lakes

Percent ofEvaporation

Percent ofEvaporation

AnnualNonfractionaring 7 5Fractionating 92 95

Wet season (March-May):--.Jonfractionating 3Fractionating 94 96

Dry season (August-October)Nonfractionating 15 IIFractionating 84 89

"The location of the modeled grid point is roughly at Manaus.

7 of 25

017104 \lCGUFFIE A]\;DHE:\DERSON-SELlERS: STABLEWATERISOTOPES 017104

Table 2a. Cl\1 Partitioning of Surface Water Budget Components With the "Standard" Surface Types andWith 20°0 Additional lakes Around the Amazon River for January-March (\VetSeason)

Standard

Absolute, 111111 d-I Percent of PrecipitationAbsolute. 111111 d-I Percent of Precipitation

Precipitation 2.2R 2.28Evaporation 1.05 46 0.95 42Runoff 1.22 53 I 13 50Total 99 91Canopy evaporation a.82 36 055 24Transpiration a.05 2 a.03 ISoil evaporation a.17 R 0.12Open water evaporation 0.25 II

sensitivity is particularly surprising and needs careful con-sideration before isotope lSSs and isotope host atmosphericmodels are coupled.

2.3. Partitioning of Reevaporated Water in CL\1[22] A recent version of the NCAR Community Climate

Model's land-surface scheme. Community land Model(ClM). can also be operated in an off-line mode [Bonanet al.. 2003]. Here we use the ClM land-surface scheme toimprove on the comparisons made by Henderson-Sellers etal. [2002] between isotope observations in the Amazonforest and simulations of the components of the surfacewater budget. Specifically. we have modified the percentageof open water surfaces in the region around the AmazonRiver itself in keeping with the data of Richey et al. [2002]and Foley et al. [2002]. Tables Ia and Ib list the compar-ative hydrological components when using the standardsurface as compared with adding 20% more lakes. Theeffects of evaporation from bare soil are minimal in thesetropical environments.

[23] With the standard ClM land types. the percentagesare improved over CCM I-Oz (BATS) as reported byHenderson-Sellers et al. [2002] as follows: (I) 51% ofprecipitation is recycled by canopy evaporation (45%)plus transpiration (6~o) annually: and (2) 15% of precip-itation is recycled by soil and lake surface evaporation(Table la). When 20% extra lakes are added to thesurface in ClM. it is found that (1) the fractionatingcomponent decreases (slightly) from 60% (45 + 15%) to58% (30 + 10 + 18~/o) of precipitation: and (2) thenon fractionating component decreases as a percentage ofevaporation from 7% to 5% (Table Ib).

[24] The ClM results are in stark contrast to the findingsof Victoria et al. [1991]. In ClM. transpiration is never alarge component of total evaporation. ranging from 15.7%

(0.15/0.95) in the dry season to 4.7%, (0.05/1.05) in the wetseason (Tables 2a and 2b). Also. the addition of 20~o extralakes [e.g .. Richey et al., 2002] increases the input offractionated water in the dry season (around April-August)(89% compared to 84~o without lakes). which is not fully inagreement with the interpretation of the deuterium excessobservations of Victoria et al. [1991] (Table Ib). Tables Iand 2 show that ClM somewhat improves the overallpercentages of fractionating to non fractionating sourcesover those reported from BATS [Henderson-Sellers et al..2002]. The transpiration t1uxes seem to be greatly under-estimated by ClM (Figure 2 and Bonan et al. [2003]) ascompared to the earlier LSM schemes. Even with extralakes. ClM cannot capture the magnitude of the recyclingof fractionated water from lakes in the dry season impliedby the isotopic observations.

2.4. Summary[2:'] In this section. two aspects of simulations relevant to

our hypothesis regarding isotopic characterization of Ama-zonian land-atmosphere exchanges have been considered:the impact of additional surface water (i.e., lakes and rivers)and the inclusion of stable water isotopes in lSSs. In bothinvestigations the LSSs have been used in an off-line (orstand alone) mode, i.e .. without feedback to the atmosphere[e.g., Chen et al.. 1997]. and for this reason we haverestricted our simulations to 5-year periods. although thislimits the impact of interannual variability.

[26] The Amazon isotopic enrichments predicted byISOlSM are plausible: the drainage to groundwater exhibitslittle change from precipitation values (in agreement withFigure 4b). while the enrichments (positive offsets) incanopy-intercepted water are similar in magnitude to thedepletion differences observed between precipitation andambient atmospheric water vapor at Manaus and Belem

Table 2b. ClM Partitioningof Surface WaterBudgetComponents With the "Standard" Surface Types and With20% Additional lakes Around the Amazon River for April- August (Dry Season)

Standard

Absolute. 111111 d-I Percent of PrecipitationAbsolute. 111111 d -! Percent of Precipitation

Precipitation 0.89 0.R9E vaporation 0.95 107 0.93 105Runoff 0.11 12 -0.14 -16Total 119 R9Canopy evaporation 0.57 6-l 038 43Transpiration 0.15 16 0.10 IISoil evaporation 0.23 26 0.16 lROpen water evaporation 0.29 JJ

R or 25

D17104 MCGUFFIE AND HEr\DERSO:-';-SELLERS: SHBLE WATER ISOTOPES DI7104

a)10I

ef_100

~.c ~~ ~'t~

-20 I + .t "- ~-30 LI ---L ----"- ----'~ "---- ...L_ __'___ ~_

o 100 200 300 400 500 600 700100 ~I=: ~ CI~o ~ ~'-~ OO/llD)~ c{'-Q&J OUOOO

~ -100 ~~~~~~-f4:~{~V'tj;;X_~±-+I -* *" + ...;..-

.200 LI ---L -----l L- L- .L- ~ _

o 100 200 300 400 500 600 700300 r-----,---------.----~-------,----~---~----r___;

o 100 200 300 400 500 600 700

~::~~~I '"""I.I""'~U~IIL]o 100 200 300 400 500 600 700Day number from 1 Aug 1978

b)I I I I I I I I I I I I I I I I I I

8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7Month

10

0000 -10.c

·20

-300 100 200 300 400 500 600 700

1°Ofo OD~O O~~ . CD. 0 0 '00 0

0

-100 ('* ~~ ~~"-¥':ftf.: -¥"t -H- ++ +++ ++ + f

-200 I I . I

0 100 200 300 400 500 600 700300

o 100 200 300 400 500 600 700

o 100 200 300 400 500Day number from 1 Aug 1978

600 700

I I I I I8 9 10 11 12 1

I I I I I I I I I I2 3 4 5 6 7 8 9 10 11 12

Month

I I2 3 4 5 6 I7

Figure 5

9 of 25

D1710-t \lCGUFF1E A"iO HE:\OERSO!\:-SELLERS: STABLE WATER ISOTOPES D171O-t

(Figure 5). However. Figure 3 demonstrates that the I'IXO v.I'D gradient of these vegetation parameter enrichments isvery sensitive to the depletions prescribed in the atmo-spheric vapor. Without feedback to the atmosphere (i.e., acoupled isotopic land-atmosphere simulation). it is notpossible to pursue the likely impact of this sensitivity. butour results suggest that valid representation of isotopicenrichments of water vapor in the atmospheric boundarylayer will be critically important in coupled isotopesimulations.

[ec] The Cl.M's partition between isotopically fraction-ating and non fractionating evaporation is sensitive to theaddition of (a plausible) 20% open surface water. Althoughthe Cl.M's representation of water recycling is better thanthat of BATS (used by Henderson-Sellers et al. [2002] inCCM I-Oz) there are still differences between the size ofthe simulated components of total evaporative flux andthose deduced from isotopic measurements in the Amazon(Tables I and 2). However. as with ISOlSM, a morecomplete interpretation of these lSSs' sensitivities to likelyAmazonian conditions is unwarranted from off-line integra-tions alone.

[eX] This preliminary investigation of two lSSs hasshown that plausible isotopic enrichments and depletioncan be generated and that the partition of evaporative fluxescan be evaluated using isotopic measurements. It seems thatthese two results could be combined so that isotopiccharacterization can be employed to assess the relativeimpact of human-induced and natural climatic variationson large basin hydroclimates. In the next section. we reviewthe large-scale hydro-climatic variability of the Amazon, itsrepresentation in GCMs and IGCMs and its likely impact onisotopic fluxes. Assuming that the atmospheric simulationsof isotope distributions are robust, this is the second stepalong the path to determining whether stable water isotopesmight be exploited in tracking and finger-printing climaticdisturbances.

3. Amazonian Hydro-Climate Variability[eq] Chen et al. [2001] have described the impact of El

Nino-Southern Oscillation (ENSO) on the Amazon's large-scale moisture convergence. An EI Nino is drier and wannerthan normal in the Amazon. while La Nifias are wetter andcooler. Amazonian temperature variations [Foley et al..2002] are quite strongly correlated with the SouthernOscillation Index (SOl): the coefficient of correlation be-tween annual mean temperature and annual mean SOl is-0.61. Similarly, basin-wide precipitation is moderatelywell correlated (r = +0.50) with the SOl with the averageEI Nifio precipitation being 83 mm yr-I (4.0%) drier thanneutral conditions. and the average La Nina 64 mm yr-](3.1 %) wetter. The observed temperature changes tend to bespatially homogeneous and widely uniform throughout thebasin and the year but hydrological responses (drier ElNines and wetter La Ninas) are stronger during the wet

season (January - February - March) and are concentrated inthe northem and southeastern portions of the basin [Fo/~\' etal.. 2002].

[)o] However. Botta et al. [2002] used principal compo-nent analysis of Amazonian climate data to show that ENSOis not the major. or even the most important. mode ofclimate variability in the Amazon Basin. They found thatENSO explains only 21 % of the total variance in annualmean precipitation and temperature and that the dominantmode of climate variability in the Amazon, which explainsabout 35% of the interannual variance of precipitation and56% of the temperature variance, has a period of 24-28 years. Other proposed drivers of climate variability inAmazonia include changes to the strength of the NorthAtlantic high-pressure system, the position of the intertrop-ical convergence zone (ITCZ). and wind stress and seasurface temperatures in the tropical Atlantic [e.g., Foley etal.. 2002]. Despite this, it seems worthwhile to try toidentify, or eliminate, the possible effects of ENSO vari-ability on isotopic measurement and simulations in theAmazon Basin.

3.1. Daily Isotope Analysis for Example "Neutral"Years

[3 I] Between August 1978 and JUI~ 1980. Matsui et al.[1983] measured daily values of I' "o and I'D in bothprecipitation and atmospheric water vapor at Belem andManaus. These data coincide with a quiet time in E:\SOterms: none of the period is rated as either a "warm" or"cool" (EI Nino or La Nina) climate by the NationalOceanic and Atmospheric Administration (NOAA) [2003]although the January/February/March 3 months of 1980 arenoted as "'W-." Notwithstanding this slight anomaly. wehere designate the full period as "neutral," i.e., classifiedneither as El Nino nor La Nina.

[3e] The expectation, in a "neutral" period such as theone selected, is that the Amazon's meteorology and hydrol-ogy will be dominated by easterly winds bringing moistureevaporated from the tropical Atlantic into the basin andupstream (Figure 5). Precipitation events will deplete thesemoist air masses of heavy isotopes leaving the atmosphericvapor more depleted than the rainfall and thus the inland site(Manaus) more depleted than Belem at the mouth of theriver (Figure Ib).

[3)] Figure 5 shows these daily isotope data (6180 andED depletions in precipitation and atmospheric watervapor) and corresponding daily mean wind direction fromNCEP reanalysis [NOAA-CIRES. 2003]. Easterly windsdominate at both locations, becoming more northeasterlyat Manaus. Winds at Manaus exhibit a greater seasonalitybut. overall. the resulting climatology is dominated bywinds from the east. As would be expected, vapor ismore depleted than the precipitation and wet seasondepletions are larger (more negative) than those in thedry season. Figure 5 also shows the seasonal change inprevailing wind direction from east and northeast in the

Figure 5. Daily values of bIXO and ED depletions in precipitation (circles) and water vapor (crosses) for (a) Belem and(b) Manaus from August 1978 to July 1980. together with wind direction (from 850 mbar of daily mean '-ICEP reanalysisfor the same period) and daily precipitation amount in mm (data from Matsui et al. [1983] and from NOAA-CIRES [2003]).See color version of this figure in the HTML.

10 of 25

D17104 "vlCGUFfIE AND HE'-:DERSOl\-SELLERS: STABLE \VATER ISOTOPES D17104

Table 3a. Numbers (11) of Daily Observations From Belem andManaus far Atmospheric Vapor and Precipitation for the Period IAugust 1978 to 31 July 1980 of a Maximum Possible of 731Together With Means of 1'1'0.1"0. and d Excess (dxs)

Atmospheric Vapor Precipitation Observations

61'0 m 61'0 son Mean n Mean dxs n Mean n Mean dxs

Belern 374 -10.7 355 -77.4 11.1 436 -1.9 299 -4.2 9.7Manaus 233 -10.6 209 -75.4 12.0 198 -3.5 157 -13.8 15.3

wet season to easterly in the dry. Some indication ofannual and interannual variability can be seen in thegraphs, for example the wet season, between Januaryand May, is characterized by large depletions, but theuneven data record, especially at Manaus (Table 3a),makes quantitative evaluation of either difficult. Matsuiet al. [1983] concluded the close correspondence betweenevents at Belem and Manaus during this ENSO-neutralperiod indicated the dominant role of weather systemstraveling in an east to west direction during this time. AtBelem, precipitation in the wet season of 1980 (days 500-600) is more depleted than that of the wet season of 1979(days 200- 300). but there is no matching differenceevident at Manaus.

[34] The use of deuterium excess as a tool in the analysisof the isotopic characteristics of precipitation has beenproposed since it reflects the conditions prevailing duringthe development of an air mass on its route to the precip-itation site (Figure Ia) [Froehlich et al.. 2002]. Values of thed excess greater than the MWL value of 10 can beinterpreted as evidence of moisture recycling during theevolution of the air mass (Figure 3b). This effect isillustrated in Table 3. which shows an increase in d excessin precipitation from Belem (9.7%0) to Manaus (15.3%0)indicative of recycling processes.

[35] The detail of the daily data offers an opportunityto examine the possible impacts of climatic excursions(e.g., due to ENSO or deforestation) on the isotopicrecord. Specifically. combining daily NCEP data withthe observational record shown in Figure 5 allows thegeneration of relationships between wind speed anddirection and total rainfall. Table 3b shows the "wet"and "dry" elSa and eD depletions in precipitation andvapor at both Manaus and Belem derived from theneutral period daily data. The "wet" and "dry" amountshave been generated by selecting the I0'1~ of days withthe highest precipitation and the 10'% of days with thelowest precipitation at both locations. Clearly the impactof rainfall amount is large, particularly for the depletions

measured in the precipitation where differences betweenthe top \0% and lowest 10% are statistically significantexcept i'D at Manaus.

3.2. Assessing Simulations of ENSO Variability in theAmazon Using AMIP II

[36] Our main interest is the use of isotopic data incharacterizing natural and human-induced hydro-climatechanges. In order to assess isotopic changes at the landsurface. particularly the partition of precipitation into frac-tionating and nonfractionating evaporation or runoff it isnecessary to eliminate as many other variables as possible.To do this. we choose a carefully controlled experimentwith a time period that encompasses both extremes of theENSO cycle.

[37] The Atmospheric Model Intercomparison Project(AMIP) comprises two experimental phases. During thesecond phase (AMIP II), the experimental design remainsfundamentally the same as AMIP 1 (i.e .. commonly speci-fied radiative forcing and ocean boundary conditions). butthe simulation period has been extended by 7 years (from1979-1988 to 1979-1995) and greater emphasis has beengiven to initialization/spin-up of soil moisture stores andconservation of continental surface energy and water. Also amore extensive set of land-surface variables is reported inthe AGCMs' output [e.g., Henderson-Sellers et al., 2003b].

[38] Of the 30 plus AGCMs participating in AMIP II, 20have had their simulations released under the protocol ofAMIP II. These AGCMs incorporate land-surface schemes(LSSs) of varied complexity ranging from the very simple"bucket" [e.g., Manabe, 1969] model to detailed soil-vegetation-atmosphere transfer schemes [e.g., Costa andFoley. 2000]. Differences among the land-surface predic-tions from AMIP II are being analyzed under the auspices ofDiagnostic Subproject Number 12. which is focused ondetermining the dependence (if any) of land-surface varia-bles (particularly latent and sensible heat fluxes) on LSScomplexity [e.g., Henderson-Sellers et al., 2003a; Irannejadet al., 2003]. The AMIP II period (I January 1979 through IMarch 1996) includes four EI Ninos (1982-1983, 1986-1987, 1991-1992, and 1994-1995) and two La Nifias(1984-1985 and 1988-1989). These climatic variationsare prescribed for the participating AGCMs because seasurface temperatures (together with sea ice amounts, atmo-spheric composition, and solar radiative forcing) are setfrom observed values in the AMIP experiments [Gates etal., 1999] (AMIP home page, available at http.z/www-pcmdi.llnl.gov/amip/amiphome.html. 2003).

[39] Figures 6a-6e shows the monthly mean surfacehydrological components averaged over the Amazon Basinas simulated during AMIP. The full 17-year average is con-

Table 3b. Means and Standard Errors of fo150. foD. and d Excess (dxs) for Upper and Lower Deciles of Precipitation Days forPrecipitation and Water Vapor Isotopic Depletions for Belem and Manaus

.'1'0 \leanIS.E.l. "bo I'D Mean .s.t.; 00" dxs Mean. "00 1'1.'0 Mean IS.E.). %0 .'D \leanIS.E.). "b" dxs Mean. "0"

BelernTop decileBottom decile

ManausTop decileBottom decile

-13.010.8)-11.010.2)

-9001(7)-78.712.0)

-12.011.4)-7.0 (1.8)

-83.3(11.4)-73.3 (- 7.0)

II of 25

3.221.2

-4.3 (0.5)-0.611.8)

-25.815.2)4.512.4)

-1.817.2

-2.928.8

-6.611.2)-2.2 (0.9)

-0.843.7

-21.4 (5.4)-13.5 14.9)

017104 \ICGLFFIE AI\D HE\DERSOl\-SELLERS: STABLE WATER ISOTOPES 017104

a) 300 A B C D E

100

200

100

o L.-o- •.••••.••~ •.•.••••.•...•.••..•.-........I

300

200

OL.............L..~...•....•....•......•.---'--'

E 300Ec 200o:;::;~ 1000-'0Q)•....

Q..

>-..c•...co~cco~ 0 L.-o- ••••••.•••~ ••••••.•••••••••.••••••••

300 Mean AGCMs CMAP ,A DOE.•... NCAR ECMWF

100

••200

o L.-o- •.••••.••~ •.•.••••.•...•.••..•.-........I

300 VIC4 7 10Month

4 7 10Month

4 7 10Month

100

200

OL.............L..~...•....•....•......•.---'--'1 4 7 10

Month

-- 17-year Mean• - .• EI Nino,A ..A La Nina

1 4 7 10Month

Figure 6. AMIP II monthly mean hydrology components: (a) precipitation, (b) evaporation, (c) runoff,(d) moisture convergence (PI' - Evi, and (e) evaporation ratio (EI'Pr) for 20 AMIP models (A-T). AMIPmean, available observations. NCEP-DOE, NCEP-NCAR, ECMWF. VIC, and MUGCM. In all plots thefull l7-year average is compared with El Nino and La Nina periods.

12 of 25

D17104

b)

EEca:;:;ct:l"--aQ.ct:l>W>-.r::+-'ca~cct:lQ.)

~

MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES 017104

200 ........--...--.-~........--...--.-....,

150100

A 8 C D E

50OL......- .•......•...~...L...............J.-----.J

200 ........--...--.-~........--...--.-~F G H J

150100

50o ~--'-~-'--'"...J

200 r-r-......-~ ...•..•....•.....•--.--,

K L150 M oN

100

50o .o-L.~ •..•.•..••••...•.•..•••••.•.••••

200 r-r-......-~ ...•..•.....,......,--.--,

p Q R s T150100

50O~"""""~...L....."""""'''''''''''200 ....-...•..~....-...•.....,

150OBSERVATIONSMean AGCMs

100

DOE NCAR ECMWF

50OL......- .•......•...~...L...............J.-----.J

200 ....-...•..~....-...•.....,VIC mugcm

4 7 10Month

4 7 10Month

4 7 10Month

150

100 17 -year Mean• - .• EI NinoA·· .. ···A La Nina

50OL......- .•......•...~...L..............J.-----.J

1 4 7 10Month

4 7 10Month

Figure 6. (continued)

trasted with the El Nino and La Nina periods as predicted by20 AMIP II AGCMs, their mean, three reanalyses (NCEP-DOE [Kanamitsll et al., 2002], NCEP-NCAR [Kistler et aI.,2001], and ECMWF), the VIC scheme [Liang et a!', 1994],and the Melbourne University AGCM (MUGCM), which is

an isotope GCM [Noone and Simmonds, 2002]. The hydro-logical components are precipitation, PI' (Figure 6a), evap-oration, Ev (Figure 6b), runoff, Ro (Figure 6c), atmosphericmoisture inflow (PI' - Ev) (Figure 6d), and the evaporationratio (EdPr) (Figure 6e).

13 of 25

D17104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES

c) 160,.........."""T""""''''''''''-''''''''''''''""T""""'''-'

120

8040

o-40 •...............•...~....o....;......•....•••....•....•

160 •....•...."'"""'T'"~""T"""""...•....•.......,

120

8040

o-40 L....-•....•....~L........ •....•....•....•160 ,.............,-~,.................-...,

120

8040

o-40 •...............•...~""O"";" ••••••••••••••••••••

160 •....•...."'"""'T'"~""T"""""...•....•.......,

120 Q

[~o] There is most agreement in the simulation of rainfallseasonality and atmospheric moisture convergence (at thebasin level) with, arguably, least agreement in the runoff(Figure 6c) and evaporation ratio (Figure 6e). Precipitation(Figure 6a) and moisture convergence (Figure 6d), beingclosely related to ENSO-driven circulation changes, exhibit

A B

F G

EE K.•....•...oC:::Ja:>-..c+-'co~ p

c~ 80Cl>~ 40

o-40 L....-.•......•...~ .•.....•...•......•.......J1 60 •....•........-r~ ....••...•.--.--........,

120

8040

o-40 •...............•...~....O....;.....•.....•.•..•....•

160 •....•...."'"""'T'"~""T"""""...•....•.......,

120

8040

o-40 L....-.•......•...~ .•.....•...•......•.......J

1 4 7 10Month

OBSERVATIONSMean AGCMs

mugcmVIC

4 7 10Month

D17104

ED

.• H

o

ECMWFDOE NCAR

4 7 10Month

4 7 10Month

4 7 10Month

17-year mean• - .• EI Nino•.......•. La Nina

Figure 6. (continued)

the largest sensitivity to ENSO forcing, with the NCEP-DOE reanalysis and the MUGCM being the most sensitiveto El :\ino and La Nina events. This result is not toosurprising as VIC is "partitioned" from distant climaticevents and the AMIP mean is an average of twenty models.However. none of the AMIP models exhibit as much

14 of 25

D17104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES D17104

d) 250 .......-...--.-~""T"""""~--.--,

20015010050o

-50 L......L- .•....•••.. ~ ....•......•.....•......••---.....J250 .......-...--.-~""T"""""~--.--,

20015010050o

-50 •..........•.....•...~ ....•..................•....•....•250 ....--•.....•~...-- •.....•...,

20015010050

o-50 L.o- .•.....•...~.L............J..........J250 ......--...,..~,............,......,

20015010050 •...--~---......~_ •.__

o-50 L...-..- •.••••....•~L.........•....•..........J250 ......--...,..~,-.--...,.......,

20015010050o

-50 L.........•....•....•~L......... •....•....•.....J250 ....--•.....•~...-- •.....•...,

20015010050

o-50 L......... .•.•..•••..•~L.........•.....•.........J

C o EA B

F H JG

EEco•...m"-o0..m>wI

Co•...m•...'0.'(3(])"-o,>...c•...co~cm(])

~

K

p Q R s T

DBSERV ATIONSMean AGCMs DOE NCAR ECMWF

4 7 10Month

4 7 10Month

4 7 10MonthVIC

17 -year Mean• - .• EI NinoA·······A La Nina

4 7 10Month

4 7 10Month

Figure 6. (continued)

sensitivity to E\SO conditions as the reanalyses: resultsless easily explained.

[41] The evaporative fluxes (Figure 6b) differ quite con-siderably with most models exhibiting very little seasonalityand very little ENSO response. Some models have asignificant seasonal cycle in tune with VIC. Although the

MUGCM has both evaporative seasonality and sensitivity toENSO. its values overall seem too small. The variety ofrunoff simulations is considerable (Figure 6c). Some modelshave very low seasonality of runoff (like the reanalyses).whereas others have seasonality of runoff comparable to themoisture convergence (PI' - Ev). Differences between the

15 of 25

017104 :'vICGLJFFIE AND HEi\DERSO?\-SELLERS: STABLE \VATER ISOTOPES 017104

e) 2 A B • c • o ~

0

2 00+-'roa:c0".+=ro 0•...0Q.

Qro 2 p R T>W>-..c+-'c ... """ -0 - 'K -'.~c 0roQ)

2 Mean AGCMs OBSERVATIONS DOE NCAR ECMWF~

OL..............L.~..L.............JL.........J

2 VIC mugcm4 7 10

Month4 7 10

Month4 7 10

Month

17 -year Mean• - .• EI Nino..•..........•.La Nina

4 7 10Month

4 7 10Month

Figure 6. (continued)

seasonality of moisture convergence and runoff are indica-tive of moisture storage (and in extreme cases. of noncon-servation of water). The evaporation ratio. shown inFigure 6e, displays the greatest intennodel differences inseasonality and in sensitivity to El Nino. Model P shows

virtually no seasonality or sensttivrty whereas Model Ldisplays a large seasonality and sensitivity.

[4c] The AMIP simulations do seem to capture an overallENSO response with La Nifias being wetter (more rain in.e.g., A, D, G. and NCEP-DOE) than the average. There is

16 of 25

D17104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES D17104

Table 4. Annual Surface Hydrologic Summaries From the AMIP II and MUGCM Models Together With aVariety of Estimates of "Truth" for the Global Land Surface (GLS) and the Amazon Basin

Source Pr. mm d 1 h!Pr Ro!Pr

GLSGRDCCMAP 1.94 0.60 040NCEP DOE [KallalllilslI et al.. 2002] 240 0.79 045NCEPI\CAR [Kislla 1'1al.. 2001] 2.37 0.77 0.53ECMWF [GihsvlI 1'1 al.. 1997] 2.21 0.72VIC [.\ijSWII CI al.. 2001] 2.07 0.67 0.33MUGCM [SOOIll' and Simmonds. 2002] 2.46 0.47 0.53A\ A\cIIP II 0.67 0.33[SD .HI!P 11 0.31 0.52 O.06j

AI1UCOl1

GRDC C\CIAP 5.00 0.39 0.61NCEP DOE [KallalllilslI 1'1 »: 2002] 571 0.68 0.43NCEP NCAR [Kislla 1'1 al.. 20011 5.93 071 0.39ECMWF [Gihsoll 1'1 al.. 19971 5.38 0.59VIC [.YijsSI'1I 1'1 at.. 200 I] 5J)9 0.67 0.33MUGCM [.\00111' and Simmonds. 2002] 5.26 0.39 0.61A\ Ar-..IlP II 4.97 0.67 0.34ISD A.\1IP 11 0.68 OJ)? O.05}

also weak agreement with the Faler et al. [2002] runoffresults in that some models (e.g., I, N, and S) show lowerrunoff during EI Nifios, especially in the early part of theyear. There is also a large variation among these AMIP IImodels, which may affect their simulation of the surfacehydro-climate and hence the extent to which they, andisotopically enabled versions of these GCMs. can correctlycharacterize surface-atmosphere water fluxes.

[.13] Table 4 summarizes the annually averaged hydrolog-ical components over the Amazon from the 20 availableAMIP II models together with observations of runoff(GRDC [Fekete et al., 2000]) and precipitation (CMAP)from the Climate Prediction Center [Xie and Arkin. 1997].As seen in Table 4, globally and in the Amazon Basin,NCEP reanalyses fail to conserve water. ECMWF [Gibsonet al.. 1997] may also have similar problems but we wereunable to assess this because we could not acquire its runoffdata (NRD). VIC [Nijssen et al., 200 I] is seen to conservethe surface water everywhere because it is constrained byobserved precipitation and tuned for large river flows.Henderson-Sellers et al. [2003b] argue that VIC providesa reliable surface water simulation, at least when averagedover a large area and a long period of time [see Wood et al.,1998].

[-1-1] In the Amazon. all reanalyses and AMIP models'estimated runoff ratios are smaller than observations(Table 4). Considering the relatively high available energyand dense vegetation canopy of the catchment, an observedmean runoff ratio of about 61% is arguably too high.Investigation reveals that the GRDC mean runoff over someareas of the Amazon Basin is greater than the CMAP meanprecipitation (Figure 7a). Comparing GRDC with the GCPC[Huffinan et al .. 1995] and Legates and Willmott [1990]precipitation climatologies provides similar results. Com-parison of the mean seasonal cycles of runoff, Ro, andprecipitation, PI', for areas of Ro 2': PI' (Figure 7b) showsthat excess runoff in comparison to precipitation is espe-cially large during the high precipitation period when it isexpected that some of the precipitation will be stored in thesoil to supply evaporation and slow drainage in the rela-tively drier months.

[-15] The closure of the simulated land-surface waterbudget is a fundamental requirement for land-surface rep-resentation. Thus

PI" - Ev - Ro - dW !dt = O. (21

Assuming that the rate of change of surface water storagewith time (t) (i.e., dldt of soil water. canopy water andsnow and ice) is close to zero over long periods (e.g., the17-year AMIP II simulation), the water imbalance (Wirnb)can be defined as the residual of precipitation less runoffand evaporation. If this term is small, that is, less than±0.05 mm d I. water is conserved. In Table 5 it can beseen that six of the AMIP II models (8, C, D, G, M, andR) do not conserve water at the land-surface adequately tobe considered of value in hydrological assessments.

[-16] The simulation of total precipitation is generallyreasonable except for Model D, which has too little rainfall.as well as failing to conserve water. The partition of theincoming precipitation into evaporation or runoff is verydifferent among the AMIP AGCMs. For example, Model Aputs less than 25% of its rain into runoff while Model L hasalmost 45% runoff This full AMlP range is significantlyless than the 610,'0 reported by GRDC, itself questionable[e.g., Henderson-Sellers et al.. 2003b]. Similarly, Model Mreevaporates just over 50% of its rain while, at the oppositeAMIP extreme, Model D reevaporates almost 85% and allare larger than the derived observation of 39%. As both ofthese fail to conserve surface water adequately. the range forAMIP models which do pass the conservation test is lower(55% to 75%).

[-17] The startlingly large range in simulated water parti-tioning shown in Table 5 suggests that characterization ofhydro-climatic changes using stable isotopes. or indeed byany other means. will be very difficult if not impossible fora number of these GCMs in the Amazon. Indeed. theanalysis performed here shows that in the Amazon Basina few of the AMIP II AGCMs and well used reanalyses failto close the surface water budget to within an acceptablemargin of =0.05 mrn .r'. Since isotopic analysis involvestracking hydrological fluxes and reservoirs several orders of

17 of 25

017104

a)

3N

EO

35

65

95

125

155

18S

215

8SW

b)15

MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES

\e---e Precipitation

\ .--. Runoff\\\

10 \'i .,

"0 <,

E ••,E ,, /X -, /:::I /u:: ,

5 '.<, /, /.,/

<, ••.... /....• /•••0

1 2 3 4 5 6 7 8 9 10 11 12Month

Figure 7. (a) Geographical distribution of the observedmean PI' - Ro (mm d-I) in the Amazon (Ro > PI' in theshaded area) and (b) the mean seasonal cycle of PI' and Ro(mm d-I) for one grid square in the shaded area. All dataare drawn from global observational sets of runoff (GRDC[Fekete et al., 2000]) and the CMAP precipitation from theClimate Prediction Center [Xie and Arkin, 1997].

magnitude smaller than those in traditional hydrologicalmodels, very much more benefit can be gained by dealingwith water budget closure in the gross hydrology beforeproceeding to address isotopic fractionation processes.

[-Ix] Among the models that do not close the waterbalance, some have not followed the AMIP protocol forreporting values of the required variables [e.g., Henderson-Sellers et al., 2003b]. Despite the emphasis in the AMIP IIprotocol, soil water initialization is still a problem in majorhydrological regions such as the Amazon. Problemsconcerning the lack of reliable global observations forevaluating land-surface simulations persist. The nonclosureof the surface water budget and very different evaporationratios from observations suggests that reanalyses are notappropriate tools for evaluating simulated surface waterbudget components. Furthermore, the GRDC runoff data

017104

set is found to be problematic in the Amazon and is notconsistent with global precipitation data sets. Owing toconstrained atmospheric forcing. off.-line land-surfaceschemes. such as VIC, close the surface water balanceand simulate evaporation ratios better than any reanalyses.but in some regions runoff ratios are very different fromobservations although these themselves may be question-able [Henderson-Sellers et af.. 2004].

3.3. Summary[-19] In this section. simulations of Amazonian basin-scale

hydro-climate variability have been examined over theAMIP II period and observations reviewed in the contextof detailed daily observations of stable water isotopes inprecipitation and near-surface vapour. As isotopic deple-tions are a strong function of precipitation amount butexhibit weaker relationships to wind speed and direction(Figure 5). isotopic characterization and model interpreta-tion will be incorrect if precipitation is inadequately pre-dicted either seasonally or under climatic variations such asEN SO-driven circulation shifts (see Figures 6a-6e andTable 4).

[50] Despite limitations in observations. which make ab-solute determination of correctness difficult (e.g .. Table 4 andFigure 7). some global model simulations are found to be toopoor to be of value (Table 5) for isotopic application andinterpretation. Overall. ENSO forcing causes responses inmany of the AMIP IImodels. the MUGCM, VIC and thereanalyses (Figures 6a-6e). These sensitivities are almostalways small relative to the seasonal cycle and generally inthe expected directions, i.e., the models' simulated sensitiv-ities agree with the observed impacts of EI Nino/La Ninaevents. The isotope AGCM (MUGCM) performs fairly wellbut tends to exhibit more variability than the AMIP AGCMsand the three reanalyses.

[51] Although it is true that, even if a simulated climatediffers widely from reality. isotopic simulations can beappropriate to this model-simulated climate. such isotopiccharacterization would be of little use for assessment of thecauses of land-atmosphere exchange mechanisms: the goalof this paper. This section has demonstrated the need forvalid hydrological simulations prior to any investigation ofisotopic hydrological exchanges.

4. GNIP Data Analyses and Interpretation4.1. GNIP Data From Recent Decades

[52] Stable isotopic data in rainfall have been collected bythe lAEAiWMO in the Amazon Basin since 196 I [e.g ..Rozanski et al .. 1993] as part of a global monitoringprogram over 550 stations. Sadly, from this total, less thanhalf have records appropriate for interannual analysis, andonly a very small fraction are still operational. Of the six orseven stations that have been established within the generalarea of the Amazon Basin (Figure l b), the most recentlyreporting (to 200 I) stations are Izobamba and Bogota ascompared with Belern, Manaus, and many other Brazilianstations which provided data until the end of the 1980s(Table 6a).

[53] Although the AMIP IIperiod includes three EI Ninosand two La Nifias, the availability of Amazon GNIP datadoes not coincide very well with the later of these events. To

18 01'25

017104 \lCGUFFIE A\iD HE\iDERSON-SELLERS: STABLE \VATER ISOTOPES 017104

A",IIP AGCM Precip .. 111md-I

Table 5. Annually Averaged Water Budget Components for the Amazon Basin From 20 AMIP II AGCMs"

RaP,.

ObservationsAEFHIJKLNoPQsT

Water imbalance unacceptable below hereB 5.\1 0.08 0.63 0.35C 5.06 0.07 0.62 0.37o 3.55 -0.55 0.R4 0.31G 5.22 -0.08 0.67 0.341\1 5.18 0.68 0.52 0.35R 5.40 -0.49 0.65 0.44

"Water imbalance (Wimb) is determined to be unsatisfactory if it exceeds ,0.05 mm d-I. These cases are highlighted byboldface and italics.

obtain adequate data for analysis, one earlier La Nifia(1973-1975) was included. and the last event of each typein the AMIP II period was set aside. Table 6b lists thenumber of months for which 6180 and 6D data are bothavailable at Manaus and lzobamba.

[~-+] Despite the relative sparseness of the data in Table 6b.the differences in depletions are remarkable (Figure 8).Specifically. the E\~O and ED measurements follow oneanother very closely in the mean (average of all monthsavailable, see Table 6a); in EI Nifios (about 20 months'data); and in La Nifias (between 22 and 33 months' data. seeTable ob). At Manaus. EI Nifios cause a very noticeablereduction in depletions from April to August (dry season).while La Nifias enhance the mean dry season depletions.For the other months of the year, EI Nino tends to reduce thedepletions as seen at Manaus. 6\ 80 and 6D depletions trackclosely and clear separations occur between E\:SO and"normal" conditions especially in the dry season. Theopposite effect is seen at lzobamba in May when EI Nifiosincrease the depletion in both isotopes. Near the Pacificcoast of South America. EI Nifios are characterized byincreased precipitation. largely derived from Pacific-sourcedmoisture which is not affected by the Amazon [HofJinan etal .. 2003] (Figure Sc ).

[,~] Figure 9a compares basin-wide averages of ~\80 anddeuterium excess for the 17 years of the AMIP II periodfrom MUGCM together with the observed values forManaus and Izobamba. For exO the model lies betweenthe two sets of station data. with a very similar seasonalpattern. For the deuterium excess, however. the MCGCMbasin averaging leads to values close to the MWL (seeFigure 3b), while data for Manaus show evidence forincreased recycling in the middle of the year. Figures 9band 9c are for the EI Nifios and La Nifias listed in Table 6b.The ~IXO values are rather similar to the means (Figure 9a),except that Manaus data and the MUGCM simulationsshow a little more depletion during La Nifias. In the

Wimb .. rnm d"l

4.9R4.254.R65RO4.735.105.473.91

0.00.02

-0.02-0.02-0.04

0.010.00

-0.040.02

-0.02-0.01

0.00-002

0.050.00

0390.750.6R0.650.60

0.610.250.320.350.410.270.34

o 7~0.660.6R0.550.64G.740.630.70

0.330.440.360.260.37G.310.260.38

5.735.914.064.035.145.355.56

0.730.62

deuterium excess seasonal cycle. however. there are cleardifferences between EI Nifio and La Nina years forMUGCM, with much larger values of d excess in August-September of EI Nino years, indicating more recycling atthese times.

[~(,] Analyses for Izobamba and Bogota show upperAmazon Basin isotopic depletions altering recently (i.e ..into the I990s and 2000s. Figures lOa and lOb). Enhanceddepletions in wet months tend to be statistically significantat both locations. At Izobamba the larger depletions haverecently become still greater. while smaller depletionshave tended to diminish (Figure lOa). As century-longanalysis of ice core isotopes from the Andes indicatesignificant interdecadal variability [Hoffinan et al., 2003].care needs to be exercised in attempting to draw conclu-sions from recent decadal-only contrasts. However. therecent enhanced depletion is interesting because it con-trasts with Henderson-Sellers et al.ts [2002] finding ofdecreasing depletions at Manaus between the 1960s andthe 1980s. One possible interpretation could be that inrecent decades the impact of deforestation [e.g., Fearnside.1993] is affecting water recycling which is the result ofcanopy interception and reevaporation of that water.Alternatively. increases in the areal extent of open water[Folel' et al.. 2002] might have decreased the relativeimportance of reinsertion of heavy isotopes into the basinhydrological system. Finally. there are known to have beenchanges in the large-scale circulation of both the atmo-sphere and the oceans between these two periods [e.g ..Vuille et al.. 2003a. 2003b] which may have affected thesource or isotopic history of precipitation and/or floodingextent and persistence.

[~7] Seasonal amplitude in isotopic depletions in theupper part of the Amazon Basin has increased over the last25 years (Figures 10e and 1Of). At Izobamba the dry seasondepletions have decreased. and the wet season depletionsincreased. At Bogota (where precipitation is influenced by

19 of 25

DI7104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES DI7104

Station Operational Period

Table 6a. IAEA/WMO Amazon Basin Isotope Collection Station Location Details and Availability"

","umber of Data Months Percent Time for 0 and "0Location Alt., m asl

Belem\lanausIzobambaIAlluriquin IBogota

1.43'S.48.48'\\3.1~'S. 60.0~'\\0.37'S. 78.55'\\(O.~'S. 78.38'\\')4.7'1\.7-1.13'\\

~472

3058(850)~5-17

1965- 19871965 -1990

1968 -date. i.e .. ~OOI -( 199~-1996)

1971-date. i.e .. ~OOI

87.5%52.Q°1)93. 1o,()

(93%)9"~o

~6.J3003~41~-1)213

"The last column is the percentage of the total months of observation (IX which both deuterium (0 I and oxygen 18 1"01 observations are available.

the Amazon for only part of the year), the wet seasondepletions have also increased. while the dry season showsno change. It is possible that the seasonal change in biROand sn signatures between the I 970s/ 1980s and 1990s/2000s may indicate the impact of deforestation (vegetationremoval prompting less recycling and less reinsertion ofheavy isotopes into the basin hydrological system) orgreater river water and flooded areas in the central basinmight have altered the balance between fractionating (evap-oration from open water) and nonfractionating (full canopyand transpiration). This latter proposition could possibly beprobed further using d excess observations and simulations.Although Figures 10c and 10d show no d excess differencesat times when bI~O depletions change. the integratingimpact of ENSO events on Andean isotopes demonstratedby Hoffman et al. [2003] and Vuille et al. [2003b] might besmoothing anticipated signals.

4.2. Using Isotopes to Evaluate AmazonianHydro-Climatic Changes

[~8] There is considerable interest in the use of isotopicinformation as a novel data source for model evaluation[e.g .. Gibson et al.. 2002]. The purpose of this paper is as anexemplar of how investigations of changing hydrologycould possibly exploit stable water isotope data and isotopicmodel simulations. It focuses on an area where isotopiccharacterization initiated GCM analysis of land-atmosphereinteractions: the Amazon Basin.

[~9] We have shown in section 2 that two lSSs (ClMand ISOlSM) exhibit significant sensitivities in their sim-ulation of gross and isotopic surface hydrology whenaddition is made of 20% open water and with a realisticdepletion in the ambient atmospheric water vapor. It islikely that isotope-enabled lSSs would result in similar. ifnot greater, disagreement on the relative importance offractionating and nonfractionating processes [e.g .. Chen etal .. 1997].

[60] In section 3 we illustrated the great variety in theperformance of AGCMs in simulating the water budget ofthe Amazon Basin. As Vuille et al. [2003a] argue that E?\SOevents effectively integrate many factors affecting theisotopic character of South American precipitation. ourresults demonstrate that. without a better consensus amongstate-of-the-art AGCMs and reanalyses on the nature of thismajor basin's hydrology. the addition of isotope tracing isunlikely to reveal much that is unambiguous about changesto the system.

[61] In this section, observations of isotopic depletionsin Amazonian precipitation have been compared withsimulations by an isotope AGCM and analyzed for possi-ble recent trends. We find that recent upper basin isotope

measurements show a tendency toward more depletion.possibly because there is less or changed water recycling(Figure 10). There is more depletion in the wet seasonarguably because relatively less non fractionating recyclingoccurs (i.e., less transpiration and full canopy evaporationand/or relatively more evaporation from open water areas).The recent increased seasonality at lzobamba (Figures 10eand 10f) could therefore be due to (I) relative changesbetween fractionating as compared to nonfractionatingevaporation (a plausible conclusion if lake and open waterareas have changed); (2) less water recycling (a possibleresult of deforestation); or (3) circulation changes.

[62] Henderson-Sellers et al. [2002] proposed that theirobserved isotopic changes might be due to greenhouseintensification of the hydrological cycle masking anyland-use change impact. Alternative explanations for theirnull result. including that isotope data to 1990 only wereavailable and the statistically significant wet seasonchanges reported might be related to ENSO events orother climatic variations that modify the regional circula-tion and hence affect the gross and isotopic hydrology ofthe Amazon. have all been shown here to have merit. Onthe other hand. it has also been demonstrated thatnumerical models (both atmospheric and land-surface)that cannot reproduce gross water budgets correctlycannot add value to isotopic interpretations.

[6.1] There is certainly potential to explore isotopic mod-ification in the Amazon by utilizing state-of-the-art isoto-pically enabled land surface schemes combined withplausible "isotope" GCMs [Noone and Simmonds, 2002;Hallinan et al., 2003; Vuille et al .. 2003a, 2003b]. Thechallenge is to choose problems appropriate to this new tooland to validate gross water predictions before interpretingisotopic simulations.

Table 6b. El Nifio and La Nina Periods Analyzed From Gl\IPStations Together With the Number of Months of Measurements ofBoth tiRO ;nd to

Period Manaus Izobamba

EI NinoJuly 198~ to June 1983 10 9July 1986 to June 1987 10 0July 1991 to June 1992 () II

La NinaJuly 1973 to June 1975" 0' 21~JJuly 1984 to June 1985 10 I

"Note that the first La Nina event is not in the AMIP II period but wasincluded because the GNIP coverage of the La Nifias in AMIP is poor.

20 of 25

DI7104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES DI7104

a) Manaus ENSO0 20

-1 10

-2 0

-3-10-- 00 -4~ -20 ~--0 -5

--- 0180 La Nina0

co -30 JQT"" -6 ----- 0180 EL Nino;JC

-7 --'lIr- Mean b180 -40

-8 - --0- - W La Nina -500 so EI Nino

-9 -60--MeanbD

-10 -701 2 3 4 5 6 7 8 9 10 11 12

b) Izobamba ENSO-5 -30

0

-7 -50

-9-70 ---- -110 0

~ -90 ~0 -13 0co --- 6180 La Nina JQT"" -15

---- 6180 EL Nino -110JQ

-17 --'lIr- Mean b180- --0- - W La Nina -130

-19·0· WEI Nino

-21-150

CJ --MeanbD

-23 -1701 2 3 4 5 6 7 8 9 10 11 12c) month

La Nina

EI Nino

Figure 8. El Nino and La Nina I'I~O and I'D depletions from G0iIP observations at (a) Manaus and(b) [zobamba for all years (mean. see Table 6a: El Nino. 20 months: and La Nina, 20-30 months. seeTable 6b). (c) Changes to the Walker Circulation during La ~ina (above average precipitation. upper panel)and EI ~ino (lower panel) periods when convection shifts to the central Pacific. reducing precipitation in theAmazon (modified from Folev et al. [2002]). See color version of this figure in the HTML.

2 J of 25

D17104 ,\lCGLiFF1E AJ\iD HP\DERSON-SELLERS: STABLE WATER ISOTOPES

a) All years01""""-----------.......,

-20 1 2 3 4 5 6 7 8 9 10 11 12

b) EI Nifios0,-------:---------.,

_ -5a

~0-10CO

o

~-10

oOJ~ -15

~- ..•. - -..' •.

-15

.•.-- •.-- .._-. ~

~/

-5 "+'- •. -- •.

-20-25 L- ..1

c) La ~r-in--a_s ---..,1 2 3 4 5 6 7 8 9 10 11 12

-15

-20L------~==~1 2 3 4 5 6 7 8 9 10 11 12

month

D17104

16,.....------------.......,15

014oQ-13000012(l)o 11X(l) 10"0 9

87L- •....•1 2 3 4 5 6 7 8 9 10 11 12

/\ 0/\- _.• - -'*-----4 - _•. - - •. \ .j-. -_•.--•._,y_

V

22,.....------------.,20

018oQ~16

~ 14oX 12(l)

"010

86L.-------------..I

.., -- '.

1 2 3 4 5 6 7 8 9 10 11 12

20.---------------,18

8

6•••••------------..1.•.

1 2 3 4 5 6 7 8 9 10 11 12

month

Figure 9. (a) Seasonal cycle of precipitation-weighted blXO (%0) and deuterium excess for all 17 yearsof the AMIP II period for MUGCM together with GNIP-derived values for Manaus and lzobamba fromall available years. (b) As Figure 9a, except for El Niiios. (c) As Figure 9a, except for La Niiias (ENSOperiods listed in Table 6b give rise to a data gap in August at Manaus for El Niiios).

22 of 25

017104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES 017104

450 a) Izobamba

~ 400E 350E~ 300c.Q 250

§ 200.is...(3 150

~ 100Q.

50

4 5 6 7

month

c) Bogota~ 400 r"o------------------TOE 350 _19805 -2_E. 300 ••••• --20005 -4-60:§~ ~~.~ 200 -10~:9- 150 -12 gRg 100 -14~C. 50 -16

-18o ~

4 6 7month

10 11 12

e)o

-2

-4

o -6

~ -8

0-10co/.:: -12

-14

-16

-18

-+-l-+-l-+-l-+-

Wet Season

Izobamba Bogota

_19805- - -~•• '19905

.~-._;)

4 6 7month

10 11 12

-=d;.:.).;;;B..;,o,:;:,g.;.ot;.;;a .,.400 r 20

E 350E.300

§ 250

'2200

:9-150uQl 100C.

50

_____ 19805

-- -.-- -200050-

15 CDX

108(f)(f)

5 ~o-'?~o

4 6 7 8 9 10 11 12month

f)

-2

T T.1 .1

-+-Dry Season

o ~4

~o -6co~..c -8

-10

-12Izobamba Bogota

Figure 10. Monthly mean precipitation (mm) and values OfblXO in precipitation at (a) Izobamba for the1980s as compared to 1990s and (b) Bogota for the 1980s and 2000s. Error bars are ± I SE. (c) and (d) AsFigures lOa and lOb. except for d excess, (e) Four-monthly wet season averages onlXo between the twobidecadal periods (I 970s plus 1980s) as compared to (1990s plus 2000s) at lzobamba (February-May)and Bogota (April/May and October/November). (f) As Figure 10e. except for dry seasons (June~September) at Izobamba and (January/February and July/August) at Bogota. Full period standard errorsshow the statistical significance.

23 or 25

D17104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES D17104

[64] Acknowledgments. P. Irannejad and L. Dyer contributed to theAMIP analysis, and 0, Noone cave access to the MUGCI\\ results. We arealso grateful to D. Noone and \V. Riley for the usc of ISOLSM.

ReferencesBonan. G. B. (1996), A land surface model (LSM version 1.0) for ecolog-

ical. hydrological. and atmospheric studies: Technical description m;duser's guide. SCAR Tech. .Yor~ .YCARTV-417+STR. 150 pp .. Natl, Cent.for Atmos. Res .. Boulder. Colo.

Bonan. G. B., K. W. Oleson. \1. Verrenstein. S. Levis. X Zeng. Y. Dai.R. E. Dickinson. and Z.-L. Yang (2003). The land surface climatology ofthe Community Land Model coupled to the NCAR Community ClimateJ\lodel..J. Clint .. 15.3123-3149.

Botta. A .. N. Ramankutty, and J. A. Foley (2002). Long-term variations ofclimate and carbon fluxes over the Amazon Basin. Geophvs. Res. Lett ..19(9),1319. doi:10.1029200IGLOI3607.

Chen. T.-C .. J.-H. Yoon. K. J. St. Croix. and E. S. Takle (2001). Suppres-sing impacts of the Amazonian deforestation by global circulationchange. Bull. Am . Meteorol. Soc .. 81. 2209 - 2216.

Chen. T. H .. et al. (1997). Cabauw experimental results from the Project forIntercomparison of Land-Surface Parameterization Schemes. J. Clim.,10(7). 1194- I215,

Costa. M. H .. and J. A. Fole v (2000). Combined effects of deforestation anddoubled atmospheric CO~ concentrations on the climate of Amazonia.J. Clim .. 13. 18- 34.

Craig. H. (1961). Isotopic variations in meteoric water, Science, 133,1702-1703.

Dai, Y.. er al. (2003). The common land model. Bull .. Am .. vleteorol. Soc ..84. 1013 -I 023.

DallOlio, A. (1976). A cornposicao isotopica das precipitacocs de Brasil:Modelos isotermicos e a influencia da ev apotranspiracao na Bracia Ama-zonica (in Portuguese). MSc. thesis. 180 pp., Piracicaba. Uni v, de SaoPaulo. Sao Paulo. Brazil.

Dansgaard, W. (1964). Stable isotopes in precipitation. Tellus. 16. 436-468.

Dickinson, R. E., A. Henderson-Sellers, P. J. Kennedy, and M. F. Wilson(1986). Biosphere-atmosphere transfer scheme (BATS) for the NCARcommunity climate model. 'yCAR Tech. .vorl.' .\'C~R TV-]75- STR. 69pp .. Natl. Cent. for Atmos. Res .. Boulder. Colo.

I' earnside, P. M. ( 1993). Deforestation in Brazilian Amazonia: The effect ofpopulation and land tenure. Ambio, n. 537-545.

Fekete. B. M .. C. J. Vorosmarty, and W. Grabs (2000). Global. compositerunoff fields based on observed river discharue and simulated waterbalances. GRDC Rep. 11. Global Runoff Data Cent.. Koblenz. Germany.(Available at http://www.grdc.sr.unh.edu/htmIlData,)

Foley. J. A.. A. Botta. M. T. Coe, and M. H. Costa (2002). EI Nino -SouthernOscillation and the climate. ecosvsterns, and rivers of Amazonia. GlobalBiogeochem. Cvcles. 16(4).1132'. doi:10.10292002GBOOI872.

Froehlich. K .. J. J. Gibson. and P. Aggarwal (2002). Deuterium excess inprecipitation and its climatological significance: Study of environmentalchange using isotope techniques. C&S Pal', SCI: l i/P. pp. 54-65.lnt. At.Energy Agency. Vienna. Austria.

Gat. J. R. (2000). Atmospheric water balance- The isotopic perspective.Hvdrol. Proc .. 14. 1357 - 1369.

Gat: J. R .• and E. Matsui ( 1991 ). Atmospheric water balance in the AmazonBasin: An isotopic evapotranspiration model, 1. Geophys. Res .. 96(D7).13.179-13.188.

Gates. W. L.. et al. (1999). An overview of the results of the AtmosphericModcl Inrercomparison Project (AMIP I). BlI/I. Am .. vleteorol, Soc. 80,29-55.

Gibson. J. 1.. et al. (2002). Isotope studies in large river basins: A newglobal research focus. Eos Trans. AGe. 83(52). 613-617.

Gibson. J. K .. P. Kallberg, S. Lppala, A. Hernandez. A. Normura, andE. Serrano (1997). E~"'. description, ECMTVF Re-Anal. Proj Rep.Ser. 1, 72 pp .. Eur. Cent. for Medium-Range Weather Forecasts. Read-ing. UK,

Henderson-Sellers. A .. and v Gornitz ( 1984), Possible climatic impacts ofland cover transformations, with particular emphasis on tropical defor-estation. Clim. Change, 6. 231-258.

Henderson-Sellers. A .. K, McGuffie. and H. Zhang (2002 I. Stable isotopesas validation tools for global climate model predictions of the impact ofAmazonian deforestation. J. Clint .. 15. 2664-2677.

Henderson-Sellers. A .. P. lrannejad, K. ''-'leGuffie. and A. J. Pitman (2003a).Predicting land-surface climates Better skill or moving targets". Geo-phvs. Res, Lett .. 30(14).1777. doi: 10. I 029/2003GLO I 7387.

Henderson-Sellers. A .. P. Irannejad. K. McGutTie. S. Sharmeen, T. J.Phillips. and H. Zhang (2003b), Evaluating GEWEX CSES' simulatedland-surface water budget components. GEHEX\,C\I'S. 13(3). 3-6.(Available at http.z-www.gewex.org/)

Henderson-Sellers ... \. K. McGuffie. D. Noone. and P. lrannejad (2004),Using stable water isotopes to evaluate basin-scale simulations of surfacewater budgets. 1. Hvdrometeorol., in press.

Hoffman. G .. et al. (2003). Coherent isotope historv of Andean ice coresover the last century. Geophys. Res. Lett .. 30(41. 1179. doi:10.I0292002GL014870.

Huffman. G, 1.. R. F. Adler. B. Rudolf. C. Schneider. and P. R. Keehn(1995), Global precipitation estimates based on a technique for combin-ing satellite-based estimates. rain gauge analysis. and NWP model pre-cipitation information. 1. Clint .. 8. 2810- 2823,

Ingraham. N, L.. and R. C. Craig (1986). Hydrogen isotope study of large-scale meteoric water transport in northern California and Nevada.1. Hvdrol., 85. 183-197.

International Atomic Energy Agency (IAEAI 120(3), Global Network forIsotopes in Precipitation: The GNIP database. release 3. October 1999.Vienna. (Available at http:isohis.iaea.org I

Irannejad, P.. A. Henderson-Sellers, and S. Sharmeen (2003), Importance ofland-surface parameterization for latent heat simulation in global atmo-spheric models. Geophvs. Res, Lett., 30( 17l. 1904. doi: 10.10292003GlO 18044.

Joussaume, S .. R. Sadourny, and J. Jouzel (1984 I. A general circulationmodel of water isotopes cycles in the atmosphere. Nature, 311. 24 29,

Jouzel, 1.. G. L. Russell. R. J. Suozzo, R. D. Koster. J. \V. C. White. and\V. S. Broecker (1987), Simulations of the HDO and H2180 atmo-spheric cycles using the NASA GISS General Circulation Model: Theseasonal cycle for present-day conditions . .J. Geophvs. Res .. 91.14.739-14.760.

Kanamitsu, M .. W. Ebisuaki, J. Woolen. S. Yang. J. J. Hnilo, M. Fiorino.and G. L. Potter (2002). NCEP/DOE AMIP-Il Reanalysis (R-2). Bull.Alii. Meteorol. Soc .. 83. 1631-1643.

Kistler. R .. ct al. (2001). The NCEP-NCAR 50-vcar reanalvsis: Monthlvmeans CD-ROM and documentation. Bull. Alii ~\fereorol. Soc .. 81. 247':'268.

Legates. D. R .. and C. J. Willmott (1990 I. Mean seasonal and spatialvariabilirv in gauge-corrected. global precipitation. Int. 1. Cliniatol .. 10.111-127.

Leopoldo, R, R. (1981), Aspetos hidrologicos at tlorista amazonica dengana regiao de Manaus, Ph.D. thesis. 180 pp .. Univ. Natl. Estado San Paolo.Botucato. Brazil.

Liang. X .. D. P. Lertenmaier, E. F. Wood. and S. J. Burges ( 1994). A simplehydrologically based model of land surface water and energy fluxes forGSMs.1. Geophys. Res., 99(D7). 14,415 14,428.

Manabe, S. (1969 I. Climate and ocean circulation: I. The atmosphericcirculation and the hydrology of the earth's surface. MOil. WeatherRev.. 97. 739- 774.

Martinelli. L. A. (2003). Isotope tracing of hydrological processes in largeriver basins: Design criteria for a network to monitor isotope composi-tions of runoff in large rivers, Int. At. Energy Agency. Vienna.

Martinelli. L. A .. R. l.-Yictoria. L. S. L. Ste~berg~ A. Ribeirio. and \1. Z.Moreira ( 1996). Using stable isotopes to determine sources of evaporatedwater to the atmosphere in the Amazon Basin, J. Hvdrol .. 183, 19 1-204.

Matsui. E.. E. Salati. M. N. G. Ribeiro. C. M, Reis. A. C. S. N. F. Tancredi.and J. R. Gat ( 1983). Precipitation in the central Amazon Basin: Theisotopic composition of rain and atmospheric moisture at Belern andManaus. Acta .tmazonica, 13.307-369,

McGuffie, K., A. Henderson-Sellers. and H. Zhang (1988), Modelling cli-matic impacts of future rainforest destruction. in Human Activities andthe Tropical Rainforest, edited by B. K. Maloney, pp. 169 - 193. KluwerAcad .. Norwell. Mass.

McGuffie. K .. A. Henderson-Sellers. H. Zhang. T. Durbidge, and A. J.Pitman (1995), Global climate sensitivity to tropical deforestation, GlobalPlaner. Change, 10. 97 -128.

Moreira. M. Z .. L. D. L. Sternberg. L. A. Martinelli. R. L. Victoria. E. M.Barbosa. L. C. Bonates. and D. C. Nepstad (1997). Contribution oftranspiration to forest ambient vapor based on isotopic measurements.Global Change Bioi .. 3, 439-450.

National Oceanic and Atmospheric Administration (NOAA) (2003). ENSOimpacts on the L.S.: Previous events. Silver Spring. Md. (Available athttp: www.cpc .ncep.noaa.gov products analysis_monitoring 'ensostuffensovears.hrml)

Nation;1 Oceanic and Atmospheric Administration - Cooperative Institutefor Research in Environmental Sciences (\I0AA-CIRES) (2003). NCEPReanalysis data provided by the '\OAA-CIRES Climate DiagnosticsCenter. Boulder. Colo. (Available at http: www.cdc.noaa.gov )

Nijssen. B .. R, Schnur. and D, P. l.ettenmaier 12(01). Global retrospectiveestimation of soil moisture usinu the VIC land surface model. 1980-1993. J. Clint .. 14{ 8 I. 1790 - 1808.

Noone. D .. and I. Simmonds (2002 I. Association between t\.sO of waterand climate parameters in a simulation of atmospheric circulation for1979-95. J. Clint .. 15.3150-3169.

24 of 25

DI7104 MCGUFFIE AND HENDERSON-SELLERS: STABLE WATER ISOTOPES D17104

Pitman. A. J .. et al. 119991. Kev results and implications from phase IICI ofthe Project for Intercomparison of Land-Surface ParameterizationSchemes. Clim. Dvn .. 15. 673 -684.

Richey. J. E .. J. M. Melack. A. K. Aufdenkampe. V. 1\1. Ballester. and L. L.Hess 12(02). Outgassing from Amazonian rivers and wetlands as a largetropical source of atmospheric CO2 . .\"alllre. 416. 617 -620. -

Riley. W. 1., C. J. Still, M. S. Tom. and 1. A. Berry 120021, A mechanisticmodel of H~xO and CI'OO fluxes between ecosystems and the atmo-sphere: Mod-c1 description and sensitivity analyses. Global Biogeochem.Cvcles. 1(514). 1095. dodO.1029 2002GB001878.

Rozanski. K .. L. Araguas-Araguas. and R. Gonfianrini 11993 I. Isotopicpatterns in modem global precipitation. in Climate Change ill Continen-tal Isotopic Records, Geophvs .. vlonogr. Ser.. vol. 78. edited by P. K.Swart et al.. pp. 1-36. AGe. Washington. O. C.

Salati. E .. and P. B. Vose 11984 I. Amazon Basin: A system in equilibrium.Science. ::25. 129-137.

Salati. E .. A. O. Olio. E. Matsui. and J. R. Gat 11979). Recycling of water inthe Amazon Basin: An isotopic study. Hiller Rcsoin: R~s.. /515). 1250-1258.

Stone. O. w.. A. Henderson-Sellers. P. Airey. and K. \leGut1ie 120(3).Murray Darling basin isotope observations: An essential component ofthe Australian CEOP. Eos Trans. AGU. 114(46). AGU Fall Meet. Suppl..H22I-08.

Townsend. A. R .. G. P. Asner. 1. \V. C. White. and P. P. Tans 120(2). Landuse effects on atmospheric I 'c imply a sizable terrestrial CO2 sink intropical latitudes. Gcophvs. Res. L!!II .. 29( 10). 1426. doi: 10.10292001 GLOl3454

Victoria. R. L.. L. .-\. Martinelli. 1. Mortatti, and J. Richey 11991 I. Mecha-nisms of water recycling in the Amazon Basin: Isotopic insights, .4I11hio.20181.384-387.

Yuille. \1.. R. S. Bradley. M. Werner. R. Healy. and F Keimig 12003a).Modeling <"'0 in precipitation over the tropical Americas: I. Interannualvariability and climatic controls. 1. Geophvs. RI!S .• 1081D6). 4174.doi:IO.10292001J0002038.

Yuille. M .. R. S. Bradley. R. Healy. M. Werner, O. R. Hardy. L. G. Thompson.and F. Keimig 12003b). Modeling <''''0 in precipitation OWl' the tropicalAmericas: 2. Simulation of the stabile isotope signal in Andean ice cores.1. Geophv». RI!S .• f 118106), 4175, doi: 10.1 029 200 1J0002039.

Wood, E. F.. et al. (1998), The Project for lnrercomparison of Land-SurfaceParameterization Schemes IPILPS) Phase 2( c) Red-Arkansas river basinexperiment: I. Experiment description and summary intercomparisons.Global Planet. Change, 19. 115-135.

Xie. P. P.. and P. A. Arkin (1997). Global precipitation: A 17-year monthlyanalysis based on gauge observations. satellite estimates, and numericalmodel outputs. Buft. A~l/. Meteorol. Soc. 78.2539-2558.

A. Henderson-Sellers. Environment. Australian Nuclear Science andTechnology Organisation. Lucas Heights. 'S\\" 2234. Australia.

K. "leGume. Department of Applied Physics. Lnivcrviry of Technology.Sydney. P.O. Box 123. Broadway. 1\S\\' 2007. Australia. (kendal.mcguffie«. uts.edu.au)

25 of 25

Journal Format For Print Page: lSI http://www.isinet.com/cgi-bin/jrnlst/jlresults.cgi

I S I..•.

THOMSON SOENTlF1C

(Use Browser Back button to return to previous page)

Thomson Scientific Master Journal ListJOURNAL LIST

Search terms: 0148-0227Total journals found: 6

1. JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES

MonthlyISSN: 0148-0227AMER GEOPHYSICAL UNION, 2000 FLORIDA AVE NW, WASHINGTON, USA, DC,20009

2. JOURNAL OF GEOPHYSICAL RESEARCH-EARTH SURFACE

MonthlyISSN: 0148-0227AMER GEOPHYSICAL UNION, 2000 FLORIDA AVE NW, WASHINGTON, USA, DC,20009

3. JOURNAL OF GEOPHYSICAL RESEARCH-OCEANS

MonthlyISSN: 0148-0227AMER GEOPHYSICAL UNION, 2000 FLORIDA AVE NW, WASHINGTON, USA, DC,20009

4. JOURNAL OF GEOPHYSICAL RESEARCH-PLANETS

MonthlyISSN: 0148-0227AMER GEOPHYSICAL UNION, 2000 FLORIDA AVE NW, WASHINGTON, USA, DC,20009

5. JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH

MonthlyISSN: 0148-0227AMER GEOPHYSICAL UNION, 2000 FLORIDA AVE NW, WASHINGTON, USA, DC,20009

6. JOURNAL OF GEOPHYSICAL RESEARCH-SPACE PHYSICS

101'2 1/0412005 10:31 AM

~AGU Become an AGU MemberSubscribe to AGU Journals

View full article: HTML; PDF (1820408 bytes)

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, D17104, doi:10.102912003JD004388, 2004

Stable water isotope characterization of human andnatural impacts on land-atmosphere exchanges in theAmazon Basin

K. McGuffieDepartment of Applied Physics, University of Technology, Sydney, Broadway, NewSouth Wales, Australia

A. Henderson-SellersEnvironment, Australian Nuclear Science and Technology Organisation, LucasHeights, New South Wales, Australia

Abstract

Stable water isotopes have been employed as a means of challenging, validating,and improving numerical models of the Amazon Basin since the 1980s. This paperserves as an exemplar of how characterization of human and natural impacts onsurface-atmosphere water exchanges could beneficially exploit stable water isotopedata and simulations. Interpretations of Amazonian isotopic data and modelsimulations are found to be seriously hampered by (1) poor simulation of the grosswater budget (e.g., lack of surface water conservation in models); (2) considerablemodel differences in the fate of precipitation (i.e., between reevaporation andrunoff); (3) wide ranging characterization of natural causes of water isotopicfluctuations (especially El Nino and La Nina events); (4) isotopic land-atmosphereflux sensitivity to the prescription of boundary layer atmospheric water vaporisotopic depletion; and (5) significantly different characterization by currentland-surface schemes of the partition of evaporation between isotopicallyfractionating (from lakes and rivers) and nonfractionating (transpiration) processes.Despite these obstacles, we find features in the recent isotopic record that might bederived from circulation and land-use changes. ENSO events may cause decreaseddepletion in the dry season, because of reduced convective precipitation, whileincreases in upper basin isotope depletions in the wet season may result fromrelatively less nonfractionating recycling because there are fewer trees. The promisefor isotopic fingerprinting of near-surface continental water cycle changes dependsupon fixing shortcomings in current atmospheric and land-surface models.

Received 25 November 2003; accepted 30 June 2004; published 14 September 2004.

Keywords: stable isotopes; Amazon; models.

Index Terms: 1615 Global Change: Biogeochemical processes (4805); 1610 Global Change:Atmosphere (0315, 0325); 1833 Hydrology: Hydroclimatology; 3322 Meteorology and AtmosphericDynamics: Land/atmosphere interactions.

View full article: HTML; PDF (1820408 bytes)

Citation: McGuffie, K., and A. Henderson-Sellers (2004), Stable water isotope characterization of

human and natural impacts on land-atmosphere exchanges in the Amazon Basin, J. Geophys. Res., 109,017104, doi:lO.102912003JD004388.

Copyright 2004 by the American Geophysical Union.