Interpolação de dados meteorológicos utilizando covariáveis para a região metropolitana de Belo Horizonte

Abstract

Making continuous spatially distributed climates surfaces are important for a range of applications. Many methodologies are being studied to improved sparsely distributed meteorological data interpolation, both the study of the best method, being Thin Plate Spline (TPS) the most used for this application, as the use and combination of covariates that improved the interpolation. This work has the aim to test the use of covariates, as sensor MODIS images, like LST and cloud cover, DEM and distance to coast, using TPS, for the Belo Horizonte metropolitan region. The general model accuracy was good, especially for the temperature, with a RMSE from 1 to 4 oC. The covariates have marginally effects in Temperature and Precipitation estimation, because the models that didn‟t use any covariates had errors very similar of the models that used them. But for the wind speed the use of covariates, mainly cloud cover, was essential to improve the model.