Integration of aspect and slope in snowmelt runoff modeling in a mountain watershed

Shlmu Audu*,Zhu-ping Sheng,Chun-ling Cui,Mutter Sydi,Hmed-Zmni Szi, Jmes Phillip King

aTexas A&M AgriLife Research and Extension Center at El Paso,El Paso,TX 79927-5020,USA

bXinjiang Water Resources Research Institute,Urumqi 830049,China

cCivil Engineering Department,New Mexico State University,Las Cruces,NM 88003-8006,USA

Integration of aspect and slope in snowmelt runoff modeling in a mountain watershed

Shalamu Abudua,b,*,Zhu-ping Shenga,Chun-liang Cuib,Muatter Saydib,Hamed-Zamani Sabzic, James Phillip Kingc

aTexas A&M AgriLife Research and Extension Center at El Paso,El Paso,TX 79927-5020,USA

bXinjiang Water Resources Research Institute,Urumqi 830049,China

cCivil Engineering Department,New Mexico State University,Las Cruces,NM 88003-8006,USA

This study assessed the performances of the traditional temperature-index snowmelt runoff model(SRM)and an SRM model with a fi ner zonation based on aspect and slope(SRM+AS model)in a data-scarce mountain watershed in the Urumqi River Basin,in Northwest China. The proposed SRM+AS modelwas used to estimate the meltrate with the degree-day factor(DDF)through the division ofwatershed elevation zones based on aspect and slope.The simulation results of the SRM+AS model were compared with those of the traditional SRM model to identify the improvements of the SRM+AS model's performance with consideration of topographic features of the watershed.The results show that the performance of the SRM+AS model has improved slightly compared to that of the SRM model.The coefficients of determination increased from 0.73,0.69,and 0.79 with the SRMmodelto 0.76,0.76,and 0.81 with the SRM+AS modelduring the simulation and validation periods in 2005,2006,and 2007,respectively.The proposed SRM+AS model that considers aspect and slope can improve the accuracy of snowmelt runoff simulation compared to the traditional SRM model in mountain watersheds in arid regions by proper parameterization,careful input data selection,and data preparation.

Snowmelt runoff model(SRM);Degree-day factor(DDF);Aspect and slope;Snow cover area;Temperature;Precipitation

1.Introduction

In the arid regions of the world,mountain-fed rivers are the only available water resources that cover the needs of public water supply,agricultural irrigation,hydropower,and other uses(Li and Williams,2008).In Northwest China,the mountain snowmeltwater is regarded as the main water source of the oases scattered throughout the river basins in the piedmontregions where humans live and engage in production activities.In recent years,global climate change,such as global warming,has affected the local hydrology and water cycle in these areas.Particularly,the glacier and snow covered areas have been greatly influenced as they are extremely sensitive to temperature factors and can be signifi cantly affected by the timing and magnitude of the streamfl ow in the watershed(Saydi et al.,2013;Ma and Cheng,2003;Ma etal., 2004).For the continental river basins in Northwest China, snowmelt water is vital to the occurrence of streamfl ow. Glacier and snowmelt water derived from the northern slopes of Tianshan Mountains accounts for almost 80%of the annual runoff.Hence,to accurately simulate and forecast the snowmelt runoff is necessary in this region.However,there is asparse network of hydro-meteorological observations and limited ground-based data because of the high-elevation and rugged mountain landforms in the watersheds of the northern Tianshan Mountains(Ma and Cheng,2003).

Numerous snowmeltrunoff models have been developed to simulate or forecast snowmelt runoff in different hydrologic conditions.They can be categorized into two major groups based on their approaches:empirical temperature-index models and physical energy-balance models(Wu et al., 2011;Pellicciotti et al.,2005;Hock,2003;Mehta et al., 2004;Kang and Merwade,2011;Boudhar et al.,2009).The applicability of the two approaches is usually restricted to the simulation of melt rates,which in turn motivates the development of various methods regarding the calculation procedure(Pellicciotti et al.,2005).The standard temperature-index model calculates the melt rate as a linear function of temperature,with the factor of proportionality being the degreeday factor(Hock,1999,2003;Pellicciotti et al.,2005; Martinec et al.,2008;Wang et al.,2010).In this approach, calculation of the melt rate is specified by associating the temperature lapse rate with the elevation change in respect to time and space,the localized critical temperature is defined, and the standard temperature-index model calculates the amount of melt water by converting the degree-day above criticaltemperature into the meltdepth.Itis a simplification of complex physical surface energy process by means of temperature and the degree-day factor.Karimi et al.(2016) assessed the performance of the WetSpa and SRM models in snowmelt runoff simulation.In their research,the WetSpa model outperformed the SRM model since the former is a distributed spatial-physicalmodel that simulates snow melting in a cellular manner at hourly and daily time steps using energy balance and degree-day approaches.Zhang et al.(2009) proposed a method for simulation of the snowmelt process using both the WinSRM 1.10 and a self-developed snowmelting runoff simulation and forecast software,SRSFS 1.0. They forecasted daily runoff of the study basin in the spring of 2004 with the self-developed SRSFS 1.0 based on T213 meteorological data from China Meteorological Administration (CAM) to obtain zonal mean temperature and precipitation.

An important problem in application of snowmelt runoff models is data scarcity due to the difficulties in the physical accessibility and the lack of hydro-meteorological measurements in high-altitude mountain watersheds,particularly in the northwestern region of China where most snow-dominated basins are located and data are insufficient.A typical example was the analysis of the snowmeltrunoff process in the Kaidu River Basin in Xinjiang,China by Zhang et al.(2006). In their research,based on the basin's characteristics,the authors introduced two coefficients for temperature and precipitation,with the corresponding products between the two groups of coefficients and variables chosen as the fi nal inputs to the SRM model.The coefficients'values were then adjusted empirically according to the simulation results through iterative experiments.It has been considered that in such datascarce remote mountain basins where data extrapolated from hydro-meteorological stations are probably unreliable,this method can be recommended and is of real signifi cance to improving the modeling and forecasting accuracy.Li and Wang(2008)studied the application of the SRM model to the Heihe River Basin,where water resources are supplied mostly by snowmelt runoff and limited hydro-meteorologica1 data are available.Results show that the SRM model is a reliable and appropriate method of examining the snowmelt runoff patterns of river basins in arid lands where human observation is sparse.

The shortwave solar radiation is the dominant source of melt energy.In addition,topographic features and surface conditions,particularly,and surface roughness and albedo, have significant influence on melt rate calculation(Hock, 1999;Kondo and Yamazaki,1990;Pellicciotti et al.,2005). Therefore,new approaches including the shortwave solar radiation and snow albedo have been developed to calculate the meltrate.An extensive discussion about the calculation of the melt rate can be found in Pellicciotti et al.(2005).Improvements in melt rate calculation have contributed to the advancement in snowmelt runoff modeling.Vafakhah et al. (2015)developed an SRM radiation model for simulating snowmelt runoff in the Taleghan Watershed in Iran.With the new method of melt rate calculation,Li and Williams(2008) developed an enhanced temperature-index model that incorporated the shortwave solarradiation and snow albedo with the degree-day factor and applied the model to the Yarkant River Basin,which is the largest tributary of the Tarim River Basin, in Xinjiang,in China.Their research explored the feasibility of modeling snowmelt runoff in a data-sparse mountain watershed by modifying existing snowmelt models to develop an enhanced temperature-index model,which uses satellitederived snow cover data and varied degree-day factors based on the shortwave solar radiation and snow albedo.

In snowmelt runoff models,temperature is a dominant factor that controls the whole process of snowmelt,and the models'performances are extremely sensitive to temperature input.Usually,the temperature values recorded at one or more stations within a basin or ambientareas are extrapolated to the whole basin with a certain locallapse rate,and the temperature data are adjusted only against the elevation change.However, in a mountain watershed,the localized topography generally affects the temperature.The topographic features such as aspect and slope exert considerable influence on the temperature variation(Hock,1999;Tang and Fang,2006).Kang and Lee(2014)explored the snowmelt runoff mechanism by relating the temperature changes to the elevation band in the North Fork American River Basin.In their study,a distributed hydrologic model was used to evaluate the orographic effects on the snowmelt runoff based on the snowfall-snowmelt routine in the soil and water assessment tool(SWAT).The outcomes of their research suggest that the snowmelt runoff model associated with the elevation band represents the snowmelt runoff mechanism well.Hence,ways of adjusting the temperature againstthe topographic factors such as aspect, slope,and elevation and accounting for their effects on the snowmelt simulation are of vital importance,particularly inthe mountain watersheds.In this study,an enhanced snowmelt runoff modelwas developed using the temperature-index SRM model with a zonation based on aspect and slope in the mountain watershed of the Urumqi River,in Xinjiang,in China.The simulation results of the proposed model were compared with those of the traditional SRM model to evaluate the performance of the proposed model with consideration of topographic features of the watershed.

2.Study area and data

The glacier and snow stored on the northern slopes of the Tianshan Mountains in Xinjiang melt to feed many rivers in spring and summer.The discharge ofthe UrumqiRiverismainly fed by snowmeltand precipitation with proportions of 37%and 36%,respectively(Gong et al.,2009).The Urumqi River originates at Glacier No.1 near Tengri Peak No.IIin the Tianshan Mountains.The river basin is located between longitudes of 86°45′E and 87°56′E and latitudes of 43°00′N and 44°07′N.It covers an area of4684 km2,with a totalriverlength of214 km.A length of 62.6 km of the river is situated in the mountain area upstream of the Yingxiongqiao Hydrological Station.The river course flows across the city of Urumqi,and finally disappears in the Gurbantunggut Desert.According to the statistics from the Yingxiongqiao HydrologicalStation overthe period from 1958 to 2009,the average annualrunoffis 2.42×108 m3,the average temperature is 1.6°C,and the average annual precipitation is 456.9 mm(Saydietal.,2013).

Fig.1.Geographic location of Urumqi River Watershed and distributions of hydrological and meteorological stations.

As shown in Fig.1,the watershed area located between the Daxigou Meteorological Station at the headwaters and the Yingxiongqiao Hydrological Station at the mountain outlet in the upper reaches of the Urumqi River Basin was selected as the study watershed.The watershed covers an area of 1073.64 km2,and its elevation ranges from 1683 m to 4459 m with a mean elevation of 3066 m.Within the watershed,there are two hydrologicalstations,the Yingxiongqiao Hydrological Station located atthe outletof the watershed atan elevation of 1920 m and the Yuejinqiao Hydrological Station located in the middle of the upper reaches at an elevation of 2313 m,as well as one weather station,the Daxigou Meteorological Station in the high-altitude mountain area at an elevation of 3539 m.

3.Methodology

3.1.Model configuration and time step

The study watershed is divided into five elevation zones(A, B,C,D,and E)for snowmelt runoff modeling.The related information is specifi ed in Table 1.Three years of daily data from 2005 to 2007 were used for snowmelt runoff modeling. The model was calibrated in the year 2005,and validated in the years 2006 and 2007.The daily mean runoff,daily mean temperature,and precipitation data from the Yingxiongqiao Hydrological Station,as well as daily mean temperature and precipitation data from the Yuejinqiao Hydrological Station and Daxigou Meteorological Station were collected and compiled for snowmelt runoff modeling in the study area.

3.2.Models

3.2.1.Traditional snowmelt runoff model(SRM)

One of the most commonly used models for snowmelt runoff simulation is the SRM model,which is designed to simulate and forecast daily streamflow in mountain basins where snowmelt is a major runoff factor.It calculates the runoff mainly through snowmelt and precipitation with some additional deterministic parameters to describe the basin characteristics.The SRM model computes the daily discharge of a basin for a lag time of 18 h by

where Q is the average daily discharge(m3·s-1);c is the runoff coeffi cient,with cSreferring to snowmelt and cRreferring to rain;a is the degree-day factor(cm·°C-1·d-1);T is the number of degree-days(°C·d);ΔT is the adjustment by the temperature lapse rate(°C·d);the composition of cSa(T+ΔT)is the melt rate;S is the ratio of the snow cover area(SCA)to the totalarea;P is the precipitation contributingto runoff(cm);A is the area of the basin or zone(km2);k is the recession coeffi cient;and n is the sequence of days during the discharge computation period.Eq.(1)is builtup for a time lag between the daily temperature cycle and the corresponding discharge cycle of 18 h.The conversion from cm·km2·d-1to m3·s-1(conversion from runoff depth to discharge)is 10000/86400.The snowmelt runoff model can be used to simulate daily flows in a snowmelt season,in a year,or in a sequence of years,provide short-term and seasonal runoff forecasts,and evaluate the potential effect of climate change on the seasonalsnow cover and runoff(Martinec et al.,2008).

Table1 Elevation zones in Urumqi River Watershed.

3.2.2.Temperature-index SRM modelwith aspectand slope (SRM+AS)

The SRM+AS model,different from the SRM model, considers aspect and slope in the zonation.In this approach, the SRM model was applied to the study watershed with 14 elevation zones that subdivided the watershed based on the differences in aspect and slope.Many studies have found that there is a high correlation between the melt rate and temperature when the temperature is presented as a principalvariable (Kang,2005;Richard and Gratton,2001).Usually,the temperature data recorded at one or more stations were adjusted against the elevation change of a given basin.However,in the high-altitude mountain area,in addition to the elevation, topographic factors such as slope and aspect can have significant effects on the temperature.The aspects against sunlight to the north,south,east,and west can be the factors that affect the air temperature.Kang(2005)reported that the snow albedo,which is the snow refl ectance against sunlight,changes with the snow surface characteristics and snow melting processes.Hence,considering the importance of topography,the integration of the topographic features of aspect and slope into the determination of numbers of degree-days in the temperature-index model may improve the accuracy of the SRM model in the mountain watersheds.

In addition to the elevation,the temperature change with aspect is also signifi cant.Kang(2005)reported that the air temperature of the northern face is on average 6°C lower than that of the southern face.Variations in aspects may produce the corresponding temperature variations in the watershed.In the north,south,east,west,and the northern,southern,eastern, and western parts the temperature variations due to the aspect are given in Table 2(Kang,2005).Table 2 presents the changing values of air temperature based on aspect values.As to the temperature variations in the western and eastern aspects,solar radiation that reaches the ground surface duringdifferent time periods from sunrise to sunset decreases gradually,and the temperature changes accordingly.Hence,the parameterθwas introduced to specify the temperature variations during different periods from sunrise to sunset.

Table2 Temperature change with aspect(Kang,2005).

θcan be defi ned by the following equation:

where t is the simulation time,ranging from 0 to 24 h.

In the study watershed,the statistics of area proportions of different aspect ratio groups in relation to the watershed area can be derived from the digital elevation model(DEM).Results show that the accumulated area corresponding to aspects of 0,90,180,270,and 360°accounts for a very small proportion of the whole watershed(0.87%),and therefore can be neglected.In addition,the runoff in this study was routed at a daily step,and the model did not account for the difference in temperature due to different time periods within a day. Therefore,the parameterθ in the temperature variation calculation can also be neglected.Accordingly,the temperature variations due to aspects in this study can be regrouped as shown in Table 3.

The air temperature also changes with the slope in the watershed.The fl atter surface receives more solar radiation than the higher slope surface.Hence,the reduction of air temperature with high slope factors can be added to calculate the snowmelt.The slope factors contribute to the air temperature based on 15°(Kang,2005).Eq.(3)shows the contribution of slope to air temperature adjustment.

where TSis the adjustment value of temperature due to slope, and S0is slope in degrees.

Eq.(4)shows the contributions of final slope,aspect,and elevation to air temperature,as follows:

where TZis the adjustment value of temperature due to slope, aspect,and elevation in each sub-zone;Tais the change of air temperature due to the aspects(in Table 3);and Televis the air temperature change due to elevation.

In a mountain watershed,temperature affects the snowmelt process and runoff mainly through the snow cover area.In the study watershed,the snow coverarea is mainly distributed in the higher elevation zones C,D,and E.The lower elevation zones (zones Aand B)are covered with less orno snow,and meltaway in the very early meltseason.Hence,the higher elevation zonesare considered forthe adjustmentoftemperature with respectto topographic features,and each of the elevation zones C,D,and E is furtherdivided into foursub-zones according to theiraspect groups summarized in Table 3.These sub-zones are recorded as C1,C2,C3,C4,D1,D2,D3,D4,E1,E2,E3,and E4.Therefore,the whole study watershed is divided into 14 elevation zones, including zones A and B plus 12 sub-zones.

Table3 Temperature change with aspect utilized in this study watershed.

3.3.Model variables and parameters

Temperature,precipitation,and snow-covered area are the three input variables of the SRM model.Usually,temperature and precipitation are measured at the hydro-meteorological station within or near a basin,while the snow-covered area is extracted from satellite image.The quality of the inputdata is directly related to the model accuracy and effi ciency,and the error in input data is proportional to the resulting accumulative errors in the calculated snowmelt(Rango and Martinec,1981).There are three hydro-meteorological stations located at different elevations in the study watershed.

3.3.1.Temperature

Temperature lapse rates are calculated using daily records of the average temperature.The temperature for the 14 elevation zones is calculated as follows:the temperature in zones Aand B is evaluated directly using the lapse rates and reference station data.The temperature forsub-zones of C,D,and E is calculated with the following steps:fi rst,the temperature change due to the aspect is determined according to Table 3;second,the distribution of temperature adjustment based on the slope data is calculated with Eq.(3);third,the temperature adjusted by slope and aspects is added to the extrapolated temperature atthe mean elevation of each elevation zone(as shown in Table 1) using Eq.(4).

3.3.2.Precipitation

Precipitation data are among the most important components of the snowmelt runoff modeling,yet most diffi cult to evaluate.In watersheds with a large area,precipitation should be extrapolated to each elevation zone(as shown in Table 1) with an altitude gradient.In the precipitation extrapolation,it is of great signifi cance to obtain the representative station that can describe the given basin's characteristics well.Hence, before extrapolating the daily records,the correlation between the precipitation from three stations and the runoff measured at the Yingxiongqiao Hydrological Station is investigated.

3.3.3.Snow cover mapping

Snow cover area(SCA)is significant to the SRM model's accuracy.Errors in determining the snow-covered areas are directly proportional to the resulting errors in the calculated snowmelt(Rango and Martinec,1981).Because of the high spatial and temporal variability of snow cover,remote sensing observations are particularly useful for providing spatially detailed input data for snowmelt runoff modeling(Nagler et al.,2008).Moreover,the SRM model is optimized for input of remotely sensed snow cover data while many other models have not been designed for remote sensing data input (Abudu etal.,2012).Hence,satellite-derived snow cover is an efficient and readily available input for snowmelt runoff modeling.

There are several sensors available for mapping the snow cover area;the Landsat 5 Thematic Mapper(TM),the National Oceanic and Atmospheric Administration-Advanced Very High Resolution Radiometer(NOAA-AVHRR),the European Remote Sensing-Synthetic Aperture Radar(ERSSAR),and the Moderate Resolution Imaging Spectroradiometer(MODIS)are the most important and widely used sensors in snow cover mapping.Compared with other sensors, SCA extracted from the MODIS data is more desirable for use in the snowmelt runoff model,mainly because the MODIS data need not be processed to classify the snow from the other features,and have been improved regarding special resolution and geolocation accuracy(Hall et al.,2002;Tekeli et al., 2005).The snow-mapping algorithm is designed to generate a global snow cover product using Earth Observation System (EOS)MODIS data(Hall et al.,2002).The Level-3 products will provide daily(MOD 10A1)and eight-day composites of global snow cover(MOD 10A2)at a 500-m resolution.Snow cover over eight days is mapped as the maximum snow extent (Hall et al.,2002).MODIS/Terra 8-day composites of global snow cover products were used in this study,and the data have been available through the Distributed Active Archive Center (DAAC)at the National Snow and Ice Data Center(NSIDC) of the National Aeronautics and Space Administration (NASA).The data format is converted and projected by applying MODIS projection tools(MRT2003)to the World Geodetic System 1984(WGS84)and Universal Transverse Mercator(UTM).During the study period,24 MOD 10A2 data were available in each year,and the imagery was mostly cloud-free.

3.3.4.Recession and runoff coefficients

Mountain watersheds usually have limited observation stations and insuffi cient hydro-meteorological data.In these cases,model parameters cannot be measured precisely with the historical data,and should be estimated mostly by hydrological judgment with the basin characteristics and physical laws taken into account(Martinec et al.,2008).In a large basin covering a variety of climatic zones,strict attention is required to calibrate the model parameters,such as the recession coeffi cient and runoff coefficient,particularly (Rango and Martinec,1981;Abudu et al.,2012).

The recession coefficient k is notconstant;itvaries with the changing discharge and can be defined by the following expression(Li and Williams,2008):

The recession coefficientis an important factor of the SRM model,indicating the decline of discharge in a period without snowmelt and rainfall(Martinec et al.,2008).Eq.(5)can be reformatted by taking the natural log on both sides of the equation:

where Qnand Qn+1are the discharges on the n th and(n+1)th day,respectively;and a,b,x,and y are the regression parameters that can be obtained through linear regression with observed recession discharge data.The recession coeffi cientin this study was calculated with the regression of runoff in the true recession period at the Yingxiongqiao Hydrological Station(Li and Williams,2008;Li and Wang,2008).From the linear function,the constants a and b can be obtained through regression;and x and y can be derived accordingly.Hence,the recession coefficient can be evaluated at the Yingxiongqiao Hydrological Station as follows:

Runoff coeffi cients are essentialparameters as they account for how much of the snowmeltand precipitation can contribute to the runoff.In some semi-arid basins,the runoff coefficient can have an even lower value,particularly in the lowest elevation zones(Martinec etal.,2008).In general,runoffcoefficients are first specifi ed according to the basin characteristics over time,and then adjusted against the model performance where necessary.The degree-day factor,due to the absence of observed data,is calculated through the empirical relationship between snow density and degree-day factor recommended by Martinec et al.(2008).The snow density is referenced from related previous works in the Urumqi River Watershed(Zhang et al.,2006;Huang et al.,2007).Model parameters for the five elevation zones can be specified as shown in Table 4.

3.3.5.Performance measures

Using the 18-h lag time,the two SRM models were calibrated in the snowmelt season from March to August in the hydrologic year 2005.With the parameters adopted for the simulation period,the models were validated during the same period from 2006 to 2007.Three accuracy criteria were used to assess modelperformances.They included the coeffi cientof determination(R2),the volume difference(Dv),and the root mean square error(RMSE),as follows:

where Qiis the measured discharge on day i,is the computed discharge on day i,and Q is the average measured discharge of a given year or snowmelt season.

Table4 SRM parameters for runoff calibration in Urumqi River Watershed.

where vris the measured runoff volume andis the simulated runoff volume.

4.Results and discussion

4.1.Calculation of input variables

4.1.1.Temperature lapse rates

Temperature lapse rates are generally used in snowmelt runoff modeling to distribute point-based observations of daily average air temperature to the hypsometric mean elevation of each elevation zone.To avoid the need for estimating daily forecasted lapse rates,mean monthly lapse rates were used in this study.Mean monthly lapse rates were estimated by deriving daily average temperature data from the Yingxiongqiao Hydrological Station,Yuejinqiao Hydrological Station, and Daxigou Meteorological Station in the basin and plotting them according to their elevations.A regression line was then fi tto the data,and the daily lapse rate was equalto the slope of the line.The resulting lapse rates were then averaged monthly to obtain mean monthly values.The monthly lapse rate calculation results are shown in Table 5.

4.1.2.Precipitation

In the study watershed,the available daily precipitation data measured at three hydro-meteorological stations have a large degree of spatio-temporal irregularity,no significant correlation or any regular trend can be found,and the altitude gradient cannot be defi ned accordingly.Results show that,for all the three years,there is a relatively high correlation between the runoff at the Yingxiongqiao Hydrological Station and precipitation from the Daxigou Meteorological Station, with a three-year average value of correlation of 0.521,and the correlation between the runoff and precipitation at the Yingxiongqiao Hydrological Station and the Yuejinqiao Hydrological Station is rather small,with three-year average values of 0.227 and 0.226,respectively.Hence,the Daxigou Meteorological Station was selected as a reference for extrapolation of precipitation with an altitude gradient of 3%to the mean elevations of each fi ve zones.

However,the Daxigou Meteorological Station is located at the upper end of the watershed,so it cannot account for the rainfall over the lower and middle sections of the watershed. Taking this factor into account,the input precipitation data to the models were further adjusted by the inclusion of precipitation records from the Yingxiongqiao Hydrological Station located at the lower end of the watershed and the Yuejinqiao Hydrological Station located in the upper middle part of the watershed.Specifi cally,the precipitation data derived from the Daxigou Meteorological Station were used for all fi ve zonesfor the period from March to June.For the flood period,precipitation records from the three stations were used for the calculation of the input precipitation data for different zones. Daily precipitation records from the Yingxiongqiao Hydrological Station for zone A,the daily precipitation measured at the Yuejinqiao Hydrological Station for zones B and C,and the daily precipitation data from the Daxigou Meteorological Station for zones D and E were used as references for extrapolation with an altitude gradient of 3%.

Table5 Lapse rates for Urumqi River Watershed from 2005 to 2007.

4.1.3.Snow cover mapping

The MOD 10A2 data were used in this study for snow cover mapping.The transitory snowfall has been determined using the snow or rainfall information classified based on the precipitation data.The average cloud cover percentages over the 24 MOD 10A2 data in the three years were 0.36%,0.94%, and 0.20%,respectively,and the imagery with cloud cover appeared mainly in the beginning of the study period.Therefore,the snow cover area for each MOD 10A2 data was extracted fi rst,and the daily SCA data between the two successive periods of MODIS data were obtained with the linear interpolation method and the curve was smoothed using the B-spline interpolation.The snow depletion curves of different elevation zones for the study watershed in the years 2005, 2006,and 2007 are shown in Fig.2.The curves show that the snow in the Urumqi River Watershed was mainly distributed in the high-elevation zones D and E,and the snow in the lowerelevation zones melted away in the early meltseason.In zones A and B,the snow cover percentage reached zero at the end of the spring,and in zone C the snow vanished in the early summer.

4.2.Simulation results

The results of calibration(in 2005)and validation(in 2006 and 2007)of the models are shown in Table 6.It can be seen from the model performances that the overall performance of the SRM+AS model improved slightly as compared to the traditional SRM model.The SRM+AS model resulted in a lower volume difference(-1.17%),a smaller root mean square error(6.51 m3/s),and a higher coeffi cient of determination(0.76)as compared with the SRM model in the calibration year 2005,indicating better performance in terms of minimizing simulation errors.Similar performances were observed for the validation periods of 2006 and 2007,although the volume difference for 2006 increases through use of the new approach.The coeffi cients of determination increased from 0.69 and 0.79 with the SRMmodelto 0.76 and 0.81 with the SRM+AS model,and the root mean square errors decreased slightly from 4.74 and 6.55 with the SRM model to 4.23 and 6.22 with the SRM+AS model,in 2006 and 2007, respectively.

Fig.2.Snow depletion curves in upper Urumqi River Watershed in years 2005,2006,and 2007.

Table6 Simulation of daily discharge in Urumqi River Watershed.

The time series comparisons of measured and simulated daily runoff between the two models are illustrated in Figs.3 and 4.As shown in the fi gures,the snowmelt runoff starts increasing from the spring to summer(from April to August), mainly due to the increase in air temperature,whereas the maximum runoff occurred in July,mainly due to the maximum temperature and some fl ash fl oods caused by rainfall in the basin.The simulation confirms snowmelt runoff as the main source of freshwater in the study watershed throughout the spring and summer from April to August of the year.Both models captured the general tendency of the snowmelt runoff process very well,except for the peak volume thatoccurred in July in the summer when most of the runoff was contributed by rainfall that did not adequately refl ect the models'precipitation input due to the limited number of rain gauges in the whole basin.Hence,a proper method for data preparation of precipitation input is important to improving the models' simulation results in July,particularly for the rainfall in the middle and lower parts of the basin in the modeling process.

Fig.3.Simulated runoff with SRM model in years 2005,2006,and 2007.

Overall,the simulation results show that the proposed SRM+AS model tends to perform better in reducing simulation errors and increasing the simulation accuracy.The SRM+AS model that includes topographic factors such as aspect and slope in snowmelt calculation will have a positive effect in improving the accuracy of snowmelt runoff simulation in the mountain watershed with proper parameterization, and careful input data selection and preparation.Due to the limitation of the research to only one watershed,further applications of the SRM+AS model to other watersheds would be recommended to provide possibly better simulation results.

5.Conclusions

This study investigated the feasibility of snowmelt runoff simulation using the temperature-index snowmelt runoff model(SRM)with a fi ner zonation based on aspect and slope in a data-scarce mountain watershed in the Urumqi River Basin,in Northwest China.The traditional SRM modelwith a fi ner zonation based on aspect and slope(SRM+AS model) was developed.The proposed model was used to estimate the melt rate with DDF and division of watershed elevation zones in the study watershed.The simulation results of the SRM+AS modelwere compared with those of the traditional SRMmodelto identify the improvements of performance with inclusion of topographic features of the watershed.

Fig.4.Simulated runoff with SRM+AS modelin years 2005,2006, and 2007.

It can be concluded from this study that the performance of the SRM+AS model has improved slightly compared to the traditional SRM model.The proposed model includes topographic factors such as aspect and slope in snowmeltmodeling.Ittends to improve the accuracy of snowmelt runoff simulation in data-scarce mountain watersheds in arid regions with proper parameterization,and careful input data selection and preparation.The limited results from this study show that better results could be achieved for snowmelt runoff simulation in the mountain watersheds of Xinjiang,in China,with the inclusion of aspect and slope in snow melt rate estimation for snowmelt runoff modeling.

The runoff of the Urumqi River is mainly contributed by snow melt,ice melt,and precipitation.The contribution of precipitation is as signifi cantas the snow meltand ice melt.The snowmeltrunoffmodels developed in this study may notrefl ect the complex runoff process of this particular watershed accurately.In addition,the accuracy of the MOD 10A2 data in the Urumqi River Watershed or any ambient area has not been investigated yet,and the elimination ofsome ofthe MOD 10A2 data with transitory new snow prolonged the time intervaloftwo successive MODIS imageries for linear interpolation up to 24 days in this study,which may be one of the factors that contribute negatively to the model's efficiency.Moreover,water losses via evapotranspiration from vegetation and sublimation of snow were notconsidered in this study.In fact,the effectof sublimation is one of the dominant ablation processes in aridand semi-arid regions.Due to the limitation ofthe study to only one watershed,further applications of the SRM+AS modelto other watersheds would be recommended forfuture research to justify possibly better simulation results.

Abudu,S.,Cui,C.L.,Saydi,M.,King,J.P.,2012.Application of snowmelt runoff model(SRM)in mountainous watersheds:A review.Water Sci.Eng. 5(2),123-136.http://dx.doi.org/10.3882/j.issn.1674-2370.2012.02.001.

Boudhar,A.,Boulet,G.,Hanich,L.,Duchemin,B.,Berjamy,B., Chehbouni,A.,2009.Evaluation of the snowmelt runoff model in the Moroccan High Atlas Mountains using two snow-cover estimates.Hydrol. Sci.J.54(6),1094-1113.http://dx.doi.org/10.1623/hysj.54.6.1094.

Gong,L.,Zhang,H.F.,Xie,X.,2009.Studies on change of urbanization and its simulation and forecast in Urumqi during the past 20 years.J.Arid Land Resour.Environ.23(5),185-208.http://dx.doi.org/10.13448/ j.cnki.jalre.2009.05.032.

Hall,D.K.,Riggs,G.A.,Salomonson,V.V.,DiGirolamo,N.E.,Bayr,K.J., 2002.MODIS snow-cover products.Remote Sens.Environ.83(1-2), 181-194.http://dx.doi.org/10.1016/S0034-4257(02)00095-0.

Hock,R.,1999.A distributed temperature-index ice-and snowmelt model including potential direct solar radiation.J.Glaciol.45(149),101-111.

Hock,R.,2003.Temperature index melt modelling in mountain areas.J. Hydrol.282(1-4),104-115.http://dx.doi.org/10.1016/S0022-1694(03) 00257-9.

Huang,W.J.,Huang,Z.,Cui,C.X.,Feng,Z.M.,Li,Q.,Wang,L.,2007. Temporaland spatialdistribution of snow density and its characteristics in Xinjiang.J.Glaciol.Geocryol.29(1),66-72(in Chinese).

Kang,D.H.,2005.Distributed Snowmelt Modeling with GIS and CASC2D at California Gulch,Colorado.M.E.Dissertation.Colorado State University, Fort Collins.

Kang,K.,Merwade,V.,2011.Development and application of a storagerelease based distributed hydrologic model using GIS.J.Hydrol. 403(1-2),1-13.http://dx.doi.org/10.1016/j.jhydrol.2011.03.048.

Kang,K.,Lee,J.,2014.Hydrologic modelling of the effect of snowmelt and temperature on a mountainous watershed.J.Earth Syst.Sci.123(4), 705-713.http://dx.doi.org/10.1007/s12040-014-0423-2.

Karimi, H., Zeinivand, H., Tahmasebipour, N., Haghizadeh, A., Miryaghoubzade,M.,2016.Comparison of SRM and WetSpa models efficiency for snowmelt runoff simulation.Environ.Earth Sci.75(8),1-16. http://dx.doi.org/10.1007/s12665-016-5490-z.

Kondo,J.,Yamazaki,T.,1990.A prediction modelfor snowmelt,snow surface temperature and freezing depth using a heat balance method.J.Appl. Meteorol. 29, 375-384. http://dx.doi.org/10.1175/1520-0450(1990) 029＜0375:APMFSS＞2.0.CO;2.

Li,H.Y.,Wang,J.,2008.The snowmelt runoff model applied in the upper Heihe River Basin.J.Glaciol.Geocryol.30(5),769-775(in Chinese).

Li,X.G.,Williams,M.W.,2008.Snowmelt runoff modeling in an arid mountain watershed,Tarim Basin,China.Hydrol.Process.22(19), 3931-3940.http://dx.doi.org/10.1002/hyp.7098.

Ma,H.,Cheng,G.D.,2003.A test of snowmelt runoff model(SRM)for the Gongnaisi river basin in the Western Tianshan Mountains,China.Chin. Sci.Bull.48(20),2253-2259.http://dx.doi.org/10.1007/BF03182862.

Ma,Y.,Shi,Q.D.,Pan,X.L.,2004.Analysis on the dynamic trends of landscape pattern in the arid regions in West China.Arid.Land Geogr.27(3), 516-519(in Chinese).

Martinec,J.,Rango,A.,Roberts,R.,2008.Snowmelt Runoff Model(SRM) User's Manual.New Mexico State University,Las Cruces,pp.19-39.

Mehta,V.K.,Walter,M.T.,Brooks,E.S.,Steenhuis,T.S.,Walter,M.F., Johnson,M.,Boll,J.,Thongs,D.,2004.Application of SMR to modeling watersheds in the Catskill Mountains.Environ.Model.Assess.9(2), 77-89.http://dx.doi.org/10.1023/B:ENMO.0000032096.13649.92.

Nagler,T.,Rtto,H.,Malcher,P.,Mu¨ller,F.,2008.Assimilation of meteorological and remote sensing data for snowmelt runoff forecasting.Remote Sens.Environ.112(4),1408-1420.http://dx.doi.org/10.1016/j.rse.2007. 07.006.

Pellicciotti,F.,Brock,B.,Strasser,U.,Burlando,P.,Funk,M.,Corripo,J., 2005.An enhanced temperature-index glacier melt model including the shortwave radiation balance:Development and testing for Haut Glacier d'Arolla,Switzerland.J.Glaciol.51(175),573-587.http://dx.doi.org/ 10.3189/172756505781829124.

Rango,A.,Martinec,J.,1981.Accuracy of snowmelt runoff simulation. Hydrol.Res.12(4-5),265-274.

Richard,C.,Gratton,D.J.,2001.The importance of the air temperature variable for the snowmelt runoff modelling using the SRM.Hydrol.Process. 15(18),3357-3370.http://dx.doi.org/10.1002/hyp.1031.

Saydi,M.,Abudu,S.,Cui,C.,Basite,M.,2013.Hydrological characteristics oftypicalwatersheds in North Tianshan.J.China Hydrol.33(2),87-92(in Chinese).

Tang,Z.Y.,Fang,J.Y.,2006.Temperature variation along the northern and southern slopes of Mt.Taibai,China.Agric.For.Meteorol.139(3-4), 200-207.http://dx.doi.org/10.1016/j.agrformet.2006.07.001.

Tekeli,A.E.,Akyu¨rek,Z.,S?orman,A.A.,S?ensoyc,A.,S?orman,A.U¨.,2005. Using modis snow cover maps in modeling snowmeltrunoff process in the eastern part of Turkey.Remote Sens.Environ.97(2),216-230.http:// dx.doi.org/10.1016/j.rse.2005.03.013.

Vafakhah,M.,Nouri,A.,Alavipanah,S.K.,2015.Snowmelt-runoff estimation using radiation SRM model in Taleghan watershed.Environ.Earth Sci. 73(3),993-1003.http://dx.doi.org/10.1007/s12665-014-3449-5.

Wang,J.,Li,H.,Hao,X.,2010.Response of snowmelt runoff to climate change in an inland river basin,Northwestern China,over the past 50 years.Hydrol.Earth Syst.Sci.14(10),1979-1987.http://dx.doi.org/ 10.5194/hess-14-1979-2010.

Wu,L.H.,Li,H.L.,Wang,L.,2011.Application of a degree-day model for determination of mass balance of Urumqi Glacier No.1,Eastern Tianshan, China.J.Earth Sci.22(4),470-481.http://dx.doi.org/10.1007/s12583-011-0201-x.

Zhang,P.,Wang,J.,Liu,Y.,Li,Y.,2009.Application of SRMto flood forecast and forewarning of Manasi River Basin in spring.Remote Sens.Technol. Appl.24(4),456-461(in Chinese).

Zhang,Y.C.,Li,B.L.,Bao,A.M.,Zhou,C.H.,Chen,X.,Zhang,X.R.,2006. Simulation ofsnowmeltrunoffmodelin Kaidu River Basin.Sci.China Ser. D Earth Sci.36,24-32.http://dx.doi.org/10.1007/s11430-007-5007-4.

Received 18 March 2016;accepted 1 July 2016

Available online 12 November 2016

This work was supported by the National Natural Science Foundation of China(Grant No.51069017)and the International Collaborative Research Program of Xinjiang Science and Technology Commission(Grant No. 20126013).

*Corresponding author.

E-mail address:shalamu.abudu@ag.tamu.edu(Shalamu Abudu).

Peer review under responsibility of Hohai University.

?2016 Hohai University.Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license(http:// creativecommons.org/licenses/by-nc-nd/4.0/).