The influence of cloud cover on the reliability of satellite-based solar resource data

Satellite-based solar resource data are generally developed and validated using binary cloudiness categories: either clear or cloudy. To investigate the reliability of solar resource data in partially cloudy conditions, we estimate cloud fraction using two distinct algorithms: a physical retrieval using surface observed global horizontal irradiance (GHI) and direct normal irradiance (DNI) and a temporal average of cloudiness estimated using observed DNI. Our analysis reveals a significant presence of scattered clouds, broken clouds, and mismatches between satellite- and surface-based cloud information at 17 surface sites across the contiguous United States, though confidently clear and cloudy conditions collectively occur more than 70% of the time. Solar radiation is computed using the National Solar Radiation Database (NSRDB) algorithm and validated using surface observations. Our findings suggest that, in the presence of scattered clouds, NSRDB data for clear-sky conditions can be significantly overestimated. Under cloudy-sky conditions classified by satellite data, DNI computed by the Fast All-sky Radiation Model for Solar applications with DNI (FARMS-DNI) can be underestimated when fractional cloudiness is identified using surface observations. The bias observed in several cloudiness categories indicates that the NSRDB is exceptionally accurate in confidently clear conditions. However, clear-sky conditions with scattered clouds and mismatched cloud data contribute significantly to the overall uncertainties in the NSRDB. Therefore, future improvements in solar resource data should involve development and implementation of satellite-derived cloud fraction and should consider radiative transfer modeling that accounts for amplified cloud reflection and fractional cloudiness.