Improved techniques for the derivation of sea surface temperatures from ATSR data

Abstract

A new technique is developed for the analysis of satellite data to provide accurate measurements of seasurface temperature (SST). Satellite data sets are partitioned into subsets depending on the value of a selected parameter (for example, latitude, total water vapor, and water vapor content of an atmospheric layer) to provide a suite of algorithm coefficients that reduces the errors associated with the derivation of SST. For data sets obtained with the along‐track scanning radiometer (ATSR) the data themselves can be used to modify the coefficients used in the SST algorithm, but for other instruments it may be necessary to use additional data sets to select a correct set of algorithm coefficients. A simulated set of ATSR data is used to develop algorithm coefficients for different atmospheric conditions, and the improvement in SST derivation is demonstrated. For ATSR, when all the data for three infrared channels in both the nadir and forward views are available, the improvement is marginal, but for situations when there are limited data the improvement is considerable. The analysis suggests that in the future the best SST analyses will be obtained by developing an interactive system, where the satellite data are ingested into a numerical weather forecast model so that algorithms can be selected and applied with a forecast or analysis of the atmospheric state. The new techniques are tested by application to a large global data set of ATSR brightness temperatures. This analysis highlights the future need for large SST validation data sets that include coincident satellite and surface‐based measurements.