1. Introduction
Snow and glaciers are natural reserves of freshwater in the Himalayan region that benefit billions of people downstream. The perennial rivers originating from the Himalayas are sources of energy for hydropower plants and are essential for agriculture and maintaining biodiversity on a local and regional scale. It has been demonstrated that Himalayan water resources, often called water towers, are essential to or supportive for downstream regions (Viviroli et al. 2007; Immerzeel et al. 2010). Seasonal snow cover is an important component of the Himalayan environment as precipitation occurs in solid form in the cold climate and regions of high elevation. Snow, with inherent properties such as high albedo, low roughness, and low thermal conductivity, has considerable spatial and temporal variability, which greatly governs the energy and water interactions between the atmosphere and land surface. From a hydrological point of view, the temporal and spatial variability of the snow distribution on a basin scale plays a key role in determining the timing and magnitude of spring snowmelt runoff. Considering the effect of snow on land and atmospheric processes, it is essential that hydrological models accurately describe seasonal snow evolution (Liston 1999). To this end, many diverse approaches have been taken in developing hydrological models for better representation of snow processes, ranging from simplified temperature index models (e.g., Martinec et al. 1983; Bergström 1992) to physically based single- or multilayer energy-balance snowmelt models (e.g., Blöschl et al. 1991; Kustas and Rango 1994; Tarboton and Luce 1996; Marks et al. 1999; Zanotti et al. 2004; Garen and Marks 2005; Liston and Elder 2006; Letsinger and Olyphant 2007). An advantage of using a model based on the energy balance over a model based on a temperature index is that the former accurately simulates the complex melt–refreeze snow physics; however, such models require a larger input dataset and have significant computation costs. In addition, multilayer snow models have shown better performance in capturing the diurnal freeze and thaw cycles, which in turn allows for the accurate simulation of the timing and total amount of snowmelt runoff (Sun et al. 1999; Slater et al. 2001; Xue et al. 2003; Bowling et al. 2003).
Many previous studies have used satellite snow-cover information to simulate snowmelt runoff in the Himalayan region (Rango et al. 1977; Dey et al. 1983; Wang and Li 2001; Singh and Jain 2003; Konz et al. 2007; Immerzeel et al. 2009; Bookhagen and Burbank 2010). However, it would be beneficial to society to provide the spatial snow-cover information through physical modeling of the snowmelt, which would be a supportive tool for studies on water resource management and climate-change adaptation. This is because future snow-cover information can obviously only be modeled (or predicted) and not sensed remotely.
In a previous study (Shrestha et al. 2010b), the snow physics of the Water and Energy Budget–based Distributed Hydrological Model (WEB-DHM; Wang et al. 2009a,b) was significantly improved (WEB-DHM-S). WEB-DHM-S was developed by incorporating the three-layer snow scheme of the Simplified Simple Biosphere, version 3 (SSiB3; Sun and Xue 2001; Xue et al. 2003) and the prognostic albedo scheme of the Biosphere–Atmosphere Transfer Scheme (BATS; Dickinson et al. 1993; Yang et al. 1997) into the WEB-DHM hydrological model. WEB-DHM-S has been rigorously evaluated with comprehensive point measurements of snow (Shrestha et al. 2010b, 2011). The present study investigates the spatial distribution of the seasonal accumulation and melting of snow in the Dudhkoshi region of the eastern Nepal Himalayas using the new distributed system (WEB-DHM-S). The Dudhkoshi region has topography ranging from midaltitude mountains to the high Himalayas and contains the most extensive and rugged high-elevation areas on earth. The region is one of the most glaciated regions of the Nepal Himalayas. The extreme terrain, hostile climate, and poor accessibility are major impediments to snow measurements. Many studies have been conducted in the Nepal Himalayas regarding snow and ice melt modeling (e.g., Fukushima et al. 1991; Braun et al. 1993; Rana et al. 1997; Sharma et al. 2000; Chalise et al. 2003; Kayastha et al. 2005; Rees and Collins 2006; Konz et al. 2007; Shrestha et al. 2010a; Alford and Armstrong 2010) but there has been no previous study on the modeling of the spatial distribution of snow cover not only in the Dudhkoshi region but also in the entire Nepal Himalayas. The main objective of this paper is to provide the basin-scale snow-cover distribution in the Dudhkoshi region of the Nepal Himalayas, along with an assessment of the land/snow surface temperature.
2. Model description
The hydrological model used in this research is a WEB-DHM (Wang et al. 2009a,b) with improved snow physics (WEB-DHM-S; Shrestha et al. 2010b). The WEB-DHM realistically simulates the land surface and hydrological processes, providing a consistent description of water, energy, and CO2 fluxes at a basin scale (Wang et al. 2009a,b). It was developed by fully coupling the Simple Biosphere, version 2 (SiB2; Sellers et al. 1996) with a geomorphology-based hydrological model (Yang et al. 2002, 2004). WEB-DHM-S (Shrestha et al. 2010b) has been developed by coupling the three-layer energy-balance snow physics of the Simplified Simple Biosphere, version 3 (SSiB3) and the prognostic albedo scheme of the BATS into the WEB-DHM. WEB-DHM-S adds more features to the WEB-DHM for simulating the spatial distribution of snow variables such as the snow depth, snow water equivalent, snow density, liquid water and ice contents in each snow layer, snow albedo, snow surface temperature, and snowmelt runoff. Following the basic model structure (Wang et al. 2009a), the basin and subbasins are delineated employing the Pfafstetter scheme, and subbasins are divided into a number of flow intervals based on the time of concentration. All external parameters (e.g., land use, soil type, hillslope properties, and vegetation parameters) and a meteorological forcing dataset including precipitation are attributed to each model grid, in which water, energy, and CO2 fluxes are calculated. A hillslope-driven runoff scheme employing a kinematic wave flow routing method is adopted in calculating runoff. For snow-covered model grids, a three-layer energy-balance-based snow accumulation and melting algorithm is used when the simulated snow depth is greater than 5 cm; otherwise, a one-bulk-layer snow algorithm is used. Each model grid maintains its own prognostic snow properties (temperature, density, and ice/water content) and/or land surface temperature and soil moisture content. In this study, a grid with 1-km resolution is used where water and energy flux calculations are carried out in hourly time steps. The three-layer snow model coupled to the soil model is illustrated in Fig. 1. Major equations of snow processes are discussed in this section; details of the snow processes were given by Shrestha et al. (2010b).
The soil model coupled with a three-layer snow model in WEB-DHM-S.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
a. Energy-balance equations
b. Mass balance equations
3. Study area, data, and methods
The study area is the Dudhkoshi region of the Koshi basin, located in the northeast Nepal Himalayas (see Fig. 2). The topography of this region varies from midaltitude mountains to the high Himalayas, with elevation ranging from 452 to 8848 m (Mt. Everest) above mean sea level (MSL). It has complex physiographic variability with tropical forest in low-lying areas to semiarid and arctic environments in high-elevation regions. The catchment area is about 3700 km2. The annual precipitation averages about 1850 mm and has high altitudinal variability, from around 2500 mm at low elevation (below 3000 m) to around 600 mm at high elevation (above 4500 m). The climate has four seasons: winter (December–February), premonsoon (March–May), monsoon (June–September), and postmonsoon seasons (October–November). The premonsoon season is characterized by hot and dry weather with scattered rainfall. The summer monsoon season is hot and humid with overcast skies and intense precipitation (around 80% of the annual amount). The onset of monsoon is around mid-June, subject to interannual variation. The summer monsoon precipitation occurs in solid form at higher altitudes and the glaciers in this area are thus termed summer-accumulation glaciers. The postmonsoon season brings good weather with a substantial reduction in precipitation. The winter season is characterized by clear skies, low temperature with a wide diurnal range, and much less precipitation. The source of runoff in this region is monsoon rainfall and the melting of snow and ice.
There are 12 meteorological stations in or near the Dudhkoshi basin (see Fig. 2 and Table 1). Stations 1–6 and station 12 are operated by the Department of Hydrology and Meteorology, Nepal (DHM), and stations 7–11 belong to the Himalayan reference site of the Coordinated Enhanced Observing Period (CEOP; Koike 2004) Asian Monsoon Project (CAMP). The elevation of CAMP automatic weather stations (AWS) in this region ranges from 2660 to 5035 m MSL over a distance of about 40 km. Pheriche (4260 m) and Syangboche (3833 m) AWSs are located in a wide flat field, whereas Namche (3570 m) and Pyramid (5035 m) AWSs are located in small hilly locations on ridges and the Lukla (2660 m) AWS is on a river terrace of a V-shaped valley. Pyramid and Syangboche AWSs have provided continuous datasets since 1994. The Pyramid AWS was established by the Ev-K2-CNR project of the Epson Meteo Center, a research unit of the Italian National Research Council, in collaboration with the Nepal Academy of Science and Technology. The Syangboche AWS was established in 1994 within the framework of the Glaciological Expedition in Nepal.
Summary of the meteorological stations in and near the study area.
a. Model input
The datasets required to drive the model are discussed here. A grid with 1-km resolution is employed as the computation grid, which is aggregated from a digital elevation model with 90-m resolution obtained by the Shuttle Radar Topography Mission (SRTM). The hillslope parameters (length and slope) for each model grid are obtained by preprocessing data from the SRTM digital elevation model. Details of the subgrid parameterization scheme used to capture topographical characteristics were given by Wang et al. (2009b). Land-use data are prepared according to SiB2 type using a land-use map obtained from the U.S. Geological Survey (USGS) global land-cover dataset with 1-km resolution. The parameters used are the saturated hydraulic conductivity, saturated and residual water content, and van Genuchten’s constant. The soil types within the basin are cambisols, lithosols, and gleysols. The leaf area index and fraction of photosynthetically active radiation absorbed by the green vegetation canopy are obtained from satellite data. These are eight-day composite products (MOD15A2) of the Moderate Resolution Imaging Spectroradiometer (MODIS) on board the Terra satellite with 1-km spatial resolution. The model parameters are presented in Table 2. Visible albedo is optimized through trial and error method based on observed snow depth, whereas other parameters are taken from literature values, with sources as indicated in Table 2. Moreover, initial conditions for soil moisture and groundwater storage are obtained by running the model several times, until hydrological equilibrium was reached.
Model parameters used in this study.
Surface meteorological datasets (hourly values of downward shortwave radiation, U and V components of wind speed, specific humidity, air temperature, surface atmospheric pressure, and precipitation) were recorded at CAMP stations. CAMP datasets are obtained from the Earth Observing Laboratory of the National Center for Atmospheric Research. This study is limited to one snowmelt season (2 October 2002–25 June 2003) of the third Enhanced Observing Period (EOP-3). DHM stations (except station 12) provide precipitation datasets only with daily time steps. The dataset recorded by the Dingboche station (station 12 in Fig. 2), 5 km east to the Pheriche site, is not used in this study owing to the lack of data for the study period. Furthermore, we utilize the MODIS cloud product (MOD06_L2; collection 5) for cloud cover fraction, which is available at 5-km spatial resolution. All the MODIS product data are obtained using the Warehouse Inventory Search Tool (WIST) of the National Aeronautics and Space Administration (NASA). In all calculations, UTC+6 is used for simplicity; however, the actual local time is UTC+5.45.
1) Correction of precipitation for snowfall
The snowfall measurement in the high-elevation region is critical, but direct measurements of snowfall are unfortunately not available at all rain gauge stations in this region. Manual recording is employed for precipitation measurements with rain gauges operated by the DHM (rain gauges 1–6; Fig. 2). At the CAMP Himalayan reference observational stations, the total precipitation is measured with a tipping-bucket rain gauge and air temperature is used to classify the precipitation as rainfall or snowfall. However, the precipitation measurement itself has huge error in the snowfall season because the tipping bucket does not measure precipitation accurately in the case of snowfall or when the bucket is frozen as the rain gauges are not heated. Moreover, none of the gauges are shielded. The recorded data show that snowfall that accumulates in the gauge at night mostly melts the following morning owing to insolation. Hence, the precipitation recorded at the CAMP stations, where winter precipitation is dominated by snowfall, requires correction. The precipitation is corrected using available datasets of either all or any one of albedo (calculated from upward and downward shortwave radiation), snow depth, and surface air temperature. This approach was also employed by Lang and Barros (2004) and Ueno et al. (2008).
The snow depth, measured by a sonic ranger, is available only at the Pyramid AWS. This dataset is the key information for this study; however, it has measurement uncertainty of 1 cm. A time series of the observed hourly snow depth and daytime albedo derived at the Pyramid AWS are shown in Fig. 3. There is an obvious relatively high value of albedo during the snow-cover season. The albedo is around 0.2 and 0.8 for snow-free days and snow-cover days, respectively. Figures 4a,b present two examples to show the necessity of precipitation correction. The snow depth increases to 180 mm owing to nighttime snowfall on 31 December 2002 and there is an abrupt increase in albedo on 1 January 2003, but no precipitation is recorded by the rain gauge (see Fig. 4a). A similar increase in snow depth is observed but precipitation is observed at 13:00 and 14:00 local time on 29 January 2003, where the data on snow depth and albedo show that the snowfall occurred in the morning (see Fig. 4b). Such delayed recording of precipitation is probably due to the melting of snow by strong insolation. The precipitation at the Pyramid AWS is corrected using mainly the differential change in the observed snow depth per unit time. A 1-cm change in snow depth without change in albedo is neglected and the new snowfall density is assumed to have a constant value 100 kg m−3 in the reconstruction of precipitation. The precipitation recorded by the gauge and the reconstructed precipitation at the Pyramid AWS are 108 and 265 mm, respectively, during the study period. At the Syangboche AWS, delayed precipitation records are shifted on the basis of air temperature and derived albedo. The data for precipitation at the Lukla AWS are of poor quality during EOP-3. Diurnal variability of precipitation at Syangboche is used in adjusting the delayed precipitation at the Namche and Pheriche AWSs because of the lack of albedo measurements at these stations.
(top) Daytime hourly albedo and (bottom) hourly snow depth observed at the Pyramid AWS from 2 Oct 2002 to 25 Jun 2003.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
(a) Hourly observations of air temperature, precipitation, snow depth, and albedo at the Pyramid AWS on 31 Dec 2002 and 1 Jan 2003. (b) As in (a) but for 28 and 29 Jan 2003.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
2) Interpolation of input data on a basin scale
The air temperature is interpolated in each model grid on the basis of the hourly lapse rate calculated from the air temperature observed at the five CAMP stations using a detrended inverse distance weight interpolation method. The downward shortwave radiation, wind speed, and relative humidity are simply interpolated on the basis of hypsometry. The incoming longwave radiation is estimated from the interpolated air temperature and bulk atmospheric emissivity factor. Atmospheric emissivity factor is estimated from cloud fraction (MOD06_L2), interpolated air temperature, and relative humidity following Crawford and Duchon (1999). Precipitation data recorded at stations 1–11 are interpolated to each model grid according to the elevation, although it involves a certain level of uncertainty.
b. Model evaluation
Snow depth and upward shortwave and longwave radiation at the Pyramid AWS are used for point-scale evaluation, whereas the snow-cover product and nighttime land surface temperature (LST) product from the MODIS on board the Terra satellite are used for basin-scale evaluation. The MODIS snow-cover product has been widely used in multidisciplinary works; for instance, the product has been used as forcing data to drive a snowmelt runoff model (e.g., Li and Williams 2008; Immerzeel et al. 2009), used as a validation tool for model output (e.g., Shamir and Konstantine 2007; Brown et al. 2008), and used in data assimilation schemes (e.g., Andreadis and Lettenmaier 2006; Zaitchik and Rodell 2009; Tang and Lettenmaier 2010). The MODIS Terra snow-cover product (MOD10A2) is an eight-day composite snow-cover product with 500-m spatial resolution, derived from eight-day periods of the MOD10A1 product starting on the first day of each year. In the case of MOD10A2, a pixel is classified as cloud only when the pixel is cloud covered continuously on all eight days, the pixel is classified as having snow cover if snow cover is observed on any of the eight days, and the pixel is classified as not having snow cover if no snow is observed on any of the eight days. Hence, MOD10A2 represents the maximum extent of snow cover over eight days, which effectively provides a temporal filter of MOD10A1 data minimizing cloud coverage. In this study, the MODIS snow-cover product from 30 September 2002 is used to initialize the snow-cover area (SCA) in the simulation, considering that these MODIS snow grids are perennially snow covered (see Fig. 2). All grids with snow are assigned with a snow depth of 1 m.
The nighttime MODIS LSTV5 (MOD11A2) is the eight-day product at 1-km resolution. It is simply a composite generated by averaging eight days of daily LST product (MOD11A1). The MODIS LST products are validated over a range of representative conditions with well-defined product uncertainties (Wan 2008), and they have been satisfactorily used in a wide variety of scientific studies (Hall et al. 2008; Bookhagen and Burbank 2010; Neteler 2010; Corbari et al. 2010; Gutmann and Small 2010). Tile h25v06 covers the entire study area and both MODIS datasets were downloaded from the WIST of NASA. The MODIS Reprojection Tool (MRT) is used to reformat the downloaded tile data from the Earth Observing System (EOS) hierarchical data format to binary format using a spatial subset for the desired grid resolution of 1 km2 and reprojecting to the Universal Transverse Mercator projection system.
Contingency table used to compute evaluation scores.
4. Results and discussions
a. Point-scale evaluation at the Pyramid site
The vertical processes of the model physics are validated at the Pyramid station with a limited dataset using the observations of snow depth, upward shortwave radiation (USR), and upward longwave radiation (ULR) from 2 October 2002 to 25 June 2003. The comparison between the observed and simulated hourly snow depths is shown in Fig. 5 with bias error (bias) and root-mean-square error (RMSE). The results show that the model simulates the variability in snow depth well. Two phases of continuous snow cover are observed: the first from 31 December 2002 to 14 January 2003 and the second from 29 January to 25 April 2003. In the first phase, the snow depth is slightly underestimated by the model, whereas in the second phase, the model simulation agrees quite well with the observation, but there is strong bias at the end of the melting period. The simulated snow-cover days continue until 30 April 2003 because of the slow decrease in simulated albedo compared with the observed case. The albedo derived for 24 April 2003 is 0.5, but the simulated albedo is around 0.63, and the decay in albedo continues for six more days.
Comparison of the observed and simulated hourly snow depth at the Pyramid AWS from 2 Oct 2002 to 25 Jun 2003.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Simulation results for the diurnal cycle of USR and ULR are presented in Figs. 6 and 7. The bias and RMSE for USR are 13.42 and 64.53 W m−2, respectively. The USR is overestimated in the melting season owing to the simulation of high albedo. On snow-cover days, the ULR is governed by the snow surface temperature and the model thus provides a constant ULR of about 315.6 W m−2 for the melting season as the snow surface temperature is 0°C during this period. However, the observation values are higher by an average of 30 W m−2. This discrepancy may be attributable to the presence of impurities on the snow surface, with surface temperature above 0°C. From 24 to 30 April 2003, the bias is remarkably large owing to a longer simulation of snow-cover days by the model. The average bias and RMSE for the ULR during the study period are 0.04 and 24.25 W m−2, respectively. It should be mentioned that the USR and soil surface temperature at the Syangboche CAMP Himalayan reference site have already been validated using the WEB-DHM (Shrestha et al. 2010a).
Comparison of the observed and simulated hourly USR at the Pyramid site from 2 Oct 2002 to 25 Jun 2003.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Same as Fig. 6 but for observed and simulated hourly ULR.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
b. Basin-scale evaluation
The MODIS snow-cover product and simulated spatial distribution of snow cover are shown in Fig. 8. Model output is taken as the maximum snow extent over eight days, corresponding to the dates of the MODIS dataset. In general, the simulated SCA is comparable to the MODIS snow-cover product. The model is able to reproduce the seasonal evolution of snow cover; however, large discrepancies are observed on some days. This difference is probably the result of precipitation interpolation. The model greatly overestimates the snow coverage in the low-lying area of the high Himalayas on 1 January 2003. The rapid increase in SCA on 1 January is due to the nighttime precipitation event on 31 December 2002. All CAMP stations recorded the precipitation and, hence, all model grids with elevation above 2600 m MSL are considered to receive precipitation during interpolation. However, there may not have been precipitation in the valleys, since MODIS indicates no snow in these areas. It is difficult to obtain the spatial distribution of precipitation in the Himalayan region even with advanced interpolation techniques as the spatiotemporal distribution of the precipitation system is complex and dominated by local orographic convection at a scale of individual ridges and valleys (Barros et al. 2006). A similar result is obtained for 2 February 2003 following midseason ablation, as seen in the snow-depth simulation at the Pyramid AWS and snow-cover images for 17 January 2003. The results further show that melting at low elevations accelerates more quickly in the MODIS product than in the model output (22 March and 7 April 2003; Fig. 8). The MODIS snow-cover products after 10 June are highly corrupted with cloud cover and, hence, are not used in the comparison. The areal extents of SCA derived from MODIS data and the model simulation are presented in Fig. 9. The model is able to track the seasonal evolution of the SCA in this region. Both MODIS and the model predict a small SCA until the end of December (with snow cover mostly at high elevations), with there being an abrupt increase at the beginning of January and again at the beginning of February. It should be noted that the MODIS product from 27 December reflects the maximum snow extent from 27 December 2002 to 3 January 2003. The high SCA on 27 December 2002 in Fig. 9 is from snowfall on 1 January 2003. The trend for midseason ablation between January and February is found to be highly correlated with MODIS observations; however, there is a remarkable overestimation of absolute values of SCA on 1 January and 2 February 2003, as discussed above. Owing to increased insolation and air temperature, the depletion of SCA accelerates from the beginning of March, and the model follows the MODIS SCA except for a remarkable overestimation during May. We believe that this overestimation is caused by misclassification of glaciers by the MODIS. It is possible that MODIS maps debris-covered glaciers as bare land, whereas the model simulation treats these glacier grids as SCA. On average, the performance of the model is satisfactory with average positive bias of 5.2% and a coefficient of correlation of about 92%.
Comparison of the MODIS snow product (MOD10A2) and the simulated snow-cover area in the Dudhkoshi region.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Areal extent of snow cover derived from the MODIS snow product and model simulations from 2 Oct 2002 to 25 May 2003.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
To explore the seasonal variability in the accuracy and bias, a 2 × 2 contingency table analysis (a pixel-by-pixel comparison of the simulated SCA and MODIS SCA) is performed for each period of eight days and the evaluation indices (as described in section 3) are shown in Fig. 10. The POD is found to be less than 0.75 on 9 November 2002 and 9 and 25 January 2003, whereas the average POD is 0.90. This also reflects that in these three scenes, the MODIS snow pixels are underestimated by more than 25%. However, the FAR is remarkably high on 1 January 2003 (or 27 December 2002) and 2 February 2003 (see Fig. 10), showing the overestimation of snow pixels. The BRS calculation is very high (above 1.4) at the end of the melting season. The model has an average KSS of 0.78, showing its potential to predict equitably snow and no-snow events with about 89% accuracy. The KSS shows a strong seasonal correlation to POD except for the dates with high FAR. It can be concluded that the model is very likely (90% on average) to correctly predict snow cover if the MODIS snow-cover product indicates snow, but the model has a significant tendency (10% on average) to predict snow cover when the MODIS snow-cover product does not indicate snow cover (see Table 4). The overall pixel agreement accuracy (the PC) is about 90%, however.
Seasonal variability in evaluation indices, calculated from similarity-categorized pixels of MODIS and model simulation.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Average values of evaluation indices during the study period.
Figure 11 shows the hypsographic curve corresponding to the images in Fig. 8, where the cumulative SCA below a given elevation is presented for both the MODIS and model predictions. The model is able to produce the snow line well but large bias is observed on some days (1 January and 2 February 2003 in Fig. 11). The overestimation of SCA varies in different elevation zones as shown in Fig. 12. The bias for SCA is negligible (less than 0.2%) in the elevation zone below 2000 m and that above 6000 m in this region. The elevation zone of 4500–5500 m is very sensitive as this is the zone where there is a variation in the seasonal snow line. The cumulative bias is large in the winter season, whereas the overestimation of the SCA is high in the elevation zone of 4000–6000 m in the premonsoon season (melting season).
Cumulative SCA (%) below a given elevation in the Dudhkoshi region corresponding to the images in Fig. 8.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Difference in SCA (%) between the MODIS data and the model prediction for a given elevation band. SCA(MODIS−MODEL) for the elevation band below 2000 m and that above 6000 m is negligible (<0.2%).
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Figure 13 compares the spatial distribution of the eight-day MODIS LST compared with the simulated LST at night (around 2200 local solar time). Only nighttime MODIS LST is used for comparison with the model simulation since nighttime MODIS LST products are more accurate and stable (Wang et al. 2008). Figure 14 shows the relative error and the box plot for the seasonal variation in percentiles of absolute error between the MODIS data and model simulation shown by Fig. 13. The 50th percentile (median) of the absolute error values demonstrates that the error is stable in the postmonsoon season, but shows an increasing trend in the melting season. The relative error also shows an overestimation of LST in March–May. This increasing trend is highly influenced by the bias in the lower elevation (snow free) areas, which may be caused by the uncertainty in the interpolation of meteorological variables; however, this should be more fully investigated in the future. In general, model- and MODIS-derived LST agree during the study period, with a mean absolute bias = 2.42°C and mean relative bias = 0.77°C. Figure 15 shows the altitudinal variation of the errors, indicating that the absolute error is consistent (2°–3°C) in elevation bands up to 5500 m. The relative bias is between 1° and 2°C up to 5000 m, above which a negative bias is observed. Both the errors are high in grids with elevation greater than 5500 m. The error in LST might be attributed to the combined effects of uncertainty in interpolated air temperature and atmospheric emissivity, to the ignorance of glaciers and to the MODIS LST algorithm itself. The designation of land cover type and misclassification of snow pixels greatly affects surface emissivity, which in turn introduces significant variation in the MODIS LST algorithm. In addition, the atmospheric temperature, water vapor, and cloud mask directly affect the LST algorithm. Another reason for the large bias might be the deficiency in Monin–Obukhov similarity theory in calculating the turbulent fluxes under a highly stable condition. Turbulence is not completely absent and has intermittent bursts under such conditions, yet it is modeled as continuous and stationary in Monin–Obukhov similarity theory. This results in an increase of cold bias in nighttime snow surface temperature (e.g., Van de Wiel et al. 2002; Brown et al. 2006). Ignoring the effect of glacier dynamics may also increase LST bias. In this study, the initial SCA is treated as snow instead of debris-free and debris-covered glacier. The albedo of glaciers is low (about 0.4 for clean ice and 0.2 for debris-covered ice) and hence the surface receives more energy than a snow surface does; that is, the evolution of the surface temperature will be quite different from that of a snow surface. Furthermore, the surface emissivity will differ if the ice is covered by a thin or thick layer of debris or by fresh snow. Nevertheless, when glaciers are ignored, the simulation of SCA is less sensitive.
Comparison of nighttime MODIS LSTV5 (MOD11A2) with the simulated nighttime LST in the Dudhkoshi region.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Box plot of the absolute error in nighttime LST between the MODIS data and model simulation, along with mean relative error. The horizontal line within each box represents the 50th percentile, the top (bottom) line of the box the 75th (25th) percentile, and the end of the top (bottom) line with whiskers the 95th (5th) percentile.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Radial plot for mean absolute and relative error in nighttime LST between the MODIS and model simulation in different elevation bands.
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
c. Sensitivity runs
There are several sources of uncertainty in the model simulation. The input data present a significant source of uncertainty, since the precipitation recorded by 11 stations and meteorological forcing from five stations are interpolated to each grid of the study area. However, this is the best available dataset for the study region. In future work, we intend to synthesize the spatial distribution of precipitation at a finer scale with the use of satellite products [e.g., the Tropical Rainfall Measuring Mission (TRMM), Global Satellite Mapping of Precipitation (GsMAP), and Multifunctional Transport Satellite (MTSAT)] and a reanalysis dataset. A series of sensitivity runs is performed to observe the effects of the air temperature lapse rate, initial snow depth, and snow albedo. Additionally, the sensitivity studies of the initial depth of snow for a predefined SCA, and of the threshold temperature for snow versus rain, suggest that the results are much less sensitive to these parameters (data not shown).
Figure 16a shows the 24-h moving average of the lapse rate of air temperature. The gray band shows the range of the upper and lower limits of the lapse rate, which are calculated through a combination of hourly lapse rates from the five CAMP stations. A stable lapse rate is observed from mid-February, whereas the lapse rate is highly variable from mid-October to mid-February (postmonsoon and winter seasons). Several simulations are carried out for the prescribed range of the lapse rate. The results for the upper and lower limits of lapse rate only are presented here. Figure 16b presents the absolute error in LST, Figs. 16c–g show the evaluation indices for the SCA, and Fig. 16h presents the areal extent of SCA. Compared to the control run (Table 5), the absolute difference in LST between MODIS and the model simulation is very large for the lapse rate increase in the postmonsoon (increased by 35%) and winter seasons (increased by 22%). Afterward, no significant variation is simulated. The period with high error is characterized by a clear-sky condition with weak winds and low atmospheric emissivity because of which the land surface cools rapidly through strong longwave radiative cooling. The bias in SCA is increased by 1.48% in winter, and 4.65% and 4.92% in post- and premonsoon seasons, respectively. Compared with the case of lapse rate increase, the lapse rate decrease (increase in temperature) has lower values of POD and FAR. The BRS and PC show significant variations in the melting season where the KSS shows no significant changes. The bias in SCA does not vary in winter, but it increases by 2.82% in postmonsoon season and decreases by 2.23% in premonsoon season. The bias in LST improves by 2.96% in postmonsoon and 1.20% in winter, but remains unchanged in the premonsoon/melting season. It is concluded that the air temperature perturbations have little effect on the SCA in postmonsoon and winter seasons (accumulation phase; see Fig. 16h), but strongly affects the simulation of LST.
Effect of air temperature lapse rate showing 24-h moving average for air temperature lapse rate with upper and lower limit. Mean absolute error in (a) LST, (b) POD, (c) FAR, (d) KSS, (e) BRS, (f) PC, and (g) SCA (in fraction).
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
Details of sensitivity runs, showing percentage increase or decrease in mean absolute bias in LST and SCA, w.r.t. the control run [postmonsoon (POM), winter (WIN), and premonsoon (PRM)].
The effect of fresh snow albedo on basin-wide snowpack simulation (SCA and LST) is examined by changing the predefined control value (0.85) of VIS albedo to the arbitrary values 0.75, 0.80, 0.90, and 0.95. The results shown in Fig. 17 demonstrate that the FAR, BRS, and KSS are better when the albedo is reduced but the POD is poor in the melting season owing to rapid melting of snow. Moreover, the SCA bias is more sensitive to increase in albedo (+4.26% for albedo 0.95) than decrease in albedo (−0.56% for albedo 0.75) in the melting season. The albedo has less effect on the SCA during the accumulation season, as expected (see Table 5). Meanwhile, the absolute error in LST simulation is high when the albedo is larger than in the control simulation. The variation in the error is large in the postmonsoon (−2.63 to +17.50%) and winter seasons (−10.74 to +3.34%), following the similar seasonal sensitivity of the air temperature perturbations. From the analysis of LST error and evaluation indices, it may be concluded that the fresh snow albedo value assigned in the control simulation produced a kind of trade-off result among all the sensitivity runs. This shows that the albedo parameterization would be a promising way to improve simulation results. In addition, remotely sensed distributed albedo products (e.g., Molotch et al. 2004; Painter et al. 2009) could be used in future in modeling of the snow-cover dynamics of such complex terrains.
Effect of snow albedo: (a) POD, (b) FAR, (c) KSS, (d) BRS, and (e) PC, and mean absolute error in (f) LST and (g) SCA (in fraction).
Citation: Journal of Hydrometeorology 13, 1; 10.1175/JHM-D-10-05027.1
In this study, we used the MODIS cloud fraction dataset as a proxy of atmospheric emissivity for calculating downward longwave radiation. If, in contrast, we were to neglect the effect of spatial variability of remotely sensed cloud coverage in the calculation of downward longwave radiation, we would have an answer to the question of how much would be the bias in the SCA and LST. Thus, we also performed the simulation by assigning the same atmospheric emissivity factor (derived from the observed meteorological dataset at Pyramid site) to the entire study area, assuming the same cloud structure. With comparison to the control run, this sensitivity demonstrates that the average absolute bias in LST increases by 2.14°C during the study period, with seasonal variations of +2.57°C in postmonsoon, +3.84°C in winter, and +0.01°C in the melting season. Likewise, the bias in SCA increases by 0.24%, 2.13%, and 4.64% in premonsoon, winter, and postmonsoon seasons, respectively. It may be concluded that ignoring the effect of spatial variability in cloud coverage greatly increases bias in LST simulations during winter and postmonsoon seasons and in SCA simulations during melting season.
5. Conclusions
In this study, a three-layer energy-balance snow model coupled with a distributed biosphere hydrological model is applied to simulate the spatial distribution of snow cover in the Dudhkoshi region of the eastern Nepal Himalayas. Point-scale evaluation was carried out at the Pyramid AWS (5035 m) against the observed snow depth and upward short- and longwave radiation, whereas basin-scale evaluation was performed by comparing the simulated snow cover with the MODIS eight-day snow-cover product (MOD10A2) and MODIS eight-day nighttime LST product (MOD11A2). The results for Pyramid illustrate that the model is capable of simulating snow accumulation and melting processes well; however, there were biases at the end of the melting season. The basin-scale SCA and snow line were well captured by the model relative to the MODIS snow product, but there was a large discrepancy on 1 January and 2 February, probably owing to uncertainty in the interpolation of precipitation. The evaluation indices derived from qualitative pixel-to-pixel analysis between the model- and MODIS-derived SCA illustrate that the model is capable in predicting snow and no-snow events with average accuracy of 90%; however, it has a tendency to overestimate the SCA by 10%. Hypsography analyses of the cumulative snow-cover area versus elevation indicate that the SCA in the elevation zone from 4500–5500 m is remarkably overestimated (about 6%) during the premonsoon season. Simulated nighttime land surface temperatures (LST) are comparable to the MODIS land surface temperatures (MOD11A2), with mean absolute bias of 2.42°C and mean relative bias of 0.77°C, during the study period.
Sensitivity runs were performed by varying air temperature lapse rates calculated from five CAMP stations and fresh snow albedo. These runs indicate that the effect of air temperature lapse rate perturbations on LST bias is more sensitive to the increased lapse rate simulations and in postmonsoon (up to +35%) and winter seasons (up to +22%), but less sensitive in the melting season (up to +7.1%). The runs with snow albedo show a seasonal sensitivity similar to that of the air temperature perturbations. However, the LST bias for the control run increases in the melting season, as opposed to the postmonsoon/winter seasons. The reasons for this increase should be investigated further. The impact of lapse rate perturbations on SCA bias is less sensitive in winter (+0.1% to +1.48%) and considerably sensitive in premonsoon (−2.23% to +4.92%) and postmonsoon (+2.82 to +4.65%) seasons. The SCA bias is highly sensitive to albedo in the melting season (−0.56 to +4.26%). The evaluation indices show that no single albedo value gives the best result. We conclude that the fresh snow albedo value assigned in the control simulation produces a kind of trade-off result among all the sensitivity runs. In the future, therefore, there should be greater focus on albedo parameterization and the use of remotely sensed distributed albedo products to treat snow-cover dynamics in such complex terrains. In addition, it is found that ignoring the spatial variability of remotely sensed cloud coverage greatly increases bias in LST and SCA simulations. Thus, increased temporal and spatial monitoring of meteorological phenomena and snow variables is required to compensate for the input errors and to properly simulate the snow processes in high-elevation regions. Through these evaluations, WEB-DHM-S has demonstrated its capacity to address basin-scale snow processes in the Himalayan region. In future work, WEB-DHM-S will be further coupled with a frozen soil scheme (e.g., Wang et al. 2010) and a glacier melt model to investigate complete cold-region processes in highly elevated river basins.
Acknowledgments
This study was supported by the Ministry of Education, Culture, Sports, Science, and Technology of Japan. The authors express deep gratitude to the Department of Hydrology and Meteorology, Kathmandu, Nepal, for providing hydrometeorological data and the Earth Observing Laboratory of the National Center for Atmospheric Research, United States, for providing the CAMP/EOP-3 dataset. The authors are grateful to the three anonymous reviewers for their insightful comments on an earlier version of this manuscript.
REFERENCES
Alford, D., and Armstrong R. , 2010: The role of glaciers in stream flow from the Nepal Himalaya. Cryosphere Discuss., 4, 469–494, doi:10.5194/tcd-4-469-2010.
Anderson, E. A., 1976: A point energy and mass balance model of a snow cover. NOAA Tech. Rep. NWS 19, 150 pp.
Andreadis, K. M., and Lettenmaier D. P. , 2006: Assimilating remotely sensed snow observations into a macroscale hydrology model. Adv. Water Resour., 29, 872–886.
Barros, A. P., Chiao S. , Lang T. J. , Burbank D. , and Putkonen J. , 2006: From weather to climate—Seasonal and interannual variability of storms and implications for erosion processes in the Himalaya. Tectonics, Climate, and Landscape Evolution, S. D. Willett et al., Eds., Geological Society of America Special Papers, Vol. 398, Geological Society of America, 17–38.
Bergström, S., 1992: The HBV model—Its structure and applications. SMHI Rep. RH 4, 35 pp.
Blöschl, G., Kirnbauer R. , and Gutknecht D. , 1991: Distributed snowmelt simulations in an Alpine catchment: 1. Model evaluation on the basis of snow cover patterns. Water Resour. Res., 27, 3171–3179.
Bookhagen, B., and Burbank D. W. , 2010: Toward a complete Himalayan hydrological budget: Spatiotemporal distribution of snowmelt and rainfall and their impact on river discharge. J. Geophys. Res., 115, F03019, doi:10.1029/2009JF001426.
Bowling, L. C., and Coauthors, 2003: Simulation of high-latitude hydrological processes in the Torne–Kalix basin: PILPS Phase 2(e) 1: Experiment description and summary intercomparisons. Global Planet. Change, 38, 1–30.
Braun, L. N., Grabs W. , and Rana B. , 1993: Application of a conceptual precipitation-runoff model in the Langtang Khola basin, Nepal Himalaya. Proc. Int. Symp. on Snow and Glacier Hydrology, Kathmandu, Nepal, IAHS, 221–237.
Brown, L., Robin T. , and Ming W. K. , 2008: Using satellite imagery to validate snow distribution simulated by a hydrological model in large northern basins. Hydrol. Processes, 22, 2777–2787.
Brown, R., Bartlett P. , Mackay M. , and Verseghy D. , 2006: Evaluation of snow cover in CLASS for SnowMIP. Atmos.–Ocean, 44, 223–238.
Brun, E., Martin E. , Simon V. , Gendre C. , and Coleou C. , 1989: An energy and mass model of snow cover suitable for operational avalanche forecasting. J. Glaciol., 35, 333–341.
Chalise, S. R., Kansakar S. R. , Rees G. , Croker K. , and Zaidman M. , 2003: Management of water resources and low flow estimation for the Himalayan basins of Nepal. J. Hydrol., 282, 25–35.
Corbari, C., Sobrino J. A. , Mancini M. , and Hidalgo V. , 2010: Land surface temperature representativeness in a heterogeneous area through a distributed energy-water balance model and remote sensing data. Hydrol. Earth Syst. Sci., 14, 2141–2151, doi:10.5194/hess-14-2141-2010.
Crawford, T. M., and Duchon C. E. , 1999: An improved parameterization for estimating effective atmospheric emissivity for use in calculating daytime downwelling longwave radiation. J. Appl. Meteor., 38, 474–480.
Deardorff, J. W., 1978: Efficient prediction of ground surface temperature and moisture, with inclusion of a layer of vegetation. J. Geophys. Res., 83, 1889–1903.
Dey, B., Goswami D. C. , and Rango A. , 1983: Utilization of satellite snow cover observations for seasonal streamflow estimate in the western Himalayas. Nord. Hydrol., 14, 257–266.
Dickinson, R. E., Henderson-Sellers A. , and Kennedy P. J. , 1993: Biosphere–Atmosphere Transfer Scheme (BATS) Version 1e as coupled to the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-387+STR, 72 pp.
FAO, 2003: Digital soil map of the world and derived soil properties. Land and Water Digital Media Series, United Nations Food and Agriculture Organization, CD-ROM disc 1.
Fukushima, Y., Wantabe O. , and Higuchi K. , 1991: Estimation of stream flow change in global warming in a glacier covered high mountain area of Nepal Himalaya. Proc. Int. Symp. on Snow, Hydrology and Forests in High Alpine Areas, Vienna, Austria, IAHS, 181–188.
Garen, D. C., and Marks D. , 2005: Spatially distributed energy balance snowmelt modeling in a mountainous river basin: Estimation of meteorological inputs and verification of model results. J. Hydrol., 315, 126–153.
Gutmann, E. D., and Small E. E. , 2010: A method for the determination of the hydraulic properties of soil from MODIS surface temperature for use in land-surface models. Water Resour. Res., 46, W06520, doi:10.1029/2009WR008203.
Hall, D. K., Box J. E. , Casey K. A. , Hook S. J. , Shuman C. A. , and Steffen K. , 2008: Comparison of satellite-derived and in-situ observations of ice and snow surface temperatures over Greenland. Remote Sens. Environ., 112, 3739–3749.
Immerzeel, W. W., Droogers P. , de Jong S. M. , and Bierkens M. E. P. , 2009: Large-scale monitoring of snow cover and runoff simulation in Himalayan river basins using remote sensing. Remote Sens. Environ., 113, 40–49.
Immerzeel, W. W., van Beek L. P. H. , and Bierkens M. F. P. , 2010: Climate change will affect the Asian water towers. Science, 328, 1382–1385.
Jordan, R., 1991: A one-dimensional temperature model for a snow cover. U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory Special Rep. 91-16, 49 pp.
Kayastha, R. B., Ageta Y. , and Fujita K. , 2005: Use of positive degree day methods for calculating snow and ice melting and discharge in glacierized basins in the Langtang Valley, Central Nepal. Climate and Hydrology of Mountain Areas, C. de Jong, D. Collins, and R. Ranzi, Eds., Wiley, 7–14.
Koike, T., 2004: The Coordinated Enhanced Observing Period—An initial step for integrated global water cycle observation. WMO Bull., 53, 115–121.
Konz, M., Uhlenbrook S. , Braun L. , Shrestha A. , and Demuth S. , 2007: Implementation of a process based catchment model in a poorly gauged, highly glacierized Himalayan headwater. Hydrol. Earth Syst. Sci., 11, 1323–1339.
Kustas, W. P., and Rango A. , 1994: A simple energy budget algorithm for the snowmelt runoff model. Water Resour. Res., 30, 1515–1527.
Lang, T. J., and Barros A. P. , 2004: Winter storms in the central Himalayas. J. Meteor. Soc. Japan, 82, 829–844.
Letsinger, S. L., and Olyphant G. A. , 2007: Distributed energy-balance modeling of snow-cover evolution and melt in rugged terrain: Tobacco Root Mountains, Montana, USA. J. Hydrol., 336, 48–60.
Li, X. G., and Williams M. W. , 2008: Snowmelt runoff modeling in an arid mountain watershed, Tarim Basin, China. Hydrol. Processes, 22, 3931–3940.
Liston, G. E., 1999: Interrelationships among snow distribution, snowmelt, and snow cover depletion: Implications for atmospheric, hydrologic, and ecologic modeling. J. Appl. Meteor., 38, 1474–1487.
Liston, G. E., and Elder K. , 2006: A distributed snow-evolution modeling system (SnowModel). J. Hydrometeor., 7, 1259–1276.
Marks, D., Domingo J. , Susong D. , Link T. , and David G. , 1999: A spatially distributed energy balance snowmelt model for application in mountain basins. Hydrol. Processes, 13, 1935–1959.
Martinec, J., Rango A. , and Major E. , 1983: The Snowmelt-Runoff Model (SRM) user’s manual. NASA Goddard Space Flight Center Reference Publication 1100, 110 pp.
Molotch, N. P., Painter T. H. , Bales R. C. , and Dozier J. , 2004: Incorporating remotely-sensed snow albedo into a spatially-distributed snowmelt model. Geophys. Res. Lett., 31, L03501, doi:10.1029/2003GL019063.
Neteler, M., 2010: Estimating daily land surface temperatures in mountainous environments by reconstructed MODIS LST data. Remote Sens., 2, 333–351, doi:10.3390/rs1020333.
Painter, T. H., Rittger K. , McKenzie C. , Slaughter P. , Davis R. E. , and Dozier J. , 2009: Retrieval of subpixel snow covered area, grain size, and albedo from MODIS. Remote Sens. Environ., 113, 868–879.
Rana, B., Nakawo M. , Fukushima Y. , and Ageta Y. , 1997: Application of conceptual precipitation runoff model (HYCYMODEL) in a debris covered glacierized basin in Langtang Valley, Nepal Himalaya. Ann. Glaciol., 25, 347–352.
Rango, A., Salomonson V. V. , and Foster J. L. , 1977: Seasonal streamflow estimation in the Himalayan region employing meteorological satellite snow cover observations. Water Resour. Res., 13, 109–112.
Rees, H. G., and Collins D. N. , 2006: Regional differences in response of flow in glacier-fed Himalayan rivers to climatic warming. Hydrol. Processes, 20, 2157–2169.
Sellers, P. J., and Coauthors, 1996: A revised land surface parameterization (SiB2) for atmospheric GCMs. Part I: Model formulation. J. Climate, 9, 676–705.
Shamir, E., and Konstantine P. G. , 2007: Estimating snow depletion curves for American River basins using distributed snow modeling. J. Hydrol., 334, 162–173.
Sharma, K. P., Charles J. V. , and Berrien M. , 2000: Sensitivity of Himalayan hydrology to land use and climatic changes. Climatic Change, 47, 117–139.
Shrestha, M., Wang L. , and Koike T. , 2010a: Investigating the applicability of WEB-DHM to the Himalayan river basin of Nepal. Annu. J. Hydraul. Eng., 53, 54–60.
Shrestha, M., Wang L. , Koike T. , Xue Y. , and Hirabayashi Y. , 2010b: Improving the snow physics of WEB-DHM and its point evaluation at the SnowMIP sites. Hydrol. Earth Syst. Sci., 14, 2577–2594, doi:10.5194/hess-14-2577-2010.
Shrestha, M., Wang L. , and Koike T. , 2011: Simulation of interannual variability of snow cover at Valdai (Russia) using a distributed biosphere hydrological model with improved snow physics. Annu. J. Hydraul. Eng., 54, 73–78.
Singh, P., and Jain S. K. , 2003: Modelling of streamflow and its components for a large Himalayan basin with predominant snowmelt yields. Hydrol. Sci. J., 48, 257–276.
Slater, A. G., and Coauthors, 2001: The representation of snow in land surface schemes: Results from PILPS 2(d). J. Hydrometeor., 2, 7–25.
Sun, S., and Xue Y. , 2001: Implementing a new snow scheme in Simplified Simple Biosphere Model (SSiB). Adv. Atmos. Sci., 18, 335–354.
Sun, S., Jin J. M. , and Xue Y. , 1999: A simple snow–atmosphere–soil transfer model. J. Geophys. Res., 104 (D16), 19 587–19 597.
Tang, Q., and Lettenmaier D. P. , 2010: Use of satellite snow-cover data for streamflow prediction in the Feather River Basin, California. Int. J. Remote Sens., 31, 3745–3762.
Tarboton, D. G., and Luce C. H. , 1996: Utah Energy Balance Snow Accumulation and Melt Model (UEB): Computer model technical description and users guide. Utah Water Research Laboratory and USDA Forest Service Intermountain Research Station Rep., 63 pp.
Ueno, K., Toyotsu K. , Bertolani L. , and Tartari G. , 2008: Stepwise onset of monsoon weather observed in the Nepal Himalayas. Mon. Wea. Rev., 136, 2507–2522.
Van de Wiel, B. J. H., Ronda R. J. , Moene A. F. , de Bruin H. A. R. , and Holtslag A. A. M. , 2002: Intermittent turbulence and oscillations in the stable boundary layer over land. Part I: A bulk model. J. Atmos. Sci., 59, 942–958.
Verseghy, D. L., 1991: CLASS—A Canadian land surface scheme for GCMs. I. Soil model. Int. J. Climatol., 11, 111–133.
Viviroli, D., Dürr H. H. , Messerli B. , Meybeck M. , and Weingartner R. , 2007: Mountains of the world, water towers for humanity: Typology, mapping, and global significance. Water Resour. Res., 43, W07447, doi:10.1029/2006WR005653.
Wan, Z., 2008: New refinements and validation of the MODIS land-surface temperature/emissivity products. Remote Sens. Environ., 112, 59–74.
Wang, J., and Li W. , 2001: Establishing snowmelt runoff simulating model using remote sensing data and GIS in the west of China. Int. J. Remote Sens., 22, 3267–3274.
Wang, L., Koike T. , Yang K. , Jackson T. J. , Bindlish R. , and Yang D. , 2009a: Development of a distributed biosphere hydrological model and its evaluation with the Southern Great Plains Experiments (SGP97 and SGP99). J. Geophys. Res., 114, D08107, doi:10.1029/2008JD010800.
Wang, L., Koike T. , Yang K. , and Yeh P. , 2009b: Assessment of a distributed biosphere hydrological model against streamflow and MODIS land surface temperature in the upper Tone River Basin. J. Hydrol., 377, 21–34.
Wang, L., Koike T. , Yang K. , Jin R. , and Li H. , 2010: Frozen soil parameterization in a distributed biosphere hydrological model. Hydrol. Earth Syst. Sci., 14, 557–571, doi:10.5194/hess-14-557-2010.
Wang, W., Liang S. , and Meyers T. , 2008: Validating MODIS land surface temperature products using long-term nighttime ground measurements. Remote Sens. Environ., 112, 623–635.
Woodcock, F., 1976: The evaluation of yes/no forecasts for scientific and administrative purposes. Mon. Wea. Rev., 104, 1209–1214.
Xue, Y., Sun S. , Kahan D. , and Jiao Y. , 2003: The impact of parameterizations in snow physics and interface processes on the simulation of snow cover and runoff at several cold region sites. J. Geophys. Res., 108, 8859, doi:10.1029/2002JD003174.
Yang, D., Herath S. , and Musiake K. , 2002: A hillslope-based hydrological model using catchment area and width functions. Hydrol. Sci. J., 47, 49–65.
Yang, D., Koike T. , and Tanizawa H. , 2004: Application of a distributed hydrological model and weather radar observations for flood management in the upper Tone River of Japan. Hydrol. Processes, 18, 3119–3132.
Yang, Z.-L., Dickinson R. E. , Robock A. , and Ya Vinnikov K. , 1997: Validation of the snow submodel of the Biosphere–Atmosphere Transfer Scheme with Russian snow cover and meteorological observational data. J. Climate, 10, 353–373.
Zaitchik, B. F., and Rodell M. , 2009: Forward-looking assimilation of MODIS-derived snow-covered area into a land surface model. J. Hydrometeor., 10, 130–148.
Zanotti, F., Endrizzi S. , Bertoldi G. , and Rigon R. , 2004: The GEOTOP snow module. Hydrol. Processes, 18, 3667–3679.
