An analysis of sea surface temperature (SST) is described. It incorporates in situ observations and retrievals from one microwave and three infrared sensors. Statistical interpolation is used to update the analysis daily on a global grid with a resolution of 1/3°. The background or first-guess field is essentially the analysis from the previous day. Satellite retrievals and buoy observations undergo a thinning and all data is subjected to a careful quality control. A scheme to remove large-scale biases from the satellite data is included, and its impact is assessed.
Analysis error is estimated using two sources of independent data with similar results. The global average r.m.s. error is less than 0.4 K. Zonally averaged errors were computed over 15°-wide latitude bands giving errors in the range 0.25 K to 0.5 K. The contributions from infrared and microwave data are found to be roughly of equal importance. The two data types are shown to be complementary, producing a significant improvement in analysis error when used together compared with the error obtained when only one is used. The analysis also compares favourably with SST analyses produced by three other centres. Copyright © 2008 Crown in the right of Canada. Published by John Wiley & Sons, Ltd.