OPTIMAL AVERAGING OF NOAA NASA PATHFINDER SATELLITE SEA-SURFACE TEMPERATURE DATA/

Statistical interpolation is applied to 8 years of cloud-interrupted s atellite radiometer data to produce estimates of sea surface temperatu re (SST) at 10-day intervals in the Indo-Australian region. The data a re 9-km resolution daily ''best SST'' values calculated globally by th e NOAA/NASA Pathfinder project reanalysis of advanced very high resolu tion radiometer (AVHRR) data. The optimal averaging technique determin es an unbiased estimate of the signal that has the minimum mean square variance from the data, within the limits of the expected measurement error. Previous studies have shown that the method is superior to oth er linear averaging techniques, especially that of simple composite av eraging. The method is applied in the time domain only, preserving the 9-km spatial resolution of the data. The signal and noise covariances were evaluated from the data. This was done with care so that accurat e estimates of the error bounds that bracket the optimally averaged va lues might be obtained. These error bounds were then verified against in situ data. A Markov function, (1 + tau/a) exp (-tau/a), where tau i s the time lag and a is a characteristic timescale, was fitted to the data and used for the signal correlation function. This was selected a fter evaluation of functional forms proposed in other studies. The eff ect on the analysis of geographical variation in the correlation funct ion was considered. The computational demand of the repeated matrix op erations in optimal interpolation was reduced by using a limited durat ion data window. The complete analysis procedure for the 8-year data s et, comprising over 10(6) time series, was tractable on a modern works tation. The result is a set of SST maps for 1987-1994 at an interval o f 10 days and a spatial resolution of 9 km. The analyses are suitable for applications such as high-resolution ocean and atmosphere modeling where the timescales and space scales of interest are comparable to t he analysis (i.e., of the order of 10 days, 9 km) and for which the pr esence of gaps due to clouds is problematic. Some features of Indo-Aus tralian regional mesoscale circulation that the analysis highlights ar e examined, including examples of detailed mesoscale SST evolution and interannual variability.