Surface global irradiance assessed by three different methods

Abstract

Three methods will be used to evaluate the surface global irradiance: radiative transfer theory, empirical regression, and artificial neural networks (ANN). Radiative transfer is the fundamental theory that describes the propagation of radiation through a medium; empirical regression predicts surface global irradiance in simple parameterizations; artificial neural network, as an artificial intelligence technique, can also be tried to assess the surface global irradiance. These three approaches are studied in the present work. Data from the station “EL IDEAM” were used in the modeling experiments built upon these approaches to evaluate the daily transparency, also known as clearness index. We found out that the optimal inputs for artificial neural networks are extraterrestrial irradiance, surface relative humidity, and a pollution index based on particulate matter of sizes less than 10µm (PM 10 ). Surface relative humidity was suggested in a regression trial under meteorological conditions of “EL IDEAM”. By means of the programming code DISORT for the solution of the radiative transfer equation, daily irradiance characteristics were analyzed, and a hybrid model was created. Our results showed that artificial neural network produces higher scores than the other methods, though advantages and drawbacks are also discussed and compared.