Global and Planetary Change 118 (2014) 69–84

Contents lists available at ScienceDirect

Global and Planetary Change

j ourna l homepage: www.e lsev ie r .com/ locate /g lop lacha

Hydrologic response of a high altitude glacierized basin in the centralTibetan Plateau

Binquan Li a,b, Zhongbo Yu a,c,⁎, Zhongmin Liang a, Kumud Acharya b,⁎⁎a State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing 210098, Chinab Division of Hydrologic Sciences, Desert Research Institute, Las Vegas, NV 89119, USAc Department of Geoscience, University of Nevada Las Vegas, Las Vegas, NV 89154, USA

⁎ Correspondence to: Z. Yu, No. 1 Xikang Road, NanjChina.⁎⁎ Corresponding author.

E-mail addresses: zyu@hhu.edu.cn (Z. Yu), kumud.ach

http://dx.doi.org/10.1016/j.gloplacha.2014.04.0060921-8181/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 11 October 2012Received in revised form 21 April 2014Accepted 23 April 2014Available online 2 May 2014

Keywords:Hydrologic responseGlacier meltEnergy balanceGSSHATibetan Plateau

Hydrologic cycles of most high altitude glacierized watersheds in the Tibetan Plateau are not closely monitoreddue to their inaccessibility. Understanding the hydrologic cycle in such a basin may provide insight into therole climate plays on changes in glacier mass. Thus, hydrologic simulations with a physical perspective in theTibetan glacierized watershed are of great significance. A high altitude glacierized basin in the central TibetanPlateau, Qugaqie basin, was investigatedwith an energy-balance based glacier-melt model and the Gridded Sur-face Subsurface Hydrologic Analysis (GSSHA)model.With these twomodels, glaciermass balancewas estimatedand basin runoff from glaciers was simulated at a daily time step. Results from the simulation period (October 1,2006–September 30, 2011) demonstrated that the glaciers experienced a large negative surface mass balancewith the cumulative value of −300 cm w.e.. In other words, up to 13.93 × 106 m3 water volume was meltingout from the glaciers during these five years. In the 2007/08 year, however, the glaciers experienced a surplusmass balance because of the low air temperature and increased precipitation in the summer season. Infiltration,evapotranspiration (ET), and overland flow were also calculated using the GSSHA model. Results showed thatprecipitation, the mainwater source, contributed roughly 95% to the total mass gain of the annual water balancein the Qugaqie basin during the study period, while the glacial runoff (snow/ice melting) contributed 5% waterbalance. In the water loss, 17% of annual water volumewas consumed by the ET process. As a result, the remain-ingwater volume (83%) converted to the basin river flow to the LakeNamCo. In the summertime, the glacial run-off accounted for 15% of the total basin runoff volume, while this contribution increased in the upstream portionto 46% due to a large percentage of glacierized area. The analysis showed that the glacial runoff contributions tothe total riverflow decrease significantly due to the decreased air temperature in the summer of 2008. In general,the integrated model produced acceptable estimations of hydrologic response in this high altitude glacierizedbasin, which is jointly fed by precipitation and glacial runoff. This study suggests that, a process-based modelfor glacierized basins can provide a reasonable simulation of hydrologic response and further enhance our under-standing of this high altitude region in the Tibetan Plateau.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

The Tibetan Plateau (TP) has abundant glacier resources, which feedmany major rivers and lakes (e.g. the Brahmaputra, Ganges, Indus,Yangtze, and Yellow Rivers, and Lake Nam Co). Quantifyingmelt contri-butions from these glaciers is a difficult task due to the lack of field data(e.g. hydrometeorological). There are few mathematical models whichcanprovide reasonably good descriptions of hydrologic processes of gla-cier melt in this region. Glacier melt modeling typically falls into twocategories: temperature index and energy-balance techniques (Hock,

ing, Jiangsu Province, 210098,

arya@dri.edu (K. Acharya).

2005). The degree-day factor, in temperature index models, has alarge range of variation: from two to 20 mm day−1 °C−1, and no clearregional pattern has been found to date (Hock, 2003, 2005). Forenergy-balance models, mass balance monitoring and more detailedclimate data are needed for calibration. However, most glaciers in theTP lack long-term continuous data.

Recently, researchers have started collecting field data on glaciermass balance and hydrometeorology in the western NyainqêntanglhaRange in the central TP. A recent glacier inventory based on remotesensing indicates that the glaciers in the region are retreating (Bolchet al., 2010). Based on the observed glacier mass balance data, somestudies have been conducted for better understanding the regional gla-cier surface energy and mass balance (Mölg et al., 2012; Yu et al., 2013;Zhang et al., 2013; Mölg et al., 2014). Obviously, glacier melt-runoffplays significant roles in sustaining seasonal river flows in this region.

70 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

Thus, a hydrologic model combined with a glacier melt componentwould provide valuable information at thewatershed level andwith re-gard to streamflow contributions. There have been some attempts onthe glacierized basins of the TP in this direction, e.g., Krause et al.(2010), Gao et al. (2011) and Immerzeel et al. (2012). For the under-standing of the major hydrologic processes such as evapotranspiration(ET), surface and subsurface flows, there are many hydrologic modelsthat are currently in use, e.g. Xinanjiang (Zhao, 1992), Arno (Todini,1996), TOPographic Kinematic APproximation and Integration (TOPKAPI)(Liu and Todini, 2002) and Gridded Surface Subsurface Hydrologic Analy-sis (GSSHA) (Downer andOgden, 2004). Among these, process-baseddis-tributed hydrologic models have recently received much attention(Khakbaz et al., 2012). Most of the currently implemented distributedmodels do not consider the glacier-melt process, or even if they do, theygreatly simplify the actual melting process. However, in higher elevationareas, snowfall accumulation and glacier retreat effects can dominatethe timing of runoff and streamflow. Some hydrologic models have in-corporated the temperature index approach for snow and ice melting(Krause et al., 2010; Ragettli and Pellicciotti, 2012). One of the few de-tailed hydrologic studies that have been conducted in glacierized basinsof the TP is that of Krause et al. (2010). They modeled the basin of theLake Nam Co using a spatially distributed model which was forced bya down-scaled ECHAM5 data set, and the glacier-melt was projectedby using the degree–day method. In addition, Gao et al. (2011) applieda hydrologic model J2000 coupling a temperature index glacier-meltmethod in a glacierized subbasin of LakeNamCo. However, no previousstudies have made use of spatially distributed hydrologic modelcoupled with energy-balance based glacier-melt model to investi-gate the basin-wide hydrologic response and no study has explicitlyconsidered glacier contribution to total streamflow from this physi-cal perspective.

In this study, we attempt to develop such an integrated energy-balance based glacier-melt and surface distributed hydrologic modelfor a glacier-covered basin in the TP. The GSSHA model (Downer andOgden, 2004) was revised by adding an energy-balance based glacier-melt module, and applied to a high altitude glacierized basin in thewestern edge of the Nyainqêntanglha Range. The Qugaqie basin is par-tially glacier-covered, with amedian altitude of 5429 meters (m) abovesea level (a.s.l.). Specifically, the hydrologic response and glacier contri-bution to total streamflow were examined in the Qugaqie basin in thecentral TP. Application of the modified GSSHA at this study area allowsthe integration of the glacier-melt model and GSSHA to be examined inthe high altitude cold regions of the TP. The integrated model can ex-plore the complex hydrologic response from the physical perspectivethat is different from those lumpedmodels with simple representativesof glacier-melt and hydrologic processes.

2. Study area and data

2.1. Qugaqie basin

The study basin, named after the Qugaqie River, covers an area of59.6 km2 with 8.4% glacierized coverage at the western edge of theNyainqêntanglha Range in the central TP (Fig. 1). The Qugaqie River(length 15.4 km, average channel bed slope N 0.04) drains into LakeNam Co, which is the second largest saline lake in the TP. In the upstreamdrainage area, ZhadangGlacier, to the southeast, covers an area of 1.9 km2

and spans an elevation ranging from 5518 to 6042m a.s.l. Two other gla-ciers are the Genpu Glacier to the south (2.6 km2, 5542–6081 m a.s.l.)and the Chuxiguo Glacier to the west (0.5 km2, 5517–5887 m a.s.l.).Three automatic weather stations (AWSs) are located near the upstreamsection of the Qugaqie River (ZD1, 5400 m a.s.l., ZD2, 5800 m a.s.l. andZD3, 5665 m a.s.l.) (Kang et al., 2009; Maussion et al., 2011). A raingauge at the glacier terminus (5580 m a.s.l.) has been operational sinceMay 21, 2010 (Zhang et al., 2013). Stream stage is measured in the up-stream and downstream sections of the river (S1, 5364 m a.s.l. and S2,

4780m a.s.l.) with controlled drainage areas of 7.4 and 57.6 km2, respec-tively.Measuredwater stages can beused for calculating streamflow. S3represents the basin outlet. This area has a semi-arid subarctic climatewith an average annual precipitation of 415 mm measured at stationNAMOR (Nam Co station for Multisphere Observation and Research,30°46.44′N, 90°59.31′E, 4730 m a.s.l., 50 km north-east of the basin)from 2006 to 2008 (Gao, 2011). The average annual air temperatureand relative humidity at NAMOR are 0 °C and 52%, respectively. The re-gion is under a complex influence of both the continental climate ofCentral Asia and the Indian Monsoon system, which leads to a climatecharacterized by a strong seasonality in both temperature and precipi-tation. Little precipitation occurs during winter, while about 90% ofthe mean annual precipitation is measured from June to September;the ablation season is short (June to mid-August) but intense (Mölget al., 2012; Zhang et al., 2013). The three continental type glaciers, locat-ed in this continental summer precipitation climate, are called summeraccumulation type glaciers as the maximum of annual accumulationand ablation occurs simultaneously.

The study area mainly consists of periglacial, morainic, and aeolianlandforms (Kang, 2011). The periglacial landform, which is developeddue to the effects of frost weather and gravity, covers about 60% landsurface of the Qugaqie basin. Themorainic landform dominates the gla-cial peripheral area where freezing and thawing influences are verystrong in the downriver area. The aeolian landform mainly developedwith the increase of the newly exposed surface after glacier retreat,consisting of fine and medium sands, are found at the lake coast (closeto the basin outlet). In general, soils in this area are alpine meadow soilup to 20 cm depth and silty soil below 20 cm depth (Tian et al., 2009).The lower altitude limit of high plateau permafrost is 5300 m a.s.l.,while the remaining area in the study basin is seasonal frozen soil (Tianet al., 2009). According to the 1-km Harmonized World Soil Database(Nachtergaele et al., 2008), soil types in this basin are sandy loam andloam (Fig. 2). Land cover is composed of bare ground, grassland, wetland,water body and glacier (Ran et al., 2009; Gao et al., 2011).Wetlands existalong themainstem of the river, where the proportion of vegetation cov-erage is N70%, consisting of plants such as hassock, meadow and shrub(Tian et al., 2009). As for the geology of this region, much of it is knownonly at a cursory or exploratory level due to insufficient data. TheQugaqiebasin mainly consists of Quaternary sediments, limestone, arenite andorthogneiss (Kidd et al., 1988).

2.2. Forcing data

The study period in this paper is from October 1, 2005 to September30, 2011. Daily precipitation at ZD1; air temperature, global radiation,and wind speed at both ZD1 and ZD2 (May 18–October 17, 2007 andMay 18–October 16, 2008); and the streamflow data at S1 and S2 (forthe summer periods of 2007–2008) were collected in this study (Kanget al., 2009; Zhou et al., 2010). In addition, the ZD3 stationwhich locatesin the ablation zone has been operational since 2009. At this site, dailyair temperature, wind speed, relative humidity, global radiation, surfacetemperature, and surface albedo data (fromOctober 4, 2009 to Septem-ber 15, 2011) were also available (Zhang et al., 2013). From the raingauge at the glacier terminus, more than one year precipitation ob-servation (May 21, 2010–September 15, 2011) was collected. In ad-dition, climate data (air temperature, relative humidity and windspeed) at the NAMOR station were also collected for the wholestudy period from the Third Pole Environmental (TPE) Database(http://www.tpedatabase.cn). In the Zhadang Glacier, annual massbalance of five balance years (from 2005/06 to 2009/10) and season-al/monthly mass balance of the 2006/07 and 2007/08 balance yearswere also available. In this study basin, mass balance values were cal-culated from observations of ablation stake records, accumulatedsnow depth, and snow/ice density by the local group. The mass bal-ance data used in this study were from the TPE Database and Kanget al. (2009).

Fig. 1. Qugaqie basin, and its location in the Nam Co region, the central TP of China.

71B. Li et al. / Global and Planetary Change 118 (2014) 69–84

In the climate data time series, the gaps, defined as the period of noin-situ measurement in the study basin, were filled from the externaldata sources as described below. For precipitation data gaps, theTRMM precipitation product (Tropical Rainfall Measuring Mission,3B42 v7) was used because it shows the excellent correlations withthe measured seasonal evolutions (Fig. 3). Precipitation was assumedto be spatially homogeneous in the study basin, and its topographiceffect was neglected. Air temperature data at ZD1, ZD2 and ZD3 wereextrapolated to other basin grid points, and for the data gap periods,air temperature data at NAMOR station were used. Based on the corre-lation analysis (Fig. 4a–b), two different lapse rates (−0.64 °C/100 mand −1.01 °C/100 m) were used for extrapolating the air temperatureat non-glacierized and glacierized grid points, respectively. Relative hu-midity data at ZD2 and ZD3 stationswere used in the calculations by as-suming that it was constant over the basin. The data gaps were filledwith simulated relative humidity at ZD3. The simulated valueswere cal-culated from its air temperature and dew point temperature whichwasestimated from the NAMOR station (Fig. 4c). To fill the humidity datagap, the ZD3 locationwas selected because it has sufficient observationsto establish a statistical relationship of dew point temperature. Windspeed data at ZD2 and ZD3 were used in the calculations and were as-sumed to be spatially homogeneous. The data gaps (i.e., without obser-vations at ZD2 and ZD3)werefilledwithwind speed data at theNAMORstation. Global radiation data at ZD2 and ZD3 were extrapolated to allthe other grid points with the energy-balance based glacier model(discussed further below). For the global radiation data gaps, a hybridradiation generation model (Yang et al., 2001, 2006) was used basedon the meteorological data.

3. Integration of glacier-melt model and GSSHAmodel

3.1. Energy-balance based glacier-melt model

Snow and ice melting in the ablation period can be evaluated with afully energy-balance based glacier-melt model. For example, the previ-ous studies have examined the surface energy budget andmass balanceof the Zhadang Glacier (Mölg et al., 2012; Zhang et al., 2013). In thispaper, the surface snow/icemelting in theQugaqie basin is also calculat-ed by a surface energy-balance based glacier-melt model. In this model,the point energy balance can be expressed as:

G 1−αð Þ þ Lin þ Lout þ QH þ QE þ QG þ QR þ QM ¼ 0 ð1Þ

where G is the global radiation, α is the surface albedo, Lin and Lout arethe incoming and outgoing longwave radiation, respectively; QH is thesensible heat flux, QE is the latent heat flux (QH and QE are together re-ferred to as the turbulent heat flux), QG is the ground heat flux in the iceor snow, and QR is the sensible heat flux supplied by rain. QM is the en-ergy available formelt, which is converted intomeltwater equivalentM.The mass balance and water available for runoff are affected by the la-tent heat flux: total ablation A is obtained considering melt and (re-)sublimation, and M provides the water input for runoff modeling.

In the calculation of net radiation, the global radiation at the grid cellis obtained by (Hock and Noetzli, 1997):

G ¼ IGs

Isð2Þ

Fig. 2. (a) The index maps of the soil texture and land cover of the Qugaqie basin; and (b) Simplified geologic map of the study area (from Kidd et al., 1988).

72 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

where Gs is the global radiation from the climate station; I and Is arethe clearly-sky direct radiation at the grid cell and the climatestation, respectively. This equation assumes that the ratio of globalradiation and potential direct radiation (Gs/Is) is constant over theentire area. The clearly-sky direct radiation on the inclined surface Iis calculated by:

I ¼ I0ψP=P0ð Þ= cosZð Þ cosθ ð3Þ

where I0 is the solar constant (1368 Wm−2), ψ is the atmosphericclear-sky transmissivity, P is the atmospheric pressure, P0 is themean atmospheric pressure at sea level, Z is the local zenith angle,and θ is the angle of incidence between the normal to the gridslope and the solar beam. The detailed procedure for extrapolatingglobal radiation at grid cells can be found in Hock and Noetzli(1997).

Snow albedo is derived from the albedo of the preceding time stepconsidering air temperature and snowfall. This method is mainlybased on the study of Hock and Holmgren (2005), but we fixed thesnow albedo as a constant if new snowfall:

αt2¼

αt1−k1 ln Ta þ 1ð Þ � ek2

ffiffiffiffind

p; ndN0 and TaN0

αt1−k3 � ek4

ffiffiffiffind

p; ndN0 and Ta≤ 0

f αt1

� �; nd ¼ 0

8>><>>:

ð4Þ

whereαt2andαt1

are the snow albedo for the current and the precedingtime steps, respectively, Ta is the air temperature, nd is the number of

days since snowfall, and k1, k2, k3 and k4 are the coefficients obtainedthrough calibration. For the third case (nd = 0), snow albedo is definedas a function of its previous value. The albedo value range of the previ-ous time step is classified into several sub-ranges, and fixed snow albe-do values of the current time step is calibrated for each sub-range of theprevious step. This equation is a modification of the original snow albe-do algorithm of Hock and Holmgren (2005), which was developed forthe hourly simulation.

The outgoing longwave radiation can be approximated by using theStefan–Boltzmann law, while the surface temperature is estimatedwithiteration. Then, the incoming longwave radiation (Lin) is calculated from(Crawford and Duchon, 1999):

Lin ¼ σεT4a ð5Þ

where is the Stefan–Boltzmann constant, and ε is the atmospheric emis-sivity which is obtained by:

ε ¼ 1−clf þ clf � 1:24� e=Tað Þ1=7 ð6Þ

where clf is the ratio of the measured incoming shortwave radiation tothe incoming shortwave radiation at the top of the atmosphere, and eis the vapor pressure. This parameterization had been tested in theZhadang Glacier with a good agreement (R2 = 0.85) for the period ofOctober 4, 2009–September 15, 2011 at the ZD3 station (Zhang et al.,2013).

Fig. 3. Comparisons of the TRMM3B42 v7 precipitation product and ground gaugemeasurements: (a) accumulated precipitation betweenMay 18 andOctober 17, 2007 at the site of ZD1, (b) same as (a) but for the period ofMay 18–October 16, 2008,and (c) accumulated precipitation between May 21, 2009 and September 15, 2011 at the site of rain gauge.

73B.Lietal./G

lobalandPlanetary

Change118

(2014)69

–84

Fig. 4. Corrections of air temperature and dew point temperature: (a) vertical air temperature lapse rate between ZD1 and NAMOR for the non-glacierized area, (b) same as (a) but between ZD1 and ZD2 for the glacierized area, and (c) empiricalcorrelation of dew point temperature between ZD3 and NAMOR stations.

74B.Lietal./G

lobalandPlanetary

Change118

(2014)69

–84

75B. Li et al. / Global and Planetary Change 118 (2014) 69–84

The turbulent heat fluxes are parameterized by using the bulk aero-dynamic method, taking the form:

QH ¼ ρacPU Ta−Tsð Þκ2

ln z=z0ð Þ−ψM z=Lð Þ½ � � ln z=z0Tð Þ−ψH z=Lð Þ½ �QE ¼ ρaLVU q−qsð Þκ2

ln z=z0ð Þ−ψM z=Lð Þ½ � � ln z=z0Eð Þ−ψE z=Lð Þ½ �

8>>><>>>:

ð7Þ

where ρa is the air density, cP is the specific heat capacity of air, LV is thelatent heat of vaporization forwater,U ismeanwind speed, Ta and Ts arethe mean air temperature and surface temperature, respectively, and qand qs are the mean specific humidity at instrument height (z) and thesurface, respectively. κ is the von Kármán constant (0.4), and L denotesthe Monin–Obukhov length which is estimated by iteration (Munro,1990). z0, z0T and z0E are the roughness lengths for the logarithmic pro-files of wind speed, temperature and vapor pressure, respectively. z0Tand z0E are parameterized according to kinematic viscosity of air andfriction velocity (Andreas, 1987).ΨM, ΨH and ΨL are the stability func-tions referring to momentum, heat and water vapor, respectively. ΨM

and ΨH are calculated by applying the non-linear stability functions(Forrer and Rotach, 1997) and Businger–Dyer expressions (Paulson,1970) for the stable and far less frequent unstable cases, respectively.It assumes that ΨL = ΨH, and that the stability functions are spatiallyinvariant across the glacier.

For the ground heat flux, two previous studies employed physicalapproximation methods in the Zhadang Glacier (Mölg et al., 2012;Zhang et al., 2013). For example, Zhang et al. (2013) estimated thevalue of QM according to the thermal conductivity of snow/ice and thesubsurface temperature profiles. In this study, the ground heat flux iscrudely considered based on the procedure in Hock and Holmgren(2005) that the ground heat flux is approximately between −5 and5 W/m2 though the whole year. For each grid cell, the ground heatflux is estimated by assuming that it rises from−5W/m2 to 5W/m2 be-tween July 1 and December 1, and then drops back to−5W/m2 on July1 of the next year, with linear interpolation in these time ranges. The en-ergy supplied by the sensible heat of rain is approximated by:

QR ¼ ρacwR Tr−Tsð Þ ð8Þ

where ρa is the air density, cw is the specific heat of water, R is the rain-fall rate and Tr is the temperature of rain (assumed to be equal to airtemperature).

3.2. GSSHA description and model integration

GSSHA is a physically based two-dimensional (2-D) hydrologicmodel, with features including 2-D overland flow, 1-D stream flow,1-D infiltration, 2-D groundwater, and full coupling between thegroundwater, vadoze zone, streams, and overland flow (Downer andOgden, 2004, 2006). It incorporates the infiltration excess (Horton)and saturation excess (Dunne) runoff generation mechanisms, and hasbeen applied to different locations in the United States. GSSHA providesdifferent combinations of process approximations. In this study, griddedwater input from the glacier-melt reaches the land surface. Surfacewater will run off as 2-D overland flow, and is modeled with a diffusivewavemethodwith an Alternating Direction Explicit (ADE) scheme. Thiswater eventually enters a stream,which is then routed to thewatershedoutlet as 1-D channelized flow with two-step explicit finite volumeschemes. Infiltration is simulated using the Green andAmptwith Redis-tribution (GAR) method. In addition to GAR, the simple “bucket” soilmoisture accounting routine is used to estimate soil moisture. Thepotential ET for vegetated soils is calculated by using the Penman–Monteith method. Actual ET can be calculated from ET and soil mois-ture. In this study, Darcy's law is used in the calculations of exfiltrationin the study basin.

For all areas of theQugaqie basin, the glacier-meltmodelwasused tocalculate gridded water input from both glacier surface and seasonalsnowpack for the GSSHA model. Precipitation on the surface was firstdistinguished as snow and rain based on the wet-bulb temperature Tw(Yamazaki, 2001) and this method had been applied to the energy bud-get modeling in the Zhadang Glacier (Zhang et al., 2013). The fraction ofsnowfall in the total precipitation (Pp) is defined as:

s ¼1−0:5 exp −2:2 1:1−Twð Þ1:3

� �; Twb1:1

�C

0:5 exp −2:2 Tw−1:1ð Þ1:3� �

; Tw≥1:1�C

8<: ð9Þ

where the fraction s ranges from0 to 1, and then the amount of snowfalland rain are sPp and (1− s)Pp, respectively. The wet-bulb temperatureis calculated as:

Tw ¼ 6:336� 10−4 � AP � Ta þ e−6:0860:476þ 6:336� 10−4 � AP ð10Þ

where AP is the atmospheric pressure.Runoff amounts from glacierized and non-glacierized areas were

provided for GSSHA. Mass balances of glaciers and snowpacks wereestimated using the ablation method. Thus, only melted water plusrain (runoff from the glacier-melt model) was used in the GSSHA calcu-lations. As a consequence, the originalmodules of precipitation distribu-tion and snowfall accumulation and melting in the GSSHA model werereplaced and a new coupled hydrologic model was presented for theglacierized Qugaqie basin.

4. Model calibration and validation

4.1. Initial conditions and model setup

The energy balance based glacier-melt model was run with a dailytime step. The initial snow cover map was simply set to 10 cm w.e. be-cause the whole first mass balance year (October 1, 2005–September30, 2006) was set as the “spin-up” period for reducing the uncertaintiesof initial conditions. In addition, the specific precipitation correctionfactors in the “spin-up” period were applied in order to remove orweaken the initial condition effect: i.e., precipitation (both snow andrain) was increased by multiplying 3.5 and 2.5 for the winter andsummer, respectively. These factors were determined based on thecomparison between the glacier-melt model results and the mass bal-ance observation in the 2005/06 balance year. Initial estimates of soilhydraulic property values were adapted from Rawls et al. (1982).Other GSSHA parameter values with unknown spatial distributions,such as the soil depth and channel roughness coefficientswere assumedbased on the GSSHAuser'smanual (Downer and Ogden, 2006) and fieldstudies (Tian et al., 2009).

During the model running except for the “spin-up” period, summerrain amount was increased by the factor of 27%. This value was used byZhou et al. (2010) in the Zhadang Glacier for both snowfall and rainfallcorrections. In this paper, snowfall is corrected using a relatively highercorrection factor (40%)which is 1.5 times the increase factor for rainfall.In the winter season (October–May), however, we found that large un-derestimation still occurs even though the increase factor was applied.For the sake of expediency, the winter snowfall amount is correctedwith a new factor (3.5).

The glacier-melt model was implemented using 30-m grid cells inorder to better consider the topographic shading, slope and aspect ofglaciers. This is necessary as the glaciers are relatively small in sizewith steep terrainwhereas GSSHAwas run in 300-mgrid cells. It shouldbe noted that the glacier-melt model was applied to all grid cells of thestudy basin rather than only the glacier. Thus, the melt-runoff mapsfrom the glacier-melt model were re-sampled to the spatial resolutionof GSSHA. Elevations from a 30-m DEM were used to define basin

76 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

boundaries and stream networks using the Watershed Modeling Sys-tem (WMS, version 9.1) software, which is the graphic user interfacesupporting the development of inputs and analysis of results forGSSHA. The glacier boundary map in the basin was extracted fromASTER remote sensing images in 2007 (Chen, 2009).

4.2. Model performance

The radiation components in the glacier-melt model were calibratedusing the available observations of global radiation, albedo and surfacetemperature during the period of April 27, 2009–September 15, 2011(Fig. 5). In the summers of 2007 and 2008, both ZD1 and ZD2 stationshave daily global radiation data, which can be used to test the globalradiation extrapolation method, i.e., Eq. (2). Calculated global radia-tion at ZD1 based on measured global radiation at ZD2 show a goodcorrelation with its measurements (Fig. 5a). In the calibration ofsnow albedo, four coefficients in Eq. (4) were determined as below:k1 = 0.1051, k2 = −0.0026, k3 = 0.0771, and k4 = −1.0218. Inaddition, the calibrated fixed albedo values f αt1

� �for new snowfall

were 0.60, 0.75, 0.80, 0.82, 0.85 for different albedo sub-ranges atthe previous time step (i.e., [0, 0.549], [0.55, 0.689], [0.69, 0.749],[0.75, 0.799] and [0.8, 1.0], respectively). A constant value of 0.2 was

Fig. 5.Measured and calculated daily (a) global radiation at ZD1 for two summers of 2007 andtemperature for the same site and period as (b).

set to ice albedo in the glaciers. Comparison ofmeasured and calculatedalbedo at ZD3 (Fig. 5b) indicates that the calculated surface albedo is ac-ceptable for energy budgetmodeling. In the glacier-melt model, surfacetemperature is approximated by iteration: if the energy term for melt(QM) is negative, surface temperature would be forced to decreasewith a 0.25 K step to re-calculate Lout and turbulent heat fluxes. The sur-face temperature results show a close correlation with the observeddata at ZD3 (Fig. 5c). The accuracy of surface temperature simulationwas slightly lower than two previous studies (Mölg et al., 2012; Zhanget al., 2013) with the same forcing data at ZD3. This may be becausesome processes of the glacier mass balance were simplified in thisstudy. For example, the meltwater refreezing process was not consid-ered whichmay take place in a range of englacial and supraglacial loca-tions. However, the simulated accuracy of surface temperature in thisstudy was good and acceptable. In the modeling of turbulent heatflux, the aerodynamic roughness lengths (z0) adopted the constantvalue of 0.3 mm for snow surface and 0.8 mm for ice surface followinga previous study in the Parlung No. 4 Glacier on the southeast TP(Yang et al., 2011). In addition, the parameterization of Lin had been val-idated at ZD3, see Zhang et al. (2013) for detail. Therefore, in general,the glacier-melt model performed well in calculating the main energybudget terms in the Zhadang Glacier.

2008, (b) surface albedo at ZD3 for April 27, 2009 to September 15, 2011, and (c) surface

Fig. 6. Comparisons ofmeasured and calculated glacier-widemass balance in the ZhadangGlacier: (a) annualmass balance for five balance years, (b) seasonal (monthly)mass balance in the 2006/07 balance year, and (c) same as (b) but for the 2007/08 balance year.

77B.Lietal./G

lobalandPlanetary

Change118

(2014)69

–84

78 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

The glacier-melt model was validated using the mass balance data(water equivalent, w.e.) of the Zhadang Glacier (Fig. 6). Comparison ofcalculated and measured glacier-wide annual mass balance shows agood correlation for the balance years 2006/07, 2007/08, 2008/09 and2009/10 (Fig. 6a), while the simulated seasonalmass balance evolutionsin the 2006/07 and 2007/2008 mass balance years (Fig. 6b–c) showsome differences against the measurements especially for the periodsof 2006/07 winter, June 2007 and September 2008. Importantly, themodel captured well the temporary mass gain of August 2008 dur-ing the ablation period. This resulted in a slight summer mass loss(Jun.–Sep.) and further produced the positive annual mass balance.Therefore, on a whole, the glacier-melt model was acceptable for esti-mating the glacier-wide mass balance in the Qugaqie basin.

The GSSHA model was calibrated and validated with the measuredstreamflow data at the S1 and S2, which locates at the upstream anddownstream sections of the study basin. The streamflow data series inthe summer period of 2007 was used in the model calibration whilethe summer of 2008 was for model validation. Sensitive parameters inGSSHA are the overland roughness (no), channel roughness (nc), satu-rated hydraulic conductivity (Ks), and soil moisture depth (d0) fromwhich ET occurs for the GAR method (Zeweldi et al., 2011), and theywere calibrated carefully. The other parameters in GSSHA are thePenman–Monteith ET parameters (land surface albedo α0, vegetationheight hveg, vegetation radiation coefficient Krad, and canopy stomatalresistance ϑ) and GAR infiltration method (capillary head ψb, porosityθs, pore distribution index λ, residual saturation θr, field capacity θfc,and wilting point θwp). All these parameters and settings from themodel were listed in Table 1. Comparisons indicated that the integratedmodel could be applied to other high altitude glacierized basins, despitesome disagreements in both calibration and validation (Fig. 7). Table 2lists the measurement statistics of discharge at both S1 and S2 for thecalibration and validation periods. Shown are the coefficient of determi-nation (R2), the root mean square error (RMSE), the mean absoluteerror (MAE), the percentage bias (PBIAS) and the index of agreement(IOA). Results showed that it was difficult for the integrated model tocapture exactly the complex hydrograph shapes and peak dischargewith daily time steps in the Qugaqie basin (Fig. 7). In the calibration peri-od (summer 2007), the coefficients of determination (R2) were 0.64 and0.72 for S1 and S2, respectively, and the simulations were generally ac-ceptable. In the validation period (summer 2008), however, the modelperformance (R2) at S1 and S2 decreased to 0.51 and 0.5, respectively.The weak agreement of simulated discharge against the observationwas produced by the underestimation in September 2008. However, ingeneral, the integrated model captured well the hydrologic regime(river flow trends in the ablation season) with the same IOA and R2

values (0.84 and 0.57) for the upstream and downstream river sectionof the study basin.

Table 1Calibrated parameter values of the GSSHA model in the Qugaqie basin.

Soil types Land cov

Parameter Uniform Sandy loam Loam Baregrou

Ks (cm/h) – 0.89 0.66 –

ψb (cm) – 11.01 8.89 –

λ – 0.378 0.252 –

θs – 0.553 0.563 –

θr – 0.041 0.027 –

θfc – 0.307 0.370 –

θwp – 0.095 0.117 –

α0 – – – 0.12hveg (m) – – – 0Krad – – – 1.0ϑ – – – 0no – – – 0.55nc 0.03 – – –

d0 (m) 0.8 – – –

5. Results and discussion

5.1. Energy-balance components

The monthly evolution of energy-balance components analyzedduring the period of October 2006–September 2011 is identified inFig. 8. It showed that the net shortwave radiation with an averagevalue of 55 W/m2 was the main component in the energy balanceresults. Also, net shortwave radiation had an obvious seasonal cycle. Inthewinter (October–May), its average energy density for these five bal-ance years was 41W/m2while in the summer (June–September) it was85W/m2. The main reasons of the seasonal variability of net shortwaveradiation are the changes of sun elevation and surface albedo (Zhanget al., 2013). Net longwave radiationwas themain energy sink in the en-ergy balance with a small difference between the mean values in thewinter and summer (−39W/m2 vs−33W/m2). Followed energy com-ponent was the turbulent heat flux with−1 W/m2 and−11W/m2 forthe winter and summer seasons, respectively. The rainfall and groundheat flux in the energy balance results were small with the absolutemeanvalues b 1W/m2. For the groundheatflux,we used a simple inter-polation method which may underestimate its contribution to the totalheat flux. A previous study conducted in the Zhadang Glacier suggestedthat the average groundheat fluxeswere 2 and−3W/m2 for thewinterand summer, respectively, of the 2009–2011 period (Zhang et al., 2013).For the glaciers, the melting energy followed a very similar oscillationtrend as the net shortwave radiation with positive values in summerand zero in winter.

The contributions of energy flux components to total flux were alsoinvestigated for thewhole period. The proportional contribution of eachheat flux was calculated through dividing the sum of absolute energyterms (sum= |Snet| + |Lnet| + |QH + QE| + |QR| + |QG|) by each energyterm in absolute value, e.g., |Snet|/sum for net shortwave radiation. Theresults suggested that, on average, the net radiation accounted for 92%of the total heat flux with the net shortwave (longwave) radiationwas about 55% (37%). Followed was the turbulent heat flux accountingfor less than 8% of the total energy. The other two energy terms (rainfalland ground heat fluxes) contributed very small amounts (less than 1%).The results indicated that the energy budget in the glaciers of theQugaqie basin was largely governed by the radiation heat flux.

5.2. Glacier mass balance

Areal glacier mass balance patterns for each mass balance year wasconsiderably different (Fig. 9), but the general trend is large negativebalance (300 cm w.e. mass in total melted out for these five years). Orto say, up to 13.93 × 106 m3 water volumemelted out from the glaciersin five years. In general, the annual variation of mass balance for all

ers

nd Grassland Wetlands Waterbody Glacier

– – – –

– – – –

– – – –

– – – –

– – – –

– – – –

– – – –

0.15 0.10 0.05 0.450.14 0.20 0 00.6 0.6 1.0 1.071 51 0 00.68 0.76 0.3 0.75– – – –

– – – –

Fig. 7. Comparison ofmeasured and simulated flows at the upper river outflow (S1) and down river outflow (S2), respectively, with the time series of daily precipitation used in the anal-ysis and air temperature at ZD3 during the study period. The summers of 2007 and 2008 were the calibration and validation periods, respectively.

79B. Li et al. / Global and Planetary Change 118 (2014) 69–84

glaciers in the Qugaqie basin for the period of October 1, 2006–Septem-ber 30, 2011 followed the pattern of that in the Zhadang Glacier (Fig. 6).In the 2007/08 mass balance year, the glacier-wide mass balance was50 cm w.e. which indicated that the glaciers in the Qugaqie basinwere temporarily advancing. The winter/summer mass balances of theglaciers were 32 and 18 cm w.e.. This was confirmed by the observa-tions of 32 and −9 cm w.e. for winter and summer mass balances,respectively, in the Zhadang Glacier. The cause of the surplus mass bal-ance switch in 2008 appears to be lower air temperature and increasedprecipitation (Fig. 7) as discussed by Kang et al. (2009). Here, we alsocompared the main energy terms (global radiation, net longwave radi-ation, and turbulent heat fluxes) on these five summers. It was foundthat there was no obvious difference in these three energy sources orsinks. However, the net shortwave radiation in the summer of 2008was considerably less than the other summers (Fig. 8). It indicatedthat a large proportion of global radiation was reflected away. In otherwords, the increased precipitation and lower air temperature resultedin extended period of snow surface (with high albedo). Thus, lessenergy was available for snow/ice melting in the summer of 2008. Inaddition, the 2010/11 balance year also experienced only a slightly neg-ative mass balance (−14 cm w.e.). This showed that the negative

Table 2Model performances at S1 and S2 for the calibration and validation periods.

Year River section R2 RMSE MAE PBIAS IOA

2007 S1 0.64 0.31 0.24 −22.47 0.85S2 0.72 1.15 0.87 6.08 0.91

2008 S1 0.51 0.22 0.19 −15.96 0.79S2 0.50 1.40 1.46 −19.78 0.77

Total S1 0.57 0.27 0.22 −19.17 0.84S2 0.57 1.30 1.03 −8.86 0.84

glacier mass balance in the Qugaqie basin was not always in a rapidtrend. During the other mass balance years, the glaciers experiencedlarge negative mass balance especially for the summers of 2009 and2010. The calculated glacier-widemass balances for these two summerswere −171 and −173 cm w.e.. In general, the summer mass balanceclosely related to the patterns of the net shortwave radiation and melt-ing energy (Fig. 8).

Based on the glacier-meltmodel results, we calculated themeanmassbalance in each 30 m elevation band for the glaciers in the study basin.The vertical balance profile (VBP) was shown in Fig. 10. It found thatthe characteristic of the VBP was closely related to the surface energybudget and mass balance of the glaciers. In the lower elevations, the2008/09 and 2009/10 balance years, which were the two negative bal-ance periods, experienced the very considerable mass loss (N4 m w.e.).In the 2007/08 year, only the grid cells with altitude less than 5640 mexperienced the mass loss process while the other glacier surfaces hadsurplus mass balance. Two balance years, 2006/07 and 2010/11 withmass balances of −41 and −14 cm w.e., respectively, showed thestrongest mass loss up to −4 m w.e. in the lowest elevations. In thehigh elevations of the glaciers, all periods had the similar pattern ofmass gain, which indicated that the snowfall governed the mass balancecharacteristic in the high elevations. In general, five mass balance yearsshowed the similar VBP gradients. According to the VBP results, theequilibrium line altitude (ELA) was obtained (Fig. 10). The ELAs of thesebalance years generally agreed with the glacier-wide mass balanceexcept for the 2010/11 year which experienced a slightly negativemass balance but with the ELA as high as the moderate mass loss years(e.g., the 2006/07 balance year). The lowest ELAwas 5640m in the glacieradvancing year (2007/08)while the highest ELAwas 5940m in the stron-gest mass loss year (2008/09). On average, the ELA was 5798 m for theglaciers in the Qugaqie basin during the whole study period.

Fig. 8.Monthly variations of the glacier-wide energy balance terms in the Qugaqie basin for the period between October 2006 and September 2011.

80 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

5.3. Hydrologic response with the GSSHA simulations

In the active soil layer of the Qugaqie basin, freezing and thawinginfluence infiltration properties and can reduce model performance.Not all complexities associated with freezing and thawing processesare accounted for in the GSSHA model. However, modeling results insuch a high altitude basin of TP can be tolerant due to the effects of fro-zen soil in the glacier–land system. And even in non-glacierized regions,distributed hydrologic models such as GSSHA are not able to capture allhydrologic processes accurately because of data limitations (Downerand Ogden, 2004). This, however, does not diminish values distributedhydrologicmodels provide. The glacier-melt and rainfall–runoff interac-tion process in the TP is very complex as much information is still un-known in the remote region. Simulated discharge results suggestedthat the runoff process respond to the liquid precipitation and snow/ice melt only occurred in the short summer season. The maximum

Fig. 9. Simulated cumulativemass balance over the glaciers in the Qugaqie basin for the period oglaciers.

daily discharges were less than 2 and 10 m3/s in the upstream anddownstream river sections (S1 and S2), respectively, during the simu-lated period. In general, the integrated glacier-melt and hydrologicmodel somewhat underestimated streamflow except for the S2 riversection of 2008 (Fig. 7). A very obvious example was at S2 in September2008 that largely underestimated the observed values. Gao et al. (2011)suggested that this underestimation might have resulted from wind-induced snowfall undercatch by precipitation gauge. In our study,there might also be because of some possible model errors in latesummer/autumn. In the calibration period, the streamflow was notmeasured in September–October, 2007, thus biasing the calibration to-wards the summer period. In addition, it might also be because theeffects of the size of water volume in the frozen soils and the possibleburied glacial ice near the glacier terminus. The likelihood of existenceof reservoirs such as lakes in sub-, en-, and supra-glacier was excludedwith the field works. The frozen soils and buried glacial ice near the

f October 1, 2006–September 30, 2011. Arealmeanmass balances were calculated over all

Fig. 10.Distributions of the glacier area andmass balance of all glaciers in theQugaqie basinwith the altitude changes. The altitude values along the y-axis represent themiddle elevation ofeach 30 m band.

81B. Li et al. / Global and Planetary Change 118 (2014) 69–84

glacier terminus (if exists) in the study basinwould certainlymelt in thewarm season. In this study, the hydrologic process in the frozen soil isnot directly addressed but considered by calibrating the hydrologicparameters, e.g., overland roughness, and hydraulic conductivity, etc.The effect of the freezing and thawing in frozen soils on the hydrologiccycle was very complex, producing some biases in the summer period(Fig. 7). In the exact same basin, Gao et al. (2011) also simulated thehydrographs for the S2 river section with the J2000 hydrologic model,which produced the accepted results with R2 from 0.56 to 0.91 withan average of 0.73 for three summers (2006–2008). Even though ourmodel performances in this study had a slightly weaker capabilitythan J2000, the integrated model of energy-balance based glacier-meltmodel and GSSHA still make sense for the glacierized basin in the TPas the glacier-affected hydrologic process was fully considered. In ourstudy, we also investigated the glacier mass balance and hydrologic re-gime in the upstream area against the observations. More restrictions inthe model calibration may relate to the slightly worse than the J2000model. However, the adequate calibration with final acceptable resultswas very important for better understanding the complex hydrologiccycle in the study basin.

During these five years (October 2006–September 2011), an-nual water volumes from precipitation and glacier meltwater wereabout 42.64 × 106 m3 and 2.32 × 106 m3. In the total water source(44.96 × 106 m3), the ET process consumed 7.50 × 106 m3/a, whilethe final discharge volume was 37.46 × 106 m3/a. Results for the inte-grated model showed that total precipitation and the ice-melting inthe glaciers accounted for approximately 95% and 5% of the annualwater volume source for the study period (snowfall in glacier surfacewas viewed as the precipitation source). In contrast, 17% of the total

Table 3Observed runoff depth (mm) for the periods of: (a) May 28–Oct. 17, 2007 and May 28–Oct. 16months of May and October represent the available days.

Control gauge May June July

S1 2007 29.61 174.57 518.612008 9.19 135.78 229.70

S2 2007 45.98 132.77 227.672008 30.56 172.38 220.00

⁎ Daily values are calculated by averaging the total runoff to daily means.

water (rain and meltwater) was consumed by the ET process. Thus,the remaining water (83%) converted to the final basin runoff to theLake Nam Co.

5.4. Glacier runoff contribution

Glacier runoff was considered as the total runoff from snow/icemelting over the glaciers in the study. Observed runoff depths fromthe two sub-basins (with the outlets of S1 and S2, respectively) werecalculated to examine runoff capacity in the region (Table 3). Runoffdepth was calculated by dividing the runoff volume by the catchmentarea. The sub-basinwith the outlet of S2 represents its controlled drain-age area, not only the intersection. Results showed that the runoffcapacity in the upstream area was greater than in the downstreamarea in the summer of 2007, as the upper river catchment has larger gla-cier coverage. However, in the summer of 2008, the difference of dailyrunoff between the two control gauges was very small, indicating simi-lar runoff capacities in both control areas. Runoff capacities differedbetween years 2007 and 2008. This is supported by the glaciermass bal-ance results. In the summer of 2008, glaciers experienced a weak massloss while a large negative balance occurred in the summer of 2007.

For the ablation seasons, we calculated the monthly water volumesfrom the glacial runoff and the total runoff for the upstream area(with outlet of S1) and the whole Qugaqie basin (with outlet of S3)in Fig. 11. Results showed that, in the upstream area with a largeglacierized coverage, the average glacial runoff contribution to thetotal runoff was 46%, with the smallest (13%) and largest (78%) percent-ages in the strongest and weakest glacier mass loss summers of 2008and 2009, respectively. This closely related to the monthly evolution

, 2008 for S1; and (b) May 5–Aug. 26, 2007 and May 15–Oct. 16, 2008 for S2. Data for the

August September October Daily⁎

380.79 152.11 44.92 9.10237.87 184.78 66.99 6.09185.82 – – 5.20218.98 205.85 88.60 6.04

Fig. 11. Summer glacial runoff volumes from the Zhadang Glacier and all glaciers in the Qugaqie basin, and the proportional contributions to the total runoff volumes in the upstream areaand in the whole Qugaqie basin, respectively.

82 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

of energy budget and mass balance in the glaciers. For the wholeQugaqie basin, it was obvious that precipitation in the non-glacierizedarea was the main contribution, accounting for 85% of total runoff inthe summer. Similar to the upstream area, the glacial runoff volumesin the 2008 and 2009 summers were the smallest and largest contribu-tions, respectively. In the 2008 summer, only 4% water volume of thebasin runoff was produced from the 8.4% glacierized coverage area. Asa whole, the glacial runoff contribution to streamflow at both S1 andS3 in the summer of 2008 was smallest during the entire study period.This phenomenon was also found by Kang et al. (2009) based on themeasured runoff at S1 for the summers of 2007 and 2008. They pointedout that the runoff was relatively constant with similar monthly totalvalues in 2008 summer, and further suggested that the glacial runoffcontribution to river flow was lower. Therefore, the main contributionto the total runoff in the Qugaqie basin was still the precipitation eventhough the glaciers have been experienced considerable mass loss.

5.5. Model uncertainty analysis

In the glacierized regions, it is very difficult to accurately simulate orforecast the hydrologic response to glacier melting, as the glacier melt-ing may result in unavoidable biases. There are some uncertainties in

this paper in the glacier mass balance modeling (e.g., input uncertaintyand parameter uncertainty). The forcing climate data were collectedfrom different sources, e.g., both gauge observation and TRMM productwere used for precipitation. In order to reduce the input uncertainty, weused asmany in-situmeasurements as possible in themodel calibrationand validation. Also, the TRMM product was tested against groundgauge measurements, resulting in a very good accuracy. However,there are still some input uncertainties because some data gaps werefilled only with minor verification. For example, wind speed data atNAMOR station which is about 50 km away from the study basin wereusedwithout correcting in the data gaps. On thewhole, the input uncer-tainty has been reduced at much as possible, so as not to produce con-siderable bias in the results.

In this paper, we performed sensitivity runs for the changes ofparameters to test its uncertainty for the energy-balance based glaciermodel (Table 4). Results showed that the energy-balance based glaciermodel was somewhat sensitive to precipitation correction factorwhich controls the amount of precipitation. A 10% change in precipita-tion decreasing produced a larger mass balance change than the sameamplitude of precipitation increasing did. A 10% change in air tempera-ture lapse rate resulted in 5% cumulative mass balance change duringthe entire study period. Mass balance was somewhat sensitive to surface

Table 4Sensitivity of glacier-wide cumulative mass balance (CMB) over the period of October 1, 2006–September 30, 2011 to parameter changes. Relative changes in parentheses.

Parameter CMB change Winter CMB change Summer CMB change

Rain/snow correction factor: +10% +20 cm w.e. (+7%) Little change +20 cm w.e. (+5%)Rain/snow correction factor: −10% −60 cm w.e. (−20%) −9 cm w.e. (−7%) −51 cm w.e. (−12%)Air temperature lapse rate: +10% +14 cm w.e. (+5%) −1 cm w.e. (−1%) +14 cm w.e. (+3%)Air temperature lapse rate:−10% −15 cm w.e. (−5%) +1 cm w.e. (+1%) −15 cm w.e. (−4%)Snow albedo coefficient k1: +10% −17 cm w.e. (−6%) −1 cm w.e. (−1%) −16 cm w.e. (−4%)Snow albedo coefficient k1: −10% +22 cm w.e. (+7%) +1 cm w.e. (+1%) +21 cm w.e. (+5%)Snow albedo coefficient k2: ±10% Little change⁎ Little change Little changeSnow albedo coefficient k3: +10% −3 cm w.e. (−1%) Little change −3 cm w.e. (−1%)Snow albedo coefficient k3: −10% +4 cm w.e. (+1%) Little change +3 cm w.e. (+1%)Snow albedo coefficient k4: +10% +4 cm w.e. (+1%) Little change +4 cm w.e. (+1%)Snow albedo coefficient k4: −10% −4 cm w.e. (−1%) Little change −4 cm w.e. (−1%)Ice albedo: +0.1 +88 cm w.e. (+29%) +7 cm w.e. (+6%) +81 cm w.e. (+19%)Ice albedo:−0.1 −87 cm w.e. (−29%) −8 cm w.e. (−7%) −79 cm w.e. (−19%)Roughness length snow: +0.1 mm −2 cm w.e. (−1%) Little change −2 cm w.e. (−1%)Roughness length snow: −0.1 mm +3 cm w.e. (+1%) Little change +3 cm w.e. (+1%)Roughness length ice: ±0.1 mm Little change Little change Little change

⁎ Little change represents that the absolute CMB change b1 cm w.e. and its relative change b1%.

83B. Li et al. / Global and Planetary Change 118 (2014) 69–84

albedo specification but not sensitive to all albedo parameters or con-stants.When ice albedo value increases or decreases by 0.1 (50% change),29% more glacier mass gain or loss occurs; and the larger mass changecould be found in summer. In the snow albedo approximation, only thefirst constant (k1) seemed to be sensitive while the changes in theother constants had smaller effect on glacier mass balance. As for theroughness length for snow or ice, it was also insensitive to mass balance,while 0.1 mm change only resulted in up to 1% relative change in theglacier mass. In this study, most approximation methods in the energy-balance based glacier model were tested in the Zhadang Glacier within-situ measurements.

In the hydrologic model, the limitation in simulating frozen soilwater may result in unavoidable biases in the final results. However,the freezing and thawing processes are very complex and could bemodeled well with sufficient in-situ measurements. No field data offrozen soil were available in this remote glacierized basin. Thus, wedid not directly address the effect of frozen soil on hydrologic processbut considered it by calibrating the GSSHA model parameters. Thistype of uncertainty in model structure is expected and perhaps un-avoidable under the current data conditions in such areas.

6. Summary and conclusions

An energy-balance based glacier-melt model and GSSHA hydrologicmodel were integrated to examine the hydrologic response in the highaltitude glacierized watersheds in the TP. The glacier-melt model calcu-lates snow and icemelts at each grid cell within the basin (including thenon-glacierized area) according to resulting melt energy from energybalance. Output of the glacier-melt model, i.e. runoff at each grid cellwithin the study area, was considered as GSSHA input (similar to pre-cipitation data in the standard GSSHA model). The original snowmodel in the standard GSSHA model was replaced by the glacier-meltmodel. Results indicated that the integrated model can be applied tohigh altitude glacierized watersheds in the TP, despite some disagree-ments in river flow comparisons. In addition, the contributions of hy-drologic components to the water balance were examined for theQugaqie basin which is jointly fed by precipitation and glacier melting.

Results from the glacier-meltmodel indicated that the glaciers in theQugaqie basin experienced large mass loss between October 2006 andSeptember 2011 with−300 cmw.e. mass balance in total. All five con-sidered mass balance years were experienced the negative glacier massbalance except for the 2007/08 year. Relatively lower air temperatureand increased precipitation in the 2008 summer were considered asthe main reasons for the positive mass balance in the 2007/08 year.However, the surplus glacier mass balance was temporary. In the nextyears, the glaciers switched back to the negative mass balance trend.

An energy-balance based glacier-melt model and GSSHA hydrologicmodel were integrated to examine the summertime hydrologicresponse in the Qugaqie watershed, which is representative of highaltitude glacierized watersheds in the TP. River flow observations attwo sites located at the upstream and downstream areas were used toevaluate model performance during the summers of 2007 and 2008,producing the general acceptable results. In high altitude watershedssuch as the Qugaqie basin, the glacier-melt and runoff-yield processesare very complex as they are affected by underground ice and frozensoils, making it difficult to accurately capture the hydrologic response.Water balance analysis also indicated that the ice melting from the gla-ciers only accounted for 5% the total water sources in this 8.4% glacier-covered basin. During the summertime, the glacial runoff (snow/icemelting) contribution to river flow was important, accounting for 46%and 15% of the total runoff volumes in the upstream area and thewhole Quagqie basin, respectively. Therefore, precipitation is the mainwater balance component in the high altitude glacierized basin of theTP.

Acknowledgments

Theworkwas funded by the Urban Flood Demonstration Program forArid and Semi-Arid Regions (UFDP) (grant no. W912HZ-8-2-0021), theNational Basic Research Program of China (grant no. 2010CB951101),the Natural Science Foundation of China (grant no. 51079039), and theScientific Research Innovation Project for Graduate Students of Jiangsuprovince, China (grant no. CX10B_210Z). The soil type and land coverdata sets are provided by the Environmental and Ecological ScienceData Center for West China, National Natural Science Foundation ofChina (http://westdc.westgis.ac.cn). The mass balance data of theZhadang Glacier and a part of climate data were provided by the ThirdPole Environmental (TPE) Database (http://www.tpedatabase.cn).Additionally, we would like to thank the GSSHA model developersMichael L. Follum and Charles W. Downer for their helpful suggestionson model setup.

References

Andreas, E.L., 1987. A theory for the scalar roughness and the scalar transfer coefficientsover snow and sea ice. Bound.-Layer Meteorol. 38 (1–2), 159–184.

Bolch, T., Yao, T., Kang, S., Buchroithner, M.F., Scherer, D., Maussion, F., Huintjes, E.,Schneider, C., 2010. A glacier inventory for the western Nyainqentanglha Range andthe Nam Co Basin, Tibet, and glacier changes 1976–2009. Cryosphere 4, 419–433.

Chen, F., 2009. Glacier Changes in theMt. Nyainqêntanglha Region in Response to ClimateChange — Application of RS and GIS. Chinese Academy of Sciences, Beijing.

Crawford, T.M., Duchon, C.E., 1999. An improved parameterization for estimating effectiveatmospheric emissivity for use in calculating daytime downwelling long-wave radia-tion. J. Appl. Meteorol. 38, 474–480.

84 B. Li et al. / Global and Planetary Change 118 (2014) 69–84

Downer, C.W., Ogden, F.L., 2004. GSSHA: model to simulate diverse stream flow produc-ing processes. J. Hydrol. Eng. 9 (3), 161–174.

Downer, C.W., Ogden, F.L., 2006. Gridded Surface Subsurface Hydrologic Analysis(GSSHA) User's Manual (Version 1.43 for WMS 6.1). U.S. Army Corps of Engineers.

Forrer, J., Rotach, M.W., 1997. On the turbulence structure in the stable boundary layerover the Greenland ice sheet. Bound.-Layer Meteorol. 85 (1), 111–136.

Gao, T., 2011. Analysis and Simulation of the Hydrological Processes in the Nam Co Basin,Tibetan Plateau. Chinese Academy of Sciences, Beijing.

Gao, T., Kang, S., Krause, P., Cuo, L., Nepal, S., 2011. A test of J2000 model in a glacierizedcatchment in the central Tibetan Plateau. Environ. Earth Sci. 65 (6), 1651–1659.

Hock, R., 2003. Temperature index melt modelling in mountain areas. J. Hydrol. 282,104–115.

Hock, R., 2005. Glacier melt: a review of processes and their modelling. Prog. Phys. Geogr.29 (3), 362–391.

Hock, R., Holmgren, B., 2005. A distributed surface energy-balance model for complex to-pography and its application to Storglaciären, Sweden. J. Glaciol. 51 (172), 25–36.

Hock, R., Noetzli, C., 1997. Areal melt and discharge modelling of Storglaciaren, Sweden.Ann. Glaciol. 24, 211–216.

Immerzeel,W.W., Beek, L.P.H.v., Konz, M., Shrestha, A.B., Bierkens, M.F.P., 2012. Hydrolog-ical response to climate change in a glacierized catchment in the Himalayas. Clim.Chang. 110, 721–736.

Kang, S., 2011. Modern Environmental Process and its Variation of the Nam Co Basin,Tibetan Plateau. China Meteorological Press, Beijing.

Kang, S., Chen, F., Gao, T., Zhang, Y., Yang, W., Yu, W., Yao, T., 2009. Early onset of rainyseason suppresses glacier melt: a case study on Zhadang glacier, Tibetan Plateau. J.Glaciol. 55 (192), 755–758.

Khakbaz, B., Imam, B., Hsu, K., Sorooshian, S., 2012. From lumped to distributed via semi-distributed: calibration strategies for semi-distributed hydrologic models. J. Hydrol.418, 61–77.

Kidd, W., Yusheng, P., Chengfa, C., Coward, M., Dewey, J., Gansser, A., Molnar, P.,Shackleton, R., Yiyin, S., 1988. GeologicalMapping of the 1985Chinese–British Tibetan(Xizang–Qinghai) Plateau Geotraverse Route. Philos. Trans. R. Soc. Lond. Ser. A Math.Phys. Sci. 327 (1594), 287–305.

Krause, P., Biskop, S., Helmschrot, J., Flugel, W.A., Kang, S., Gao, T., 2010. Hydrological sys-tem analysis and modelling of the Nam Co basin in Tibet. Adv. Geosci. 27, 29–36.

Liu, Z., Todini, E., 2002. Towards a comprehensive physically-based rainfall-runoff model.Hydrol. Earth Syst. Sci. 6 (5), 859–881.

Maussion, F., Wei, Y., Huintjes, E., Pieczonka, T., Scherer, D., Yao, T., Kang, S., Bolch, T.,Buchroithner, M., Schneider, C., 2011. Glaciological field studies at Zhadang Glacier(5500–6095m), Tibetan Plateau. In: Tijm-Reijmer, C.H., Oerlemans, J. (Eds.), Workshopon the Use of Automatic Measuring Systems on Glaciers, Pontresina, Switzerland,pp. 62–68.

Mölg, T., Maussion, F., Yang,W., Scherer, D., 2012. The footprint of Asianmonsoon dynam-ics in the mass and energy balance of a Tibetan glacier. Cryosphere 6 (6), 1445–1461.

Mölg, T., Maussion, F., Scherer, D., 2014. Mid-latitude westerlies as a driver of glacier var-iability in monsoonal High Asia. Nat. Clim. Chang. 4, 66–73.

Munro, D.S., 1990. Comparison of melt energy computations and ablatometer measure-ments on melting ice and snow. Arct. Alp. Res. 22 (2), 153–162.

Nachtergaele, F., Velthuizen, H.v, Verelst, L., 2008. HarmonizedWorld Soil Database. Foodand Agriculture Organization of the United Nations.

Paulson, C.A., 1970. The mathematical representation of wind speed and temperatureprofiles in the unstable atmospheric surface layer. J. Appl. Meteorol. 9 (6), 857–861.

Ragettli, S., Pellicciotti, F., 2012. Calibration of a physically based, spatially distributed hydro-logical model in a glacierized basin: on the use of knowledge from glaciometeorologicalprocesses to constrain model parameters. Water Resour. Res. 48, W03509.

Ran, Y., Li, X., Lu, L., 2009. China land cover classification at 1 km spatial resolution based amulti-source data fusion approach. Adv. Earth Sci. 24 (2), 192–203 (in Chinese).

Rawls, W.J., Brakensiek, D.L., Saxton, K.E., 1982. Estimation of soil water properties. Trans.ASAE 1316–1320.

Tian, K., Liu, J., Kang, S., Campbell, I.B., Zhang, F., Zhang,Q., Lu,W., 2009.Hydrothermal patternof frozen soil in Nam Co lake basin, the Tibetan Plateau. Environ. Geol. 57, 1775–1784.

Todini, E., 1996. The ARNO rainfall–runoff model. J. Hydrol. 175, 339–382.Yamazaki, T., 2001. A one-dimensional land surface model adaptable to intensely cold re-

gions and its applications in Eastern Siberia. J. Meteorol. Soc. Jpn. 79 (6), 1107–1118.Yang, K., Huang, G.-W., Tamai, N., 2001. A hybrid model for estimating global solar radia-

tion. Sol. Energy 70, 13–22.Yang, K., Koike, T., Ye, B., 2006. Improving estimation of hourly, daily, and monthly solar

radiation by importing global data sets. Agric. For. Meteorol. 137, 43–55.Yang, W., Guo, X., Yao, T., Yang, K., Zhao, L., Li, S., Zhu, M., 2011. Summertime surface en-

ergy budget and ablation modeling in the ablation zone of a martime Tibetan glacier.J. Geophys. Res. 116, D14116.

Yu, W., Yao, T., Kang, S., Pu, J., Yang, W., Gao, T., Zhao, H., Zhou, H., Li, S., Wang, W., Ma, L.,2013. Different region climate regimes and topography affect the changes in area andmass balance of glaciers on the north and south slopes of the same glacierized massif(the West Nyainqentanglha Range, Tibetan Plateau). J. Hydrol. 495, 64–73.

Zeweldi, D.A., Gebremichael, M., Downer, C.W., 2011. On CMORPH rainfall for streamflowsimulation in a small, hortonian watershed. J. Hydrometeorol. 12, 456–466.

Zhang, G., Kang, S., Fujita, K., Huintjes, E., Xu, J., Yamazaki, T., Haginoya, S., Wei, Y., Scherer,D., Schneider, C., Yao, T., 2013. Energy and mass balance of Zhadang glacier surface,central Tibetan Plateau. J. Glaciol. 59 (213), 137–148.

Zhao, R.J., 1992. The Xinanjiang model applied in China. J. Hydrol. 135, 371–381.Zhou, S., Kang, S., Gao, T., Zhang, G., 2010. Response of Zhadang Glacier runoff in Nam Co

Basin, Tibet, to changes in air temperature and precipitation form. Chin. Sci. Bull. 55,2103–2110.