Impact of Snow Representation in Seasonal Forecast Systems

Current seasonal forecast systems have difficulties predicting temperature over continental regions, whereas for some regions with maritime influence their performance is better. The main driver for better skill in maritime regions is related to the ocean and its memory effect. For continental regions, the land surface can become a more important source of predictability on (sub-)seasonal time scales. Besides soil moisture, snow is a crucial component of the land surface as it stores an extensive amount of water and modulates the earth’s radiation budget each winter season. A snow-covered land surface leads to local temperature decreases in the overlying air (snow-albedo effect and high emissivity) and melting snow cools the surface air and contributes to soil moisture and river water. We compare the snow representation in seasonal forecast systems from four European weather/climate services provided by the Copernicus Climate Change Service (C3S) and their performance in predicting snow, temperature and precipitation. The goal is to identify the impact of the snow initialisation and snow modelling from the four forecasts systems. The first results show that the predicted anomalies of 2m temperature over continental regions correlate with reanalyses only for the first forecasted month, whereas anomalies in snow water equivalent can be predicted up to several months. While the biases among the forecast systems differ, the correlation skills are similar for the same variable, with precipitation having the lowest correlation skills. Furthermore, we will investigate the causal relationships between snow and 2m temperature with time-lagged correlation or similar methods and will consider the whole ensembles of the hindcasts.