Constructing Time Series of Mean Global Sea Surface Temperature Anomaly for Climate using Data from the ATSRs
Veal, Karen; Remedios, John J.; Ghent, Darren
University of Leicester, UNITED KINGDOM

The Along-Track Scanning Radiometers (ATSRs) have provided a near continuous time series of data from 1991 to 2012. Their highly accurate radiometric calibration, dual-view capability, three thermal infrared channels and the stability of the respective platforms, make the data suitable for the construction of climate time series. Furthermore, ATSR data are independent of in situ data allowing corroboration of the in situ climate record. In our previous work, we established a time series of ATSR data from ESA version 2.0 data including uncertainties in global mean sea surface temperature (SST). In this paper, we exploit the recently produced ATSR Reanalysis for Climate (ARC) sea surface temperature (SST) dataset which has regional biases smaller than ±0.05 K and stability better than 0.005 K yr^-1. The ARC dataset includes both skin and depth SSTs which also represents a considerable step forward in SST availability for climate studies. In particular, we present time series of mean global and regional skin and depth SST anomaly calculated using the ARC dataset and also time series of skin SST anomaly from ATSR version 2.0 data, with comparisons as appropriate.

The construction of time series needs to take account of a number of factors including inter-instrument bias and uncertainties at a monthly level. The ATSR series currently consists of three instruments with a period of overlap between consecutive instruments. Biases may exist between instruments due to, for example, calibration differences and observation time differences. When using data from more than one instrument in a time series any biases between data from different instruments must be accounted for and the data from different aligned. In our approach an analysis of data from the overlap periods is used to produce inter-instrument bias correction fields which are applied in order to align data from different instruments.

Calculation of uncertainties on averaged data must take into account both measurement and sampling errors. Infra-red sensors cannot observe the Earth's surface through cloud so sampling is irregular. A method for estimating the uncertainty on the monthly global mean which takes account of both temporal and spatial under-sampling is described. We show the importance of temporal sampling in the month and quantify the effect for the ATSR datasets.

Our results show the global mean SST was highly variable during ATSR mission and the time series is in good agreement with time series of in situ data from HadSST3. Uncertainties on the global mean SST anomaly are less than 0.015 K for ATSR version 2.0 data. This performance gives us good confidence in SST observations for climate over nearly two decades.

Satellite data can be used to calculate regional and global mean SSTA with low uncertainties. It is important that the Sea and Land Surface Temperature Radiometer due to fly on Sentinel 3 achieves similar accuracy and stability as achieved with the ATSRs so that the present time series can be extended into the future.