A Global Database For Quantifying Predictability of Solar Irradiance

Solar forecast is an essential subject in many engineering and geophysical fields. It is well known that the accuracy of solar forecasts may vary considerably across different geographical locations and temporal periods. In order to evaluate one's solar forecast, the root-mean-square error (RMSE) score has been widely used to compare the performances of different forecast models across various locations and time periods. Alternatively, for the same evaluation purpose, one may also employ a satellite-derived or reanalysis irradiance database with global coverage. In this paper, we propose such a global database of reference solar forecast accuracy using records from the Clouds and the Earth's Radiant Energy System over the time period 2000 to 2020. The final RMSE score can be calculated using the corresponding reference forecasts in our database, and thus provides a worldwide guideline for solar forecast research. Furthermore, We also discovered an interesting relation between RMSE on a global scale and the classification of climate.