REDUCING CLOUD CONTAMINATION IN ATSR AVERAGED SEA-SURFACE TEMPERATUREDATA

The Along Track Scanning Radiometer (ATSR) was launched in July 1991 o n the European Space Agency's first remote sensing satellite ERS-1. AT SR has the potential to measure sea surface temperature (SST) to a pre cision of 0.3 K, which is more than double the accuracy of any previou sly flown infrared radiometer. A key factor limiting ATSR's performanc e is remnant cloud contamination. Examination of the 0.5 degrees spati ally averaged ATSR SST data (version 500) from the South Atlantic for the whole of 1992 and 1993 shows the presence of regional cloud contam ination in the night SST measurements. The authors establish a figure of 5.7% as a lower limit for this nighttime cloud contamination. The c ontamination leads to differences between day and night mean SSTs and to poor comparisons with in situ thermosalinograph SST data. A new clo ud filtering process designed for postprocessing of the data is propos ed to remove the contamination. The algorithm presented here relies on assumptions that the day data are less cloud contaminated than the ni ght data and that a large proportion of the SST variability can be exp lained by an annual and semiannual model. Testing the filtering algori thm shows that differences between the day and night SST signals are s ubstantially reduced and that comparisons with the thermosalinograph S ST data improve by a factor of 3 in rms scatter and by 0.3 K in the me an difference.