AVALIAÇÃO DE MODELOS DE ESTIMATIVA DE RADIAÇÃO SOLAR PARA LOCALIDADES ÚMIDAS E SEMIÁRIDAS

Abstract

Solar radiation is the most important weather element for the development of agricultural crops. The objective of this work was to apply and assessing models to estimation the solar global radiation, using air temperature data for the weather conditions of João Pessoa-PB and Uruçuí-PI, both regions in Brazil. Weather data were obtained by INMET meteorological stations, from the years 2009 and 2010 in Uruçuí, and from 2015 to 2017 in João Pessoa. In each data set, an analysis was performed to exclude missing or wrong data from the readings. Thus, the data set in Uruçuí counted 693 days and in João Pessoa 591 days. To apply the Rs estimation models, it was necessary to obtain the solar radiation at the top of the atmosphere (Ra) and the solar radiation on clear sky days (Rso). Next, the estimation models of incident solar radiation were implemented, which were: Hargreaves-Samani Model (HS), Thornton-Running Model (TR) and Weiss Model (WS). The models were assessed with the indicators: Coefficient of determination of the linear regression (R²); Mean ratio (RM); Mean absolute error (EAM); Square root of mean square error (RQME) and the Wilmott concordance index (dW). For Uruçui-PI, the TR and HS models performed similar to each other. Two of the five indices showed these models as the more adequate. The worst model, therefore, was the WS model. For the locality of João Pessoa-PB four of the five indicators recommend the use of the TR model. The analysis of the performance of the models led to the conclusion that both HS and TR do not work as well in tropical conditions in which the decrease of Rg due to increase cloudiness or winter season does not occur together with the reduction of air temperature local.