Geostatistical comparison of UERRA MESCAN-SURFEX daily temperatures against independent data sets

Abstract

Regional reanalises are attractive new sources of meteorological data for the growing society of the end users, due to their physical consistency, dynamical coherency, and multivariate products at higher than the global reanalises spatio-temporal resolution. The assessment and quantification of uncertainties of the products of the regional reanalises and their added value are crucial for the interpretation. Hence these products could be also incorporated in the regional climatology, consistent comparisons of their long-term timeseries against independent and representative data sets have to be performed. The present study could be considered as step ahead in this direction - the MESCAN-SURFEX, which is the product with the most detailed spatial structure among all others in the UERRA (Uncertainties of Ensembles in Regional Reanalysis) project, is compared against two gridded observational data sets in South-east Europe: the well-known regional CARPATCLIM and the product of the Bulgarian National Institute of Meteorology and Hydrology ProData. The comparison aims to assess the skill of MESCAN-SURFEX to reproduce the climatological field of the mean temperature. Additionally, the daily extreme temperatures are estimated using the MESCAN-SURFEX output on sub daily basis and the results are compared against their CARPATCLIM- and ProData-counterparts. The computation of the mean and extreme temperatures with the MESCAN/SURFEX data are performed for the whole time span of this product and the comparison against the references for the whole time span of each of them on daily basis. The main conclusion of the study, which agrees with the outcomes of more detailed recent evaluations, is that MESCAN-SURFEX reproduces realistically the regional temperature field over Southeast Europe. According to the mean temperature, the differences remains under certain limits (RMSE generally below 2 °C) without, at least not apparent, systematic and spatial pattern. The estimation of the extreme temperatures produces results with biases comparable to the biases of the mean temperature, which makes the proposed method applicable for certain cases.