Projected 21st Century Changes in Precipitation and Temperature Over Italy Using <scp>CMIP6 CMCC</scp> ‐ <scp>CM2</scp> ‐ <scp>SR5</scp> Model and <scp>COSMO</scp> ‐ <scp>CLM</scp> Dynamical Downscaling

ABSTRACT This study investigates the climate change signal over Italy using the regional climate model COSMO‐CLM v6.0‐clm1, driven by the CMCC‐CM‐SR5 global model under the CMIP6 scenarios SSP1‐2.6 and SSP3‐7.0. The added value of the high‐resolution COSMO‐CLM simulation is analysed by comparison with state‐of‐the‐art references. Fine‐scale spatial patterns for temperature and precipitation were assessed using E‐OBS observational dataset and CERRA reanalysis. Precipitation analysis was additionally supported by gauge‐based datasets SCIA and CERRALND reanalysis. CERRA was used to evaluate the localised intensity and spatial patterns of extreme precipitation events. The results show that COSMO‐CLM significantly outperforms its driving model across Italy, reducing temperature biases by 50%–75% (up to 2.7°C in summer) and improving precipitation distributions, particularly for extreme events. Although persistent warm biases in autumn (approximately 0.5°C–1°C) remain, COSMO‐CLM is more effective in resolving orographic climate signals, converting apparent overestimations of precipitation into meaningful increases in intensity. The models' skill varies with the reference dataset, despite consistent spatial patterns, highlighting how observational uncertainty propagates through model evaluation. Projections reveal a scenario‐dependent north–south precipitation dipole. COSMO‐CLM reduces the magnitude of the driving signals, particularly for summer temperature, which warms up to ~5°C. Furthermore, it adds critical mesoscale features not present in the driving data. A coherent signal does not emerge for annual precipitation, but both models concur on winter drying and summer wetting patterns, which COSMO‐CLM refines through localised enhancements. These localised changes, including varying warming rates and precipitation shifts, highlight the value of high‐resolution modelling for localising where impacts will be more severe and providing essential information. The findings, although based on a single model, highlight COSMO‐CLM's potential for representing climate projections over Italy by capturing mesoscale processes and extreme events with good performances. These results can provide, if included within a EURO‐CORDEX multi‐model perspective, valuable insights to support impact and risk assessment, thereby contributing to the definition of regional and local adaptation and mitigation strategies over Italy.