Abstract
An empirical analysis program, based on finding an optimal representation of the data, has been applied to 120 observations of twenty nine 1973 and 1974 North Pacific tropical cyclones. Each observation consists of a field of Nimbus-5 Electrically Scanning Microwave Radiometer (ESMR-5) radiation measurements at 267 grid points covering and surrounding the tropical cyclone plus nine other non-satellite derived descriptors. Forecast algorithms to estimate the maximum wind speed at 12, 24, 48 and 72 h after each observation were developed using three bases: the non-satellite-derived descriptors, the ESMR-5 radiation measurements, and the combination of the two data bases. Independent testing of these algorithms showed that the average error made by algorithms developed from all three bases was less than the average error made by the persistence 24, 48 and 72 h maximum wind speed forecast and less than the average errors made operationally by the Joint Typhoon Warning Center (JTWC) 48 and 72 h maximum wind speed forecasts. The algorithms developed from the ESMR-5 base alone outperformed the JTWC operational forecast for the 48 and 72 h maximum wind speed. Also, the ESMR-5 data base, when combined with the non-satellite base, produced algorithms that improved the 24 and 48 h maximum wind-speed forecast by as much as 10% and the 72 h maximum wind forecast by approximately 16% as compared to the forecast obtained from the algorithms developed from the non-satellite data base alone.
