Using AATSR data to assess the quality of in situ sea-surface temperature observations for climate studies

In situ data are widely used to provide a ground truth for the calibration and validation of satellite sea-surface temperature (SST) retrievals. They are also used to monitor long-term changes in the climate. For these applications, and others, it is necessary to understand the uncertainties in the data. Near-coincident SST observations from the Advanced Along-Track Scanning Radiometer (AATSR) and in situ platforms were used to understand the characteristics of errors in the measurements. The mean random error on the AATSR retrievals was found to be 0.14 K. The in situ errors were modelled as a constant offset plus a random error. For ships, the standard deviation of the constant offset was estimated to be 0.71 K and the mean random error was 0.74 K. For drifting buoys, the standard deviation of the constant offset was estimated to be 0.29 K and the mean random error was 0.26 K. These results suggest that there is a need to revisit current assessments of the adequacy of in situ observing systems. The trend in global-average SST between 1991 and 2007 calculated from in situ data was compared to its counterpart calculated from the ATSR instruments. The in situ record warms more slowly than the ATSR record, probably due to a decrease in the fraction of relatively warm-biassed ship observations contributing to the global-average SST over the period and a corresponding increase in the fraction of relatively unbiassed drifting buoy observations.