Spatial interpolation of monthly mean precipitation in the Rio Bravo/Grande basin

Abstract

Precipitation is one of the primary climatic variables used to describe hydrological processes.Nevertheless, their spatial representation is difficult in areas with complex orographic effects and limited coverage of weather stations. The present study analyzed monthly rainfall data in order to reliably represent the spatial distribution of monthly mean precipitation (MMP) in the Bravo/Grande River Basin (CRB). Data were used from 201 weather stations located inside and around the basin. With information from 60% of the stations, selected randomly, multiple linear regression models were fitted to predict MMP as a function of elevation, complexity of the topography, coastal proximity and geographic location of stations, which explained between 70 and 82% of the spatial variability of precipitation occurring during the rainy period. Monthly maps of MMP were obtained, which were spatially calibrated by interpolating the residuals. Validation tests of the spatial calibration were conducted before and after for the remaining 40% of the stations. The validation tests showed efficiency values (EF) between 0.41 and 0.82 and mean absolute percentage error values (%EMA) between 19.1% and 39.5%. The best predictive months were from May to August. The calibration of the models significantly improved the reliability of the interpolations for every month (EF between 0.60 and 0.90, %EMA between 16.2% and 30.1%), making it possible to obtain reliable geographical coverage and high spatial resolution, with the potential for considering them as input variables in models to assess hydrological processes in the CRB.