Evaluation of multiple reanalyses in reproducing the spatio-temporal variability of temperature and precipitation indices over southern South America

Abstract

Several temperature and precipitation indices, with special focus on extremes, were analysed in different sub-regions of southern South America during 1979–2017 using multiple reanalyses, the CPC gridded data set and the most extended network of meteorological stations employed in regional climate studies up to date. Reanalyses generally well represented the spatial patterns of the indices, although they showed some differences in extreme indices over large portions of southern South America and tended to overestimate precipitation maximums, especially in southern Chile. Furthermore, ERA-Interim presented clear difficulties in reproducing precipitation near the Andes Mountains, exhibiting the largest overestimations. This seemed to be improved in the new generation of ERA reanalyses (ERA5). When evaluating the long-term changes, most of the data sets agreed in general warming conditions, stronger and more homogeneous for the maximum temperature. NCEP1 and NCEP2 reanalyses showed contrary temporal changes in almost all the temperature indices. Precipitation indices exhibited less consistent changes among reanalyses, although significant upward trends were detected for precipitation extremes in southeastern South America and downward trends were detected in southern Chile in the observational data sets. In addition, most of the data sets agreed in drier conditions in the arid diagonal region of Argentina as reflected by significant positive trends for dry spells and negative trends for the total annual precipitation. In terms of the inter-annual correspondence, reanalyses usually presented good correlations to the stations reference in the regional averaged series, mainly for temperature indices and more variable for precipitation indices. Overall, no reanalysis was found to perform best. The use of reanalyses data to perform regional climate studies should consider the existent differences among them and with observational data. Moreover, using multiple sources of information is strongly recommended to account for observational uncertainty, especially in regions like southern South America, where data availability and its resolution are often limited.