Evaluating models for estimation of incoming solar radiation in Cruz das Almas, Bahia, Brazil

摘要

> size="2" face="Verdana">Solar radiation at the earth's surface (R_s) is one of the essential weather variables for several agro-environmental studies. In the absence of measured R_s, it's possible to simulate them by models. Using three years of daily R_s data (2004-2006) collected by an INMET (Brazilian Meteorological Service) automatic weather station installed in the municipality of Cruz das Almas, BA, three air temperature-based R_s simulation models were evaluated. As a preliminary step, data quality was investigated and, when needed, adjustments were made based on clear sky solar radiation envelopes. The following models were evaluated: Hargreaves-Samani for both interior and coastal regions, Thornton-Running, and the Weiss model. Linear regression R2, mean ratio (MR), mean absolute error (MAE), root mean square error (RMSE), and the index of agreement of Willmott (d_W) were used for model performance evaluation against measured R_s. The results recommended the coastal-region Hargreaves-Samani for prediction of R_s in the Reconcavo of Bahia, Brazil, since it presented the MR value closest to 1.00 (1.01), the lowest MAE (2.54 MJ m-2 d-1), the lowest RMSE (3.13 MJ m-2 d-1), and the highest d_W (0.90).