Accuracy Analysis of Estimates of Total Solar Radiation in Databases and Regression Models for Eastern Russia

Abstract

The efficient use of solar energy requires an accurate assessment of the incoming solar radiation. The study involves comparing the accuracy of monthly means of the global solar radiation flux from the ERA5 reanalysis database, the SYN1deg satellite-based observation database, climate reference data, and data of regression models for seven settlements in the east of Russia. The study employs two well-known regression models, including parameters of extraterrestrial solar radiation, total cloudiness, air humidity, minimum and maximum air temperature, atmospheric pressure, and a new regression model, which additionally includes parameters of low-level cloudiness and sun elevation angle. The accuracy of databases and regression models is evaluated by comparing their data with ground measurements of weather stations. The indices of the mean absolute error, root-mean-square error, and mean bias error are calculated. The comparison shows that the data of climate reference books for the period from 1937–1957 to 1980 have the smallest deviation from the estimates of the monthly mean flux of global solar radiation for 2006–2020 at most of the points discussed. The ERA5 monthly mean estimates of the global solar radiation flux are more accurate than the SYN1deg data at five of the seven points considered. The new regression model proposed in the study makes it possible to provide greater accuracy of monthly estimates of the global solar radiation flux compared to the data of SYN1deg and ERA5 for most of the points considered.