Assessment of Snowcovered Areas Using Air Temperatures During Melt in a Mountainous Basin

Water Freedom System

Survive Global Water Shortages

Get Instant Access

PRATAP SINGH1 AND LARS BENGTSSON2 1 National Institute of Hydrology, Roorkee 247 667 (India), 2 Department of Water Resources Engineering, Lund University, S-221 00, Sweden


Snow is a very important component of the hydrological cycle and it plays a vital role in the water resources in many parts of the world. In a region or basin, snow cover is developed from a series of winter storms and is depleted during spring and summer period because of warmer climate. Depending upon the location of the basin and climatic conditions there, snow cover developed during preceding winter can deplete totally or partially. Snow cover is a major component of water storage on seasonal timescale, and changes in its extent, depth, and its water equivalent have a major impact on alpine and continental water resources. The amount and rate of runoff induced by melt processes in a basin can be related to variation in snow covered area (SCA). There are various hydrological applications of SCA on the basin scale. Modeling of snow melt runoff carried out using SCA data is very important for various practical applications in the field of runoff predictions, reservoir management, electric power production, irrigation practices, flood control, and so on. There are some hydrological models, for example, snow melt runoff model (SRM), that use SCA as an input variable on the daily basis for snow melt computation (Martinec et al. 1983). At the same time, variation in SCA plays a crucial role in modeling and simulating alterations of global change effects on water resources, the ecological conditions, the albedo, and ultimately on the radiation budget.

Application of SCA for snow melt modeling studies becomes inevitable for the large and inaccessible Himalayan basins, which experience high snowfall but do not have sufficient meteorological network to quantify it. Early use of remote sensing focused on empirical relationships between SCA and monthly or accumulated runoff (Rango etal. 1977; Ramamoorthi 1987). These simple relationships worked well for some applications and particularly in data-sparse regions of the world. Estimation of daily snow melt runoff in the Himalayan river basins using satellite-derived SCA is being increasingly recognized as an immensely useful procedure in water resources research and management (Rango et al. 1977; Ramamoorthi and Subba Rao 1981; Gupta etal. 1982; Dey etal. 1983, 1989; Dey and Goswami 1984; Jain 2001). The prediction of snow melt induced runoff in the Himalayan rivers has a great potential for application in irrigation, hydropower generation, and domestic and industrial water supply.

The SCA has also been used as an indicator of snow reserve or water equivalent in a basin (Meier 1973; 0degaard and 0strem 1977; Rango et al. 1977). Schjodt-Osmo and Engeset (1997) reported that the Norwegian Water Resources and Energy Administration (NVE) uses

Climate and Hydrology in Mountain Areas. Edited by C. de Jong, D. Collins and R. Ranzi © 2005 John Wiley & Sons, Ltd information on the changing distribution of snow during the periods of accumulation and, in particular, ablation (0strem 1974; Andersen 1983). The areal extent of snow is one of the principal variables and is directly related to the summer runoff potential. The snow water equivalent of a snowpack cannot be derived from only the current remote sensing data. However, the derivable snow cover fraction is still a very important parameter to monitor for operational flood forecasting (Schjodt-Osmo and Engeset 1997). Singh etal. (1997) and Singh and Jain (2002) used SCA in estimating snow and glacier contributions in the annual flows of Himalayan rivers using the water balance approach. Singh and Singh (2001) have discussed various applications of regional snow cover on the climate system.

Usually, information on the extent of SCA is derived from satellite data because snow can readily be identified and mapped within the visible bands of satellite imagery because of its high reflectance in comparison to nonsnow areas. Therefore, remote sensing is a valuable tool for obtaining snow data for predicting snowmelt runoff as well as climate studies. Use of satellite data for snow mapping has become operational in several regions of the world. Currently, NOAA develops snow cover maps for about 3000 river basins in North America, of which approximately 300 are mapped according to elevation for use in streamflow forecasting (Carroll 1990). NOAA also produces regional and global maps of mean monthly snow cover (Dewey and Heim 1981). Rango (1993) presented a review of the status of remote sensing in snow hydrology. Snow cover mapping with satellite data in the Swiss Alps is reviewed by Seidel etal. (1989, 1995). Haefner etal. (1997) suggested for setting up snow cover information systems for individual basins or other hydrological units, planning regions or even entire mountain ranges on a long-term perspective. On the practical side, these applications are related to the monitoring of seasonal and yearly alterations of the snow cover under presently existing climatic conditions to simulate and forecast runoff, to map the regional distribution of the water equivalent, and to document the recession processes of the snow cover during the melting period.

In order to get information on SCA, systematic and continuous mapping of snow cover becomes essential for snow hydrological applications. However, in general, there are discontinuities in the SCA data required for the studies. It is difficult to develop SCA database on the daily time scale because of the cost involved in acquiring satellite data, time consumed in the analysis of data, and inaccurate data due to presence of cloud cover. However, in some cases, depending upon the size of the study area, cost factor may not be important, but other factors dominate. In the melt season, the cloud cover represents a major obstacle when deriving information from optical imagery (Schjodt-Osmo and Engeset 1997). The issue related with filtering of cloud cover in the mountainous basins has been discussed in detail by Ranzi et al. (1999). They used NOAA-AVHRR data for monitoring areal snowpack in the Southern Alps. Haefner et al. (1997) reported that even the acquisition of 5-10 (or more) satellite scenes for a single melting season is rather costly. Even if one can afford the resources, a lot of time is consumed in the analysis. Such problems become more important for the Himalayan basins, which are larger in size and have longer ablation period with higher probability of cloud cover during premonsoon and monsoon period. A higher number of images is needed for such basins for snow melt modeling studies, which need lot of investment and also require much time for analysis.

For modeling of snow melt runoff and river discharge, the spatially distributed energy balance approach is considered as preferred method. This approach allows for computation of spatially distributed melt from the basin. Moreover, such methods also allow for prediction of the spatial distribution of SCA (Melloh et al. 1997; Bloschl et al. 1991; Bloschl and Kirnbauer 1992; Mittaz etal. 2002). For computing the melt from the basin using temperature index method, usually, satellite data are obtained for a few dates in the melt season and linearly interpolated for the period of unavailability of data. This study deals with developing a methodology to interpolate, extrapolate, or fill the missing SCA data during ablation period using temperature data, which is easily available. Following this approach, one can reduce the number of images for the melt period to obtaining SCA. Air temperature can then be used to generate SCA data for the basin.


The SCA at a particular time after first melt can be considered a function of the initial value of SCA before start of the melting and patterns of the temperature during the melt period. The use of degree-day or temperature index approach is a well-established method for snow melt estimation. At present, there are a number of snow melt models that use this approach for computing snow melt runoff from the basin (Singh and Singh 2001). Melt starts first in the warmer lower parts of a basin, where usually the snow cover is thinnest. Consequently, the snow disappears first from the lower part of the basin. As summer season progresses, the melt continues in the upper part of the basin. The SCA reduces with time and at each point of time melt can be related to air temperatures. Therefore, cumulative degree-days (CDD) over the melt period should represent the depletion of SCA. CDD is obtained simply by adding the daily mean temperature at the selected station. A seasonal snow cover will disappear at a faster rate during warmer climatic conditions, while it will follow slow depletion under a colder temperature regime. Rango and Martinec (1994) correlated SCA and cumulative depth of melt to infer the changes in SCA under warmer climatic scenarios, which indirectly supports the dependency of SCA on temperature because cumulative melt is mainly governed by the temperature. In this paper, to keep the methodology simple and directly applicable, temperature data of the stations located in or near the SCA are used. For the Himalayan basins melt season sets in about March, therefore March 1 has been considered as reference date for CDD computation and accordingly SCA data is used. However, the choice of reference date for initializing CDD may vary from region to region and accordingly the shape of the graph may vary. Broadly, the reference date should represent the time of when melting starts in the basin, that is, there is no further increase in SCA.


This study has been carried out for the Satluj river basin up to Bhakra dam (Indian part) located in the

Table 5.1 Main physical characteristics of the investigated basin



Name of study area

Satluj Basin

Mountain range

Western Himalayas

Elevation range (m.a.s.l.)





76° 10'-79°10'E

Area (km2)


Glaciers and permanent snow (%)


Mean annual rainfall (mm)


western Himalayas. The Satluj river is a highly snowfed river having about 60% contribution of snow and ice melt runoff in its annual flows (Singh and Jain 2002). The Satluj river rises in the lakes of Mansarovar and Rakastal in the Tibetan plateau at an elevation of about 4600 m and forms one of the main tributaries of the Indus river. The physiographical features of the study basin are given in Table 5.1, and location of the basin is shown in Figure 5.1. The altitude of the basin varies from about 500 m to 7000 m, although only a very small area exists above 6000 m. Mean elevation of the basin is about 3600 m. The shape and location of this basin is such that the major part of the basin area lies in the greater Himalayas where heavy snowfall is experienced during winters. A major part of the basin is covered by snow during winter. Owing to large differences in seasonal temperatures and great range of elevation in the

Mark The Satluj River Basin Map

Figure 5.1 Location map of the Satluj River basin (Indian part)

catchment, the snowline is highly variable, descending to an elevation of about 2000 m during winter and retreating to above 4000 m after summer season. The topographical setting and availability of abundant water provide a huge hydropower generation potential in this river, and hence several hydropower schemes already exist, are planned, or are coming up on this river.


The study has been carried out for the melt season (March-August) using five years SCA and temperature data. In general, SCA was available once a month, while daily mean temperatures were available for the whole study period. The daily mean temperatures of two high altitude stations, namely, Kalpa (2436 m) and Kaza (3639 m), were used in this study. March 1 was considered as reference starting point to compute the cumulative degree-days. The information on SCA was determined from the satellite images/digital data. In this study, the satellite data was processed using ERDAS IMAGINE image processing software. First, snow cover area maps were prepared for the study basin and then SCA was determined. Landsat (MSS) (80 m resolution)

data have been used for 1987, whereas IRS (LISS-I) (72.5 m resolution) data were used for 1988-91.


The depletion of SCA in Satluj basin with time during summer period and trends of CDD at Kalpa for three ablation seasons (1987, 1988, and 1989) are shown in Figure 5.2. Figure 5.3 shows the relationship between SCA with CDD for Kalpa for different years. It can be noticed from Figure 5.3 that SCA reduces exponentially with CDD for the summer season. Similar trends for all the years confirm such relationship. This relationship can be expressed as

The derived values of coefficients a and b and coefficient of determination (R2) for different years are given in the Table 5.2. Changes in the values of coefficients are possible because of changes in initial value of SCA and temperature conditions.

The high value of R2 for all years shows that the SCA and CDD in the form of Equation (5.1) are highly correlated. Results show that SCA and CDD are

Figure 5.2 Depletion of snow covered area (SCA) with time in the Satluj basin along with trend of daily cumulative degree-days (CDD) observed at Kalpa (2536 m) for different years. CDD was computed March 1 onward

Julian days

Figure 5.2 Depletion of snow covered area (SCA) with time in the Satluj basin along with trend of daily cumulative degree-days (CDD) observed at Kalpa (2536 m) for different years. CDD was computed March 1 onward

Figure 5.3 Relationship between snow cover depletion and cumulative degree-days (CDD) at Kalpa (2436 m) station

Figure 5.3 Relationship between snow cover depletion and cumulative degree-days (CDD) at Kalpa (2436 m) station

Table 5.2 Values of coefficients a and b used in Equation (5.1) and R2 for different ablation seasons using Kalpa station (2436 m) data


Values of

Values of

Coeff. of


coeff. a as in

coeff. b as in


Equation (5.1)

Equation (5.1)














not linearly related but rather are related nonlinearly. Exponential relationship implies that initial increment in temperature leads to higher changes in the snow cover area than later increments in temperature of the same magnitude. Such trends can be explained on the basis of distribution of snow in the basin. Consequently, snowpack developed in the basin during winter season is thin at lower altitudes and thick at higher altitudes. During summer, the snow line retreats from the lower altitude to higher altitude and consequently SCA in the basin is reduced. The retreat rate is reduced in the later part of the melt season because of higher depth of snow at high altitudes. Kattlemann (1997) reported rapid melting of snow at low elevations in Sierra Nevada, California, USA. Therefore, accumulation and depletion of snow and temperature conditions attribute to exponential trend of depletion of SCA with CDD. Using SCA and computed snow melt runoff for four years data, Gupta et al. (1982) reported a logarithmic relationship between SCA in the end of melt season and the volume of seasonal snow melt runoff for four Himalayan basins. Assuming that snow melt is linearly related with temperature, one can conclude that SCA and CDD should have exponential relationship.

On the basis of the conclusion that a logarithmic relationship exists between SCA in the end of melt season and the volume of seasonal snow melt runoff, as reported by Gupta et al. (1982), it can be expressed mathematically as ln(SCA) ~ V (5.2)

Assuming that snow melt is linearly related with temperature, one can conclude that SCA and CDD should have exponential relationship. Further, the preceding statement can be expressed mathematically as

where T0 is the reference temperature (0°C). Integrating both sides leads to

Figure 5.4 Depletion of snow covered area (SCA) with time in the Satluj basin along with trend of daily cumulative degree-days (CDD) observed at Kaza (3639 m) for different years. CDD was computed March 1 onward

Julian days

Figure 5.4 Depletion of snow covered area (SCA) with time in the Satluj basin along with trend of daily cumulative degree-days (CDD) observed at Kaza (3639 m) for different years. CDD was computed March 1 onward

The right side of the equation is a possible definition of CDD, so

When combined with the relationship of SCA to V leads to

Depletion of SCA for the study basin was also correlated with CDD of another station (Kaza, 3639m), which is located higher up in the basin (Figure 5.4 and 5.5). The relationship between SCA and CDD was also exponential for this station, but it was disturbed because of negative temperatures in the month of March at this station. As shown in Figure 5.5, in the beginning of melt season cumulative temperatures were negative at this station for all the three years. Cumulative negative temperature disturbed the exponential relationship for this initial period of melt season (Figure 5.5). Therefore, the stations that experience negative temperature during the melt season cannot be used for such applications. However, they can be used for a period after which they experience positive temperature and cumulative temperature is positive. For further application of this study, only Kalpa station was used. Additional snowfall during ablation season will change the relationship between SCA and CDD. Hall and Martinec (1985) have discussed this issue. This aspect has not been dealt in this study. However, there is need to involve these aspects in future.


There are three major applications of this approach, which are described below.

(a) Interpolation of SCA. Once the relationship between SCA and CDD is established using daily values of CDD and few values of SCA, this equation can be used to interpolate data during the melt period. Using CDD data in the derived equation, one can get daily values of SCA in the basin. The missing data can be generated using known relationship between SCA and CDD.

(b) Simulation of SCA. Because SCA and CDD are exponentially correlated, once the trend of depletion of SCA snow is established in the basin, it can be extended for the later part of the melt season using only CDD.

Was this article helpful?

0 0

Post a comment