Weather-State-based evaluation of Mediterranean precipitation in climate models using satellite observations

Precipitation over the Mediterranean is strongly modulated by large-scale circulation and cloud-regime variability, posing persistent challenges for climate model evaluation. Standard model assessments based on aggregated precipitation metrics often mask regime-dependent errors and limit physical interpretability. Here, we present a Weather-State(WS)-based framework for evaluating Mediterranean precipitation in climate models, combining satellite observations with model simulations in a physically consistent manner.

Daily precipitation from the TRMM dataset (1998-2016) is analyzed together with cloud-regime information derived from ISCCP-based Weather States. The evaluation focuses on historical simulations from CMIP climate models over the Mediterranean (30°–45°N, 10°W–40°E). As a baseline, model-observation differences in total precipitation, precipitation intensity and precipitation frequency are first assessed to identify large-scale discrepancies. WS-conditioned diagnostics are then used as a complementary layer to examine precipitation characteristics within dynamically coherent regimes.

Within this framework, precipitation frequency, intensity, and spatial structure are evaluated conditional on WS occurrence, allowing regime-dependent behavior to be compared across models and observations while reducing the influence of regime mixing inherent in climatological metrics. The analysis further quantifies the contribution of individual WSs to total precipitation and examines the consistency of WS-dependent patterns and trends across models.

The analysis highlights the role of Weather-State-conditioned diagnostics in structuring model evaluation and enabling physically consistent comparison in structuring model evaluation and enabling physical comparison of precipitation characteristics across regimes, supporting improved understanding of model behavior without reliance on fixed assumptions about future change.