Probabilistic post-processing of CAMS radiation services
Abstract
This study examines the feasibility of ML-based probabilistic post-processing of the Copernicus Atmospheric Services (CAMS) Surface Solar Irradiance (SSI) satellite retrieval model. We develop a probabilistic neural netowrk framework that not only produces a best estimate of SSI, but also a full probability distribution. The resulting model appeared slightly more consistent that its deterministic equivalent and, most importantly, proved its ability to deliver a condition-dependent estimate of the estimate uncertainty.
| Origin | Files produced by the author(s) |
|---|