Large‐scale sea surface temperature variability from satellite and shipboard measurements

Abstract

A series of satellite sea surface temperature (SST) intercomparison workshops were conducted under NASA sponsorship at the Jet Propulsion Laboratory. Three different satellite data sets were compared with each other, with routinely collected ship data, and with climatology for the months of November 1979, December 1981, March 1982, and July 1982. The three satellite data sets were (1) AVHRR—SST estimates produced operationally by the National Oceanic and Atmospheric Administration from the advanced very high resolution radiometer aboard the NOAA polar‐orbiting weather satellites; (2) HIRS/MSU—SST estimates produced by a research group at the NASA Goddard Space Flight Center from the 20‐channel high resolution infrared sounder and the 4‐channel microwave sounding unit, also aboard the same NOAA satellites; and (3) SMMR—SST estimates produced by another group at Goddard from the scanning multifrequency microwave radiometer on the NASA research satellite Nimbus 7. The satellite and ship data were differenced against an accepted climatology to produce anomalies, which in turn were spatially and temporally averaged into 2° latitude‐longitude, 1‐month bins. Monthly statistics on the satellite and ship bin average SST yielded rms differences ranging from 0.58° to 1.37°C and mean differences ranging from −0.48° to 0.72°C, varying substantially from month to month and sensor to sensor. The SMMR generally had the largest rms differences and time‐dependent biases, while the AVHRR and HIRS/MSU had smaller more comparable values. The monthly bins were further smoothed spatially to correspond to 600‐km averages to further suppress the noise of individual observations, particularly for the ship data. When this was done the monthly ship data standard deviations about climatology varied between 0.35° and 0.63°C. Taking these values as true SST signal standard deviation levels, and the satellite‐ship rms differences as noise levels, results in signal‐to‐noise variance ratios of about 0.25 for SMMR and 1.0 for AVHRR and HIRS/MSU. Maps of SST anomaly reveal a complex pattern of partial agreement and disagreement between ship and satellite data. Maps of satellite minus ship and satellite minus satellite SST differences were often dominated by coherent large‐scale patterns of obvious geophysical origin related to distributions of surface wind speed, atmospheric water vapor, cloudiness, and stratospheric aerosols. Unfortunately, the spatial scales of these patterns are often quite similar to those associated with actual SST anomalies. Caution must therefore be exercised when dealing with the satellite data so that errors are not misinterpreted as true SST anomalies.