Evaluation of the performance of a database for snow water equivalent in southern Quebec, obtained using a spatialised particle filter

Université de Sherbrooke (UdeS) produced a snow water equivalent (SWE) database for southern Quebec, Canada, using a spatialised particle filter, the snow module of the distributed hydrological model Hydrotel, and two types of snow observations (manual snow surveys, and automated sonic sensors). This gridded database has a spatial resolution of 10 km and covers the southern part of the province of Quebec, below 53°N. This database is used operationally as part of the government’s official flood forecasting system but has never been compared to other similar gridded datasets. This work therefore aims to compare the UdeS-produced SWE grids with reference data from the Canadian database CanSWE, and GMON stations, which measure SWE using gamma ray attenuation. This study also compares the UdeS grids with four other gridded databases that are widely used in hydrology: ERA-5 Land, SNODAS, MERRA and Crocus-ERA-5. Three indices are used to evaluate the ability of these databases to estimate the duration of the snow cover period: the start and end dates of snow cover and the start date of snowmelt. The annual maximum of SWE, the correlation coefficient, bias, and root mean square error (RMSE) are other indices used to evaluate the ability of these databases to estimate SWE values. The main results show that, despite some differences, particularly in the north-eastern part of the study area, all databases accurately estimated the duration of the snow cover period. Except for the MERRA database, which appears to underestimate the SWE in our study area, the results show that all databases perform well. ERA-5 Land appears to perform better, although it overestimates the reference data. UdeS and Crocus perform similarly but underestimate the reference data.