Calibration of Solar Irradiance Data Using Optimal Transport
Abstract
This study investigates the use of optimal transport (OT) for the site adaptation of satellite-derived solar irradiance data. Using one year of ground measurements from three distinct locations for calibration and a subsequent year for validation, we evaluate the effectiveness of OT for correcting the joint distribution of the clear sky index (kt) and diffuse fraction (kd). We compare the performance of OT against two standard techniques: quantile mapping (QM) applied to kt and kd independently, and linear regression (LR) applied to global horizontal irradiance (GHI) and diffuse horizontal irradiance (DHI). DHI estimates derived from decomposition models applied to LR-corrected GHI are also assessed. The results demonstrate that OT provides a more robust calibration, better preserving the physical relationship between kt and kd compared to the other methods. Additionally, our findings show that both the original satellite data and the tested decomposition models systematically overestimate DHI at the studied locations.
Keywords
optimal transport, clear sky index, diffuse fraction, ground measurements, site adaptation