Abstract
A very recent gridded product for the hourly global horizontal irradiance (GHI), derived from the measurements of the Earth Polychromatic Imaging Camera (EPIC) onboard the Deep Space Climate Observatory (DSCOVR) launched by a NOAA/NASA/USAF consortium, is validated at 31 locations worldwide, from January, 2017 to June, 2019. In contrast to those traditional methods that leverage (simplified) radiative transfer, this EPIC-derived product uses machine learning – a random forest model – to map out the connection between satellite-observed variables of various kinds and GHI. Nonetheless, the detailed validation conducted here shows that the quality of this EPIC-derived GHI dataset not only does not outperform those traditional gridded solar radiation datasets, but also contains undesirable artifacts that can be possibly attributed to inadequacies in the machine-learning procedure. For these reasons, it is not recommended to use this EPIC-derived dataset in its current form for solar resource assessment purposes.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
