Toward the elimination of bias in satellite retrievals of sea surface temperature: 1. Theory, modeling and interalgorithm comparison

Abstract

The along‐track scanning radiometer (ATSR), launched in July 1991 on ERS‐1, is an infrared radiometer designed to permit retrieval of skin sea surface temperature (SST) to the accuracy required for many climate research purposes. Using the prelaunch retrieval scheme, this accuracy (0.3 K) was achieved only when observations at 3.7 μm were available, i.e., SSTs derived from nighttime scenes before the failure of this channel in May 1992. Retrievals using only channels at 11 and 12 μm suffered significant biases. First, cold biases of up to 1.5 K arose from the radiative effects of the unanticipated presence of a significant loading of stratospheric aerosol following the eruption of Mount Pinatubo in June 1991. Second, cold biases of up to 0.4 K were associated with regions of high water vapor loading. We solve the first problem by choosing retrieval coefficients to be orthogonal to the modeled changes in brightness temperatures caused by variations in stratospheric aerosol optical depth. We attribute the second problem to deficiencies in radiative transfer modeling of water vapor continuum absorption and show that use of an updated parameterization reduces bias from wet atmospheres. Applying the new retrieval coefficients to ATSR data, we find good consistency between SSTs retrieved with and without the 3.7 μm channels, the global mean and standard deviation of differences between retrievals being of the order of 0.05 K and 0.25 K, respectively. We therefore anticipate that reprocessing ATSR data using our new retrieval scheme will result in a substantially improved record of ATSR SST, in that the following should be reduced to insignificant levels: (1) the artefactual trend (previously ∼0.25 K yr−1 in tropical regions) corresponding to the decaying load of post‐Pinatubo aerosol, (2) the discontinuity in SST retrievals (previously up to 0.7 K) associated with the failure of the 3.7 μm channel, and (3) cold biases (previously ∼0.4 K) in wet tropical regions. Thus this work represents a significant advance in terms of the quality of ATSR SSTs for climate research. The techniques are also applicable to both the ATSR‐2, flying on ERS‐2, and the advanced ATSR (planned for launch on the Envisat platform in 2000). However, we note that even with the improved physical modeling on which the new retrieval coefficients are based, we do not yet meet the stringent requirement of 0.1 K decade−1 stability in retrievals for climate change detection purposes.