Regional trends in Nimbus-7 OLR: Effects of a spectrally nonuniform albedo

Abstract

Data from the Earth Radiation Budget wide-field-of-view sensors on board the Nimbus-7 satellite during the period 1979-87 indicate a statistically significant downward trend in daytime outgoing longwave radiation (OLR) of up to 14 W m-2 decade-1 over the Sahara desert region of Africa. Reflected solar radiation shows a positive trend of the same magnitude. No significant trends in temperature or cloudiness over the region can be demonstrated. Atmospheric desert dust aerosols increase over the same period, but the spatial pattern of the dust trend is different from that of the OLR and reflected solar trends. A plausible explanation is that these trends are instrumental in nature and are caused by degradation of the Suprasil filter dome that covers the reflected solar channels. The filter dome degradation is greater for shorter wavelengths of the solar spectrum compared to red and near-infrared wavelengths. This enhanced degradation at shorter wavelengths, combined with the increased albedo of deserts at longer wavelengths, causes the global correction applied to the reflected solar data to be too large over desert regions. The OLR is calculated by subtracting the reflected solar irradiance from the unfiltered total outgoing irradiance (reflected solar plus emitted longwave). Thus, if the total irradiance remains unchanged, an erroneous upward trend in the solar data will cause a downward trend in the OLR estimate of similar magnitude. A preliminary calculation of the magnitude of the filter degradation effect is presented. Laboratory and thematic mapper spectral reflectance data of desert sand are used to estimate the reflectivity of the desert surface as a function of wavelength. The degradation of the shortwave filter as a function of wavelength is estimated with a quadratic fit to the known deterioration rates of the reflected solar and near-infrared channels. Using these parameters, the magnitude of the trend explained by this effect is calculated to be 11 W m-2 decade-1, which is nearly as large as the trends in the data.