Decomposition of diffuse solar irradiance into clear-sky and cloudy components using all-sky imagery in broken cloud situations

Clouds strongly modulate the radiation budget and surface radiative fluxes. An accurate representation of the cloud radiative effect by atmospheric models is thus highly important to correctly represent the energy budget. Evaluation of models with reference observations is however complicated by the point nature of surface measurements, and the high temporal variability of irradiance in particular underneath broken cloud fields. Here, the effects of temporal changes in cloud optical properties are superimposed on those resulting from changes in cloud cover and geometry relative to the sun. All-sky imagers are frequently used to estimate cloud cover, or to infer cloud type. In this study, we investigate their ability to relate cloud cover and diffuse solar irradiance during broken cloud situations. An effective cloud fraction is used to linearly interpolate diffuse irradiance between a clear-sky and a fully overcast value, which are determined from a fit. While cloud cover is traditionally defined as fraction of the solid angle covered by clouds, alternative definitions are also considered here, by introducing a cosine weighting analoguous to the definition of irradiance, and to account for enhanced forward-scattering by cloud droplets. The ability of these different effective cloud fractions to reproduce observed time series of diffuse irradiance are studied for several cases, and the best-performing model is determined.