The atmospheric component of the Mediterranean Sea water budget in a WRF multi-physics ensemble and observations

The use of high resolution atmosphere-ocean coupled regional climate models to study possible future climate changes in the Mediterranean Sea requires an accurate simulation of the atmospheric component of the water budget (i.e., evaporation, precipitation and runoff). A specific configuration of the version 3.1 of the weather research and forecasting (WRF) regional climate model was shown to systematically overestimate the Mediterranean Sea water budget mainly due to an excess of evaporation (similar to 1,450 mm yr(-1)) compared with observed estimations (similar to 1,150 mm yr(-1)). In this article, a 70-member multiphysics ensemble is used to try to understand the relative importance of various sub-grid scale processes in the Mediterranean Sea water budget and to evaluate its representation by comparing simulated results with observed-based estimates. The physics ensemble was constructed by performing 70 1-year long simulations using version 3.3 of the WRF model by combining six cumulus, four surface/planetary boundary layer and three radiation schemes. Results show that evaporation variability across the multi-physics ensemble (similar to 10 % of the mean evaporation) is dominated by the choice of the surface layer scheme that explains more than similar to 70 % of the total variance and that the overestimation of evaporation in WRF simulations is generally related with an overestimation of surface exchange coefficients due to too large values of the surface roughness parameter and/or the simulation of too unstable surface conditions. Although the influence of radiation schemes on evaporation variability is small (similar to 13 % of the total variance), radiation schemes strongly influence exchange coefficients and vertical humidity gradients near the surface due to modifications of temperature lapse rates. The precipitation variability across the physics ensemble (similar to 35 % of the mean precipitation) is dominated by the choice of both cumulus (similar to 55 % of the total variance) and planetary boundary layer (similar to 32 % of the total variance) schemes with a strong regional dependence. Most members of the ensemble underestimate total precipitation amounts with biases as large as 250 mm yr(-1) over the whole Mediterranean Sea compared with ERA Interim reanalysis mainly due to an underestimation of the number of wet days. The larger number of dry days in simulations is associated with a deficit in the activation of cumulus schemes. Both radiation and planetary boundary layer schemes influence precipitation through modifications on the available water vapor in the boundary layer generally tied with changes in evaporation.