Reanalysis data are often used to carry out scientific research, although if it is not clear the extent their errors. This article explores the inherent uncertainty about using reanalysis data of maximum temperature and minimum in southeastern South America. It was compared seasonal mean fields TX (summer) and TN (winter) observed and interpolated grid points (Tencer et al., 2011), three different multidecadal-reanalysis (NCEP, ERA40 y 20CR) and four regional climate models (LMDZ, PROMES, RCA and REMO). It was studied also the surface energy balance for each reanalysis and was found that the involved processes in this balance affect directly to the temperature. Errors in temperature are partially linked with errors arising from how regional climate models reproduce the sensible heat flux, latent heat flux and surface net radiation. The ability of the reanalysis and regional climate models to represent the geographical distribution of TX and TN it was analyzed through Taylor diagrams. Ensembles of reanalysis or ensembles of regional climate models usually have better statistics in these diagrams than individual reanalysis or models. Moreover, the statistics shown by the Taylor diagrams suggest that errors in the geographical distribution of spatial anomalies of temperature of both reanalysis and regional climate models have similar magnitudes.
Keywords: maximum temperature, minimum temperature, gridded dataset, reanalysis, regional climate models