Improving the Accuracy of the National Solar Radiation Database (NSRDB) using High-Resolution Data

The National Solar Radiation Database (NSRDB) is the most accessed, public database of solar irradiance and associated weather parameters for use in energy modeling and climate research. The NSRDB provides broadband irradiance, including global horizontal irradiance (GHI), direct normal irradiance (DNI), diffuse irradiance, and spectrally resolved irradiance. Data are delivered as time series or typical meteorological year data sets. The data is currently available for the contiguous United States, Central America, and part of South America. The Physical Solar Model (PSM) that underpins the current datasets is a unique physics-based model that has been developed at NREL and has opened the door to the use of next-generation satellite datasets for solar resource assessment and forecasting. PSM assimilates additional information content from recent satellite such as GOES-16 to improve solar radiation estimates. For 2018, NREL in collaboration with University of Wisconsin and NOAA developed satellite based-cloud properties from GOES–16 at 2km–5-minute resolution (Figure 1). Ancillary meteorological data from MERRA-2 at half-degree were downscaled to the GOES-16 product resolution. Using these NSRDB datasets for 2018 are produced at 2km–5-minute resolution by using the PSM models. This high-resolution dataset is capable of capturing the temporal and spatial variability of the sky condition and should demonstrate higher accuracy during times of high-variability. A validation of the performance of the NSRDB 2018 data is conducted to quantify the accuracy when compared to high-quality ground observations under both clear – and cloudy sky conditions. To understand the variability, various time scales will be used in the validation by averaging the ground measurement from five minutes to one hour centered on the satellite measurement time (every 5 minutes). This will provide insight on the subpixel variability and temporal variability captured by the satellite. A comprehensive uncertainty estimation by including all known sources of uncertainties including ground-based measurement uncertainty will also be presented.