Added value of a multi-model ensemble of convection-permitting rainfall reanalyses over Italy

Abstract

Regional convection-permitting meteorological reanalyses substantially improve the atmospheric representation compared to convection-parameterized counterparts. This holds particularly for multi-scale driven variables such as precipitation in terms of spatial structures, intensity and frequency rates, or the timing and peak of its summer diurnal cycle. However, the simulation of convective-related phenomena is highly model-dependent, implying the inability to sample the full range of natural variability with single-model experiments. This challenge is exacerbated for km-scale simulations owing to the intrinsic chaotic nature underlying convection. Multi-model ensembles of high-resolution climate models demonstrate to reduce the simulation errors associated with individual model outputs over Europe. When applied to retrospective estimates, such ensemble approach could then offer a comprehensive, homogeneous, and optimized assessment of past atmospheric states.

This study presents the first multi-model ensemble of regional reanalyses over Italy considering four recently produced datasets to assess the added value of their joint use. These products are derived by dynamically downscaling the global reanalysis ERA5 with different numerical models: MERIDA_HRES, MOLOCH, SPHERA and VHR-REA_IT. The reference dataset for comparison is the pluviometer-based hourly analysis GRIPHO. The investigation over 2007–2016 includes the annual and seasonal variations in daily and hourly mean rainfall intensity and frequency, heavy precipitation occurrences, and their summer diurnal cycles. No single dataset systematically outperforms the others, and substantial inter-model variability is detected for summer precipitation. The ensemble improves rainfall statistical estimates compared to individual reanalyses by providing more realistic spatial patterns, enhanced skill, and reduced biases relative to observations. These findings have potential implications for downstream reanalysis applications.