Responding to an old thread, but just wanted to mention that gridded data computed using a combination of physics-based models and satellite data is considered a more reliable source then weather station data for (at minimum) solar and wind data for use in renewable energy modeling. This would explain why it’s appropriate for SAM to link to data from the NSRDB, rather then some collection of EPW files.
I only learned this myself from @josephyang’s posts here, where he explains exactly why it is better, and also the argument for reanalysis data for BEM: Looking for Beta Testers for Weather Data Service - #16 by josephyang
Also, I would infer that the vast majority of the data in the SAM epw files comes from satellite measurements instead of on-ground measurements. I’m 99% sure that the solar radiation data in these epws is from satellite measurements and this may be the case for the air temperature measurements too. Understandably, satellite-derived data has much larger error bars around it than ground-observed data.
This is probably true, but I believe the integration with physics-based modeling and the interpolation from real data makes the gridded data ultimately more reliable then raw satellite and weather station data. The fact that Climate One Building partially updated it’s weather files with ERA5 data for radiation indicates that this is definitely the case for radiation values at least.