Snow contribution to springtime atmospheric predictability over the second half of the twentieth century

Abstract

A set of global atmospheric simulations has been performed with the ARPEGE-Climat model in order to quantify the contribution of realistic snow conditions to seasonal atmospheric predictability in addition to that of a perfect sea surface temperature (SST) forcing. The focus is on the springtime boreal hemisphere where the combination of a significant snow cover variability and an increasing solar radiation favour the potential snow influence on the surface energy budget. The study covers the whole 1950–2000 period through the use of an original snow mass reanalysis based on an off-line land surface model and possibly constrained by satellite snow cover observations. Two ensembles of 10-member AMIP-type experiments have been first performed with relaxed versus free snow boundary conditions. The nudging towards the monthly snow mass reanalysis significantly improves both potential and actual predictability of springtime surface air temperature over Central Europe and North America. Yet, the impact is confined to the lower troposphere and there is no clear improvement in the predictability of the large-scale atmospheric circulation. Further constraining the prescribed snow boundary conditions with satellite observations does not change much the results. Finally, using the snow reanalysis only for initializing the model on March 1st also leads to a positive impact on predicted low-level temperatures but with a weaker amplitude and persistence. A conditional skill approach as well as some selected case studies provide some guidelines for interpreting these results and suggest that an underestimated snow cover variability and a misrepresentation of ENSO teleconnections may hamper the benefit of an improved snow initialization in the ARPEGE-Climat model.