Estimates of hourly global irradiance based on geostationary satellite data
with a resolution of several (2 to 10) kilometres reproduce ground-measure
d values with a Root Mean Square Error (RMSE) of typically 20% to 25%. The
different components of this RMSE have been enumerated by several authors b
ut, due to the lack of adequate measurements, their respective importance i
s not well settled. In the present study we attempt to quantify these compo
nents from a practical point of view, that is from the point of view of use
rs having to rely on time/site specific irradiance data. We conclude that t
he intrinsic, or "effective" RMSE is more along the line of 12%. This effec
tive RMSE is the measure of the methodological imprecision (satellite-to-ir
radiance conversion models). The remaining part of the overall RMSE is the
amount by which spatially averaged satellite-derived estimates are, by thei
r very nature, bound to differ from ground-measured local insolation.