Desempenho de modelos de estimativa da radiação solar na região do Triângulo Mineiro

Abstract

Solar radiation is determinant on global climate behavior and weather phenomena, and data on solar radiation is required in many applications in different areas of study. In agronomic studies, net radiation or energy represents the amount of energy that is available for the processes of evapotranspiration, photosynthesis and heating the air and soil, and is one of the factors that most affects evapotranspiration and crop productivity. However, not all weather stations are equipped to measure it. This study calibrated models for estimating daily solar radiation (Rs) using data from maximum and minimum air temperatures and rainfall, and to determine the impact of estimated radiation on reference evapotranspiration (ETo) and on the simulation of productivity for the Triangulo Mineiro region in Minas Gerais. Eight models were evaluated: Bristow and Campbell (BC), Chen (CH), Donatelli and Campbell (DC), De Jong and Stewart (JS), Annandele (AN), Hargreaves (HA), modified Hargreaves (HA-1) and Hunt (HU). The models were calibrated by the method of minimization of the square root of the squared sum of the difference between observed and predicted values, and evaluated by the root of the error mean square (RMSE), coefficient of determination (R2) and mean error (EM). Fhe effect of estimated Rs by each model on the ETo and on the simulation of productivity, comparing the results obtained between the estimated and measured Rs by RMSE and R2 was observed. There was greater variation of the coefficients as their number in the models increased, making the local calibration more important. All models tended to overestimate the Rs values below average and underestimated those above it, but these trends were stronger in the models HA, HA-1, AN and HU, especially for lower values. The models BC, DC, JS and CH were more accurate, with values of RMSE and R2 ranging from 2.54 to 3.30 MJ m-2 d-1 and from 0.60 to 0.74, respectively. ETo estimates presented R2 above 0.70 and RMSE less than 0.5mm d-1 for all the models, resulting in reliable ETo values for management and design of irrigation systems. Simulations of productivity are more accurate with Rs estimated by the models of group I, regardless of location, with R2 greater than 0.90 showing productivity estimates reliable for both the producer and for aiding prediction of agricultural crops.