Validation of ECMWF climatic data, 1979–2017, and implications for modelling water balance for tropical climates

Abstract

AbstractGridded meteorological systems greatly facilitate the analysis of the impacts of climate on crop development and productivity. Comparisons of these data with actual ground data validate this data source for various analyses in agricultural areas. The impact of the use of these grid data is an important evaluation for the temporal and spatial simulation of soil‐water availability for crops. We seek to verify how meteorological (ECMWF) data represents the surface water balance for Minas Gerais state. Monthly data for air temperature (T) and precipitation (P) from ECMWF were compared with the data from 771 ground stations (National Meteorological Institute, INMET) in the state of Minas Gerais in southeastern Brazil for 1979–2017. Potential evapotranspiration was estimated by Thornthwaite method (1948), and water balance was estimated by the method proposed by Thornthwaite and Mather (1955), with an available water capacity of 100 mm. We temporally and spatially compared the two data sources, and the comparisons were evaluated for accuracy using mean absolute percentage error (MAPE) root mean square error (RMSE) and for precision using the adjusted coefficient of determination (R2adj). ECMWF T and P tended to be temporally and spatially similar to the INMET data. The largest deviation between INMET T and ECMWF T was 2.81°C, mainly in the southwest of the state (the Minas Gerais triangle) and part of the central region during winter and spring, and the smallest deviation was −0.19°C in the northeast. The largest deviation between INMET P and ECMWF P was 75 mm·mo−1 in the summer, mainly between January and February in the central region of Minas Gerais. ECMWF T and ECMWF P allowed an accurate estimation of the components of the water balance. For example, the lowest MAPEs were 1.21% for ECMWF water‐storage capacity (southern Minas Gerais), 9.16% for ECMWF water deficiency (Vale do Jequitinhonha e Mucurí), and 8.69% for ECMWF excess water (Vale do Jequitinhonha e Mucurí). ECMWF can be used to estimate WB to represent surface stations, provided they are calibrated according to the region and seasons.