Gain of the AVHRR visible channel as tracked using bidirectional reflectance of Antarctic and Greenland snow

The snow surfaces of the high plateaus of the East Antarctic and Greenland ice sheets are used to determine multi-year drift in the sensitivity of the visible channel of the Advanced Very High Resolution Radiometer (AVHRR) on the polar-orbiting satellites NOAA-9, 10, and 11. Bidirectional reflectance distribution functions are empirically derived for the months of October-February (Antarctica) and April-August (Greenland) using a simplified atmospheric model. The bidirectional reflectance of the snow surface should not change from year to year for near-nadir satellite views. Therefore, drift in the derived bidirectional reflectance distribution function is interpreted as drift in channel sensitivity. Several factors make the snow surface of an ice sheet suitable as a calibration target far visible and near-UV channels. (1) In this spectral region, snow has a very high albedo (>97%) that is invariant with grain size and incidence angle. (2) On the high plateaus the temperatures are always far below freezing so the surface consists of cold fine-grained snow, and there is negligible contamination. (3) The ice sheet surfaces are uniform and flat across large areas. (4) Ozone is the only significant variable absorber in this spectral region, and its absorption can be accounted for if the ozone amount is known. (5) Cloud detection and removal is not necessary, because the thin clouds over the high ice sheets apparently do not alter the near-nadir reflectance, as they do over dark surfaces. Our analysis indicates that the visible channel on NOAA-9 degraded linearly over the 3.5-year lifetime of the instrument by 5.3+/-0.1% per year. NOAA-10 showed non-linear behaviour that could be fitted with a fourth-order polynomial. Data from NOAA-11 prior to the eruption of Mt Pinatubo showed a linear increase in sensitivity of 2.3+/-0.2% per year. The derived drifts are not sensitive to the choice of spatial and temporal averaging scales, the choice of months and gridboxes, and whether or not a cloud-rejection scheme is used.