Inter-comparison and validation of high-resolution surface air temperature reanalysis fields over Italy

Since its release in 2019 by ECMWF, ERA5 has represented the state-of-the-art for global reanalysis, also providing initial and boundary conditions for many regional reanalysis products. These latter have been developed using parametrizations optimized for specific locations and observations not used in ERA5, to resolve physical processes which occur at smaller scales and thus better describe the atmospheric variables at finer resolutions than the 31 km of ECMWF global reanalysis. In this study, we investigate the capability of different reanalysis products to reproduce the characteristic atmospheric processes of the Italian peninsula. For example, a dynamical downscaling of ERA5 to higher resolutions can be an asset in studying the occurrence of extreme events, where the small scales play an important role, because of the Italian complex orography and coastal-sea interactions. For these reasons, inter-comparison and validation of different reanalysis products over Italy are carried on in this study, comparing the ERA5, ERA5-Land (9 km resolution), the MEteorological Reanalysis Italian Dataset (MERIDA, 7 km), the Copernicus European Regional ReAnalysis (CERRA, 5.5 km) and the Very High-Resolution dynamical downscaling of ERA5 REAnalysis over ITaly (VHR-REA_IT, 2.2 km). 

In particular, this work focuses on the evaluation of the 1991-2020 monthly climatology of surface air temperature over the Italian territory for each reanalysis product. The climatologies obtained from the different reanalyses at their native resolution are compared with the climatology obtained from a high-density station observational network. These observations were gridded for each reanalysis over its native grid, using a methodology that specifically accounts for elevation adjustments given the elevation differences between reanalysis grid points and station locations.

The results show that ERA5 underestimates temperature over most of the Italian territory. The bias is around -1 °C on average, with peaks of -3 °C over the Alps, and a positive temperature bias of around +0.5 °C over the Po valley. ERA5-Land, MERIDA, and CERRA reflect a similar cold bias distribution over the Italian region, even if some differences in the pattern distribution can be found, with local cold bias intensification in some areas. VHR-REA_IT, on the contrary, presents a warm summer bias of about +3 °C on average over the Po valley. In addition to that, ERA5 exhibits a negative bias over the Alpine region which increases with elevation in winter. A similar gradient is also found in the other higher resolution reanalyses (ERA5-Land, MERIDA, and CERRA), with the only exception of VHR-REA_IT which, conversely, shows a positive bias increase with decreasing elevation in summer.