Estimation of Global Solar Radiation Based in Temperature Observations for the Goiás State

Abstract The objective of this work was to evaluate the performance of five models for estimating solar radiation based on temperature data for dry and wet periods in Goiás State. Climate data from 10 municipalities were used. The models performance for calibration and validation were evaluated by coefficient of determination (R2), root mean square error (RMSE), mean square relative error (RRMSE), the mean absolute error (MAE) and efficiency of model by the Nash-Sutcliff method (EF). Also, the models performance considering all data set for daily and monthly estimated solar radiation were evaluated by Willmott’s agreement coefficient, Camargo’s confidence index, spline smoothing cubic regression and linear regression. It was observed that HG and DCBB models showed the worst performance and CD and DB models the best performance for estimating solar radiation values for Goiás State.摘要 本研究旨在评估基于气温数据的五种模型在巴西戈亚斯州干湿期太阳辐射估算中的表现。本研究使用了10个市镇的气候数据。模型校准与验证阶段的表现通过决定系数（coefficient of determination，R²）、均方根误差（root mean square error，RMSE）、相对均方误差（mean square relative error，RRMSE）、平均绝对误差（mean absolute error，MAE）以及纳什-萨特克利夫效率系数（Nash-Sutcliff method，EF）进行评估。此外，针对全数据集下的日、月尺度太阳辐射估算结果，本研究采用威尔莫特一致性系数（Willmott’s agreement coefficient）、卡马戈置信指数（Camargo’s confidence index）、三次样条平滑回归与线性回归对模型表现进行了评估。研究结果表明，在戈亚斯州太阳辐射估算任务中，HG与DCBB模型表现最差，而CD与DB模型表现最优。