Evaluating shortwave GEWEX Surface Radiation Budget fluxes as a function of sky cover conditions

Satellite surface radiative flux retrievals are routinely evaluated by comparison to time series of radiometric measurements from field sites. While such comparisons provide a general indication of the long-term reliability of the satellite products, distinguishing between retrieval algorithm errors and differences due to sampling mismatch between the satellite and surface measurements requires more detailed analysis. In this study, we use a radiometrically based sky cover classification scheme to segregate three-hourly GEWEX Surface Radiation Budget (SRB) shortwave flux values according to cloud conditions. We then analyze the agreement between surface and satellite downwelling shortwave flux values within these classes as a function of surface type for the midseasonal months of April, July, and October 2000 and January 2001.

In general, it was found that the satellite retrieved fluxes are lower than ground measurements for clear sky and broken cloud situations and higher than the ground measurements for overcast cases. This can be explained by higher sky cover variability over the 1° SRB grid boxes than was detected at the surface sites, even over a period of one hour. Cloud cover determined from ISCCP data, which is used as an input to the SRB shortwave algorithm, was correspondingly higher than cloud cover detected at the surface for clear and broken cloudy skies. However, ISCCP and surface determined cloud cover tended to agree for overcast cases. This suggests that both small clouds and small breaks in overcast clouds are likely to go undetected by satellites, thus complicating the retrieval of instantaneous fluxes at the surface.

Exceptions to the above results indicate cases in which sampling errors do not dominate SRB-surface measurement flux differences. For example, at tropical sites, satellite retrieved downwelling shortwave fluxes were found to be consistently higher than surface measured fluxes regardless of sky conditons. ISCCP cloud cover is nearly always lower than that determined from the surface at polar sites. The best agreement between ISCCP and surface cloud cover was found at desert sites although the flux differences are large at these locations. Inferences will be made regarding the causes of these error characteristics and how the SRB flux retrievals may be improved.