ANALYSES OF GLOBAL SEA-SURFACE TEMPERATURE 1856-1991

Global analyses of monthly sea surface temperature (SST) anomalies fro m 1856 to 1991 are produced using three statistically based methods: o ptimal smoothing (OS), the Kalman filter (KF) and optimal interpolatio n (OI). Each of these is accompanied by estimates of the error covaria nce of the analyzed fields. The spatial covariance function these meth ods require is estimated from the available data; the time-marching mo del is a first-order autoregressive model again estimated from data. T he data input for the analyses are monthly anomalies from the United K ingdom Meteorological Office historical sea surface temperature data s et (MOHSST5) [Parker et al., 1994] of the Global Ocean Surface Tempera ture Atlas (GOSTA) [Bottomley et al., 1990]. These analyses are compar ed with each other, with GOSTA, and with an analysis generated by proj ection (P) onto a set of empirical orthogonal functions las in Smith e t al. [1996]). In theory, the quality of the analyses should rank in t he order OS, KF, OI, P, and GOSTA. It is found that the first four giv e comparable results in the data-rich periods (1951-1991), but at time s when data is sparse the first three differ significantly from P and GOSTA. At these times the latter two often have extreme and fluctuatin g values, prima facie evidence of error. The statistical schemes are a lso verified against data not used in any of the analyses (proxy recor ds derived from corals and air temperature records from coastal and is land stations). We also present evidence that the analysis error estim ates are indeed indicative of the quality of the products. At most tim es the OS and KF products are close to the OI product, but at times of especially poor coverage their use of information from other times is advantageous. The methods appear to reconstruct the major features of the global SST field from very sparse data. Comparison with other ind ications of the El Nino - Southern Oscillation cycle show that the ana lyses provide usable information on interannual variability as far bac k as the 1860s.