Abstract. This paper presents the reconstruction of the 80-year time series of daily global solar radiation (GSR) at the subtropical high-mountain Izaña Atmospheric Observatory (IZO) located in Tenerife (The Canary Islands, Spain). For this purpose, we combine GSR estimates from sunshine duration (SD) data using the Ångström–Prescott method over the 1933/1991 period, and GSR observations directly performed by pyranometers between 1992 and 2013. Since GSR measurements have been used as a reference, a strict quality control has been applied based on principles of physical limits and comparison with LibRadtran model. By comparing with high quality GSR measurements, the precision and consistency over time of GSR estimations from SD data have been successfully documented. We obtain an overall root mean square error (RMSE) of 9.2% and an agreement between the variances of GSR estimations and GSR measurements within 92%. Nonetheless, this agreement significantly increases when the GSR estimation is done considering different daily fractions of clear sky (FCS). In that case, RMSE is reduced by half, to about 4.5%, when considering percentages of FCS > 40% (~ 90% of days in the testing period). Furthermore, we prove that the GSR estimations can monitor the GSR anomalies in consistency with GSR measurements and, then, can be suitable for reconstructing solar radiation time series. The reconstructed IZO GSR time series between 1933 and 2013 confirms change points and periods of increases/decreases of solar radiation at Earth's surface observed at a global scale, such as the early brightening, dimming and brightening. This fact supports the consistency of the IZO GSR time series presented in this work, which may be a reference for solar radiation studies in the subtropical North Atlantic region.
Solar Radiation Time-Series Prediction Based on Empirical Mode Decomposition and Artificial Neural Networks
