COMPARISON OF SPECIAL SENSOR MICROWAVE IMAGER VECTOR WIND STRESS WITHMODEL-DERIVED AND SUBJECTIVE PRODUCTS FOR THE TROPICAL PACIFIC

A new source of vector wind stress data is assessed relative to existi ng analyses of the surface wind field. The large-scale variability of vector wind stress generated by Atlas et al. (1991) and based on the s pecial sensor microwave imager (SSM/I) remotely sensed observations of surface wind speed is compared with five operational and subjectively analyzed wind products across the tropical Pacific basin for the firs t year of SSM/I, July 1987 through June 1988. The conventional wind st ress data considered are the operational wind products from European C entre for Medium Range Weather Forecasts (ECMWF), National Meteorologi cal Center (NMC), and Fleet Numerical Oceanography Center (FNOC), and the subjective analyses from Florida State University and the Universi ty of Hawaii. The spatial and temporal variability of the zonal compon ent, meridional component, and curl of the wind stress are examined re lative to their future use in wind-driven ocean model studies of tropi cal Pacific Ocean circulation. The basin-scale structure of the SSM/I data fall within a range bracketed by ECMWF, the University of Hawaii, NMC, and the Florida State products. The SSM/I data are shown to be m ost similar to the ECMWF analysis and the subjective analysis of satel lite-derived cloud motion wind vectors (SAWIN) performed at the Univer sity of Hawaii. The basin-wide mean rms difference between the SSM/I d ata and these two products is 0.18 dyn cm-2 for tau(x) and 0.10-0.12 d yn cm-2 for tau(y). In order to place these differences within context , the basin-wide mean standard deviation for the temporal variability is found to be 0.23-0.26 dyn cm-2 for tau(x) and 0.15-0.16 dyn cm-2 fo r tau(y). On regional scales, some of the differences among these prod ucts are greater than 0.3 dyn cm-2. The ECMWF and SAWIN analyses are t he most similar (basin-wide mean rms difference tau(x) = 0.16 dyn cm-2 , tau(y) = 0.11 dyn cm-2) of the 15 possible product versus product co mparisons. This may indicate a high weight given to cloud motion wind vectors in the ECMWF analysis for data sparse regions of the tropical Pacific. The wind data from the FNOC analysis forecast system in use i n 1987-1988 (and since upgraded) was the most dissimilar wind product. The relatively dense space-time coverage of the SSM/I satellite data (order of 27 observations per 2-degrees x 2.5-degrees grid square ever y 1.5 days), together with the large-scale similarity with conventiona l wind products, suggests that the SSM/I-based analysis represents a n ew source of surface wind information suitable for ocean modeling stud ies. As a result of the potential demonstrated here, strong considerat ion should be given to the use of these wind data in forcing ocean cir culation modeling studies. Furthermore, the prospects of processing th e surface wind speed retrievals from spaceborne passive and active mic rowave sensors dating back for more than 10 years should be examined.