This study evaluates MODIS snow cover characteristics for large number of snowmelt runoff events in 145 catchments from 9 countries in Europe. The analysis is based on open discharge daily time series from the Global Runoff Data Center database and daily MODIS snow cover data. Runoff events are identified by a base flow separation approach. The MODIS snow cover characteristics are derived from Terra 500 m observations (MOD10A1 dataset, V005) in the period 2000–2015 and include snow cover area, cloud coverage, regional snowline elevation (RSLE) and its changes during the snowmelt runoff events. The snowmelt events are identified by using estimated RSLE changes during a runoff event. The results indicate that in the majority of catchments there are between 3 and 6 snowmelt runoff events per year. The mean duration between the start and peak of snowmelt runoff events is about 3 days and the proportion of snowmelt events in all runoff events tends to increase with the maximum elevation of catchments. Clouds limit the estimation of snow cover area and RSLE, particularly for dates of runoff peaks. In most of the catchments, the median of cloud coverage during runoff peaks is larger than 80%. The mean minimum RSLE, which represents the conditions at the beginning of snowmelt events, is situated approximately at the mean catchment elevation. It means that snowmelt events do not start only during maximum snow cover conditions, but also after this maximum. The mean RSLE during snowmelt peaks is on average 170 m lower than at the start of the snowmelt events, but there is a large regional variability.
Conceptual degree-day snow models are often calibrated using runoff observations. This makes the snow models dependent on the rainfall-runoff model they are coupled with. Numerous studies have shown that using Snow Cover Area (SCA) remote sensing observation from MODIS satellites helps to better constrain parameters. The objective of this study was to calibrate the CemaNeige degree-day snow model with SCA and runoff observations. In order to calibrate the snow model with SCA observations, the original CemaNeige SCA formulation was revisited to take into account the hysteresis that exists between SCA and the snow water equivalent (SWE) during the accumulation and melt phases. Several parametrizations of the hysteresis between SWE and SCA were taken from land surface model literature. We showed that they improve the performances of SCA simulation without degrading the river runoff simulation. With this improvement, a new calibration method of the snow model was developed using jointly SCA and runoff observations. Further analysis showed that the CemaNeige calibrated parameter sets are more robust for simulating independent periods than parameter sets obtained from discharge calibration only. Calibrating the snow model using only SCA data gave mixed results, with similar performances as using median parameters from all watersheds calibration.