A comprehensive comparison between satellite‐measured skin and multichannel sea surface temperature

At present, the most widely used algorithms for the calculation of sea surface temperatures (SSTs) from infrared satellite data are based on the regression of infrared radiances with in situ buoy SST measurements (the multichannel SST, MCSST, and the cross‐product SST, CPSST). Drifting buoys measure the SST anywhere from 0.5 to 1.0 m below the surface, thus missing the important radiative skin of the ocean. It is the temperature of this skin layer that is active in the air‐sea heat exchange. To determine the differences between the algorithms for the satellite‐measured skin SST (or satellite‐measured surface skin temperature, SMSST), the MCSST and the CPSST, historical infrared imagery was used to compute coincident maps of each product from which the temperature differences were calculated. The MCSST ‐ SMSST differences were computed for five consecutive seasons. CPSST ‐ SMSST comparisons were computed with two different CPSST algorithms for periods from two different years. Also, a global map of SMSST, using both AVHRR and HIRS data, was computed for one period and compared with the AVHRR‐only SMSST, MCSST, and CPSST maps. Histograms of all differences, stratified by zonal band, clearly suggest the mean and rms differences between the SMSST, MCSST, and CPSST The MCSST has a mean offset from the SMSST of 0.45 K, while the CPSST and SMSST have a mean offset of 0.1 K. The MCSST has its greatest rms differences (∼0.7 K) with the SMSST at low latitudes with much lower rms differences at mid and polar latitudes. The CPSST and SMSST, however, exhibit larger rms differences (∼0.6 K) at mid and higher latitudes where smaller differences would be expected. The AVHRR/HIRS SMSST has a small mean offset from the AVHRR‐only SMSST (0.20 K) and CPSST (0.19 K) but shows a large amount of scatter with both algorithms (rms difference of 0.59 K with SMSST, 0.65 K with CPSST). Owing to the lack of in situ skin SST measurements, it is difficult to say which of the algorithms produces the most accurate results; only the differences may be discussed.