Group for High Resolution Sea Surface temperature (GHRSST) analysis fields inter-comparisons. Part 1: A GHRSST multi-product ensemble (GMPE)

Many sea surface temperature (SST) gap-free gridded analysis (Level 4, or L4) fields are produced by various groups in different countries. The Group for High Resolution SST (GHRSST) is an international collaboration body which has formed the inter-comparison technical advisory group (IC-TAG), to advise SST producers and users on the relative performance of these SST fields. This two-part paper describes two of the three major systems developed under GHRSST coordination towards this goal. Part one (this paper) describes the GHRSST Multi-Product Ensemble (GMPE) system, which runs on a daily basis at the UK Met Office, taking various L4 analyses as inputs, transferring them onto a common grid, and producing an ensemble median and standard deviation. The various analysis systems contributing to the GHRSST inter-comparisons are discussed, highlighting areas of commonality between the systems as well as those parts of the systems where there is less agreement on the appropriate algorithmic or parametric choices. The characteristics of the contributing L4 analyses are demonstrated by comparing them to near-surface Argo profile temperature data, which provide an independent measurement of SST and have been shown to provide a good estimate of foundation SST (the SST free of diurnal warming). The feature resolution characteristics of the L4 analyses are demonstrated by calculating horizontal gradients of the SST fields (on their original grid). The accuracy and resolution of the GMPE median are compared with those of the input analyses using the same metrics, showing that the GMPE median is more accurate than any of the contributing analyses with a standard deviation error of 0.40 K globally with respect to near-surface Argo data. For use in climate applications such as trend analysis or assimilation into climate models, it is important to have a good measure of uncertainty, so the suitability of the GMPE standard deviation as a measure of uncertainty is explored. This assessment shows that, over large spatial and temporal scales, the spread in the ensemble does have a strong relationship with the error in the median, although it underestimates the error by about one third. Crown Copyright (c) 2012 Published by Elsevier Ltd. All rights reserved.