Spatial Regimes of Heavy Precipitation over the Czech Republic: A Regime-Based Evaluation of Observed and Historical Model Daily Precipitation

Regional climate models reproduce many large-scale characteristics of daily precipitation, but substantial uncertainty remains in simulating heavy precipitation, especially in summer and over complex terrain. Previous evaluations of the historical ALADIN-CLIMATE/CZ simulation over the Czech Republic showed that standard diagnostics based on mean fields, bias, correlations, and percentile characteristics are useful for assessing overall performance, but provide only limited information on whether the model reproduces the correct spatial organization of precipitation events. This is a key issue for heavy precipitation, whose impacts depend not only on event magnitude but also on where precipitation is concentrated within the domain.

This study evaluates daily precipitation over 1995–2014 using recurrent spatial precipitation regimes derived from a Self-Organizing Map. The analysis is based on a historical model run that provides an appropriate baseline for identifying model biases relevant to interpreting future climate simulations. Observed and modeled daily fields are classified in a common regime space, allowing direct comparison of the frequency, intensity, and spatial structure of characteristic precipitation patterns. This framework is particularly suitable for heavy precipitation because it helps distinguish between errors in the occurrence of particular regimes, errors in precipitation intensity within those regimes, and errors in the representation of terrain- and season-dependent rainfall structures.

The results are consistent with previous evaluations, but provide a more process-oriented interpretation of model error. The historical run reproduces the main characteristics of daily precipitation reasonably well, yet important deficiencies persist in reproducing extreme events. In particular, the model simulates a reduced diversity of precipitation regimes, with wet days concentrated into a relatively small subset of recurrent patterns. It also overproduces annual exceedances of the observational extreme threshold, while the largest deficiencies occur in summer extreme regimes, especially those linked to complex terrain, where conditional intensities are generally too weak. Overall, the regime-based analysis complements conventional validation by showing that uncertainty in simulated heavy precipitation arises not only from bulk bias, but also from imperfect representation of the diversity, frequency, and spatial structure of daily precipitation regimes.