Estimate of daily global solar irradiance based on thermal amplitude in Bahia state, Northeastern Brazil

ABSTRACT Accurate estimation of global solar irradiance (Rs) is essential for energy, agricultural, and climate applications, particularly in regions with sparse radiometric networks, such as the state of Bahia, Brazil. To address this limitation, empirical models based on easily obtainable variables, such as air temperature, represent viable alternatives. Thus, this study aimed to calibrate and evaluate five temperature-based empirical models: Bristow-Campbell (BC), Chen (CH), Hargreaves (HA), Hunt (HU), and Richardson (RI), to estimate daily Rs across 36 municipalities in Bahia. Observed solar radiation and daily maximum and minimum air temperature data were used, and model calibration was performed using the Levenberg-Marquardt optimization algorithm. The calibrated coefficients exhibited spatial variability, reflecting the influence of local climatic conditions. Statistical analyses showed that the correlation coefficient ranged from 0.62 to 0.82 and the agreement index from 0.73 to 0.93, indicating satisfactory model performance in most municipalities. Among the evaluated models, the BC model consistently showed the highest precision and accuracy, followed by the CH, RI, HU, and HA models. The BC model was classified as “very good’’ or “good’’ in approximately 67% of the municipalities, confirming its robustness and adaptability to Bahia’s diverse climatic conditions. These findings provide a robust methodological basis for selecting empirical models according to objective performance metrics. Additionally, they demonstrate that regional calibration substantially enhances the reliability of the estimates, contributing to a more accurate assessment of solar resources and more well-supported decision-making.