The value of satellite- derived snow cover images for calibrating a hydrological model in snow- dominated catchments in Central Asia

Including satellite-derived snow cover data for hydrologic model calibration can be a good way to improve model internal consistency. This study applied a multiobjective genetic algorithm to characterize the trade-off curve between model performance in terms of discharge and snow cover area (SCA). Using a Monte Carlo-based approach, we further investigated the additional information content of an increasing number of SCA scenes used in the calibration period. The study was performed in six snowmelt-dominated headwater catchments of the Karadarya Basin in Kyrgyzstan, Central Asia, using the hydrological model WASA and snow cover data from four melt seasons retrieved from AVHRR (Advanced Very High Resolution Radiometer). We generally found only small trade-offs between good simulations with respect to discharge and SCA, but good model performance with respect to discharge did not exclude low performance in terms of SCA. On average, the snow cover error in the validation period could be reduced by very few images in the calibration period. Increasing the number of images resulted in only small further improvements. However, using only a small number of images involves the risk that these particular images cause the selection of parameter sets which are not representative for the catchment. It is therefore advisable to use a larger number of images. In this study, it was necessary to include at least 10-16 images. Key Points Calibration results in small trade-offs between discharge and snow cover Snow cover errors may be large when neglecting snow cover during calibration Value of increasing the number of snow cover scenes for calibration is studied