Calibration of solar radiation models for Europe using Meteosat Second Generation and weather station data

Abstract

Solar radiation is a key input variable for crop growth models. However, direct measurement of solar radiation is operationally done for a limited number of weather stations. Instead of direct measurements, empirical solar radiation models are used that link solar radiation to more commonly measured meteorological variables. Coefficients for these models are site-dependent and therefore generally interpolated from the limited locations where solar radiation is measured. In this study we calibrated three solar radiation models (Ångström -Prescott, Supit-Van Kappel, and Hargreaves) using a solar radiation product derived from Meteosat Second Generation data. We demonstrated that the accuracy of daily solar radiation increases when using satellite data for calibrating model coefficients, as compared to interpolated ground-based model coefficients. The average relative root mean square error for Meteosat Second Generation-based calibrated models was 1.9% lower for the Supit-Van Kappel model (p<0.001, n=137), and 1.8% lower for the Hargreaves model (p<0.001, n=222). We did not find a significant improvement for the Ångström -Prescott model. Finally, we interpolated Meteosat Second Generation-based model coefficients to create continuous coefficient maps for Europe. These allow to estimate solar radiation from sunshine duration, cloud coverage and air temperature range for every location in Europe without preceding calibration. We conclude that Meteosat Second Generation-based calibration of model coefficients improves the accuracy of solar radiation estimates used in crop growth models.