Complementarity between Two Global Satellite-Retrieved Irradiance Products: GSNO and NSRDB

Surface solar irradiance retrieved from geostationary satellites constitutes the most fundamental information for solar energy meteorology. As most satellite-retrieved irradiance products only cover the disk fields of view of the corresponding satellites, two noncommercial global products stand out, namely, the Geostationary Satellite Network Observation (GSNO) system and the National Solar Radiation Database (NSRDB). Whereas GSNO is new and employs a hybrid retrieval scheme that couples cloud-microphysical inversion with a neural network–accelerated radiative transfer solver, NSRDB has long been established and is recognized as one of the most accurate physically retrieved irradiance products. This work performs a rigorous head-to-head comparison of the two products, validating their performance at worldwide locations covering diverse radiation regimes. NSRDB generally achieves superior performance. Notwithstanding, it is discovered that the two products possess strong complementarity, in that opposite biases are more often than not observed at the validation locations, highlighting the benefits of product merging. A simple weighted fusion of the two datasets reduces the overall normalized root-mean-square error from 20.93% (NSRDB) and 24.86% (GSNO) to 17.25%.