Statistical Forecasts of 24, 48 and 72 h Typhoon and Tropical Storm Intensity Changes

Abstract

A 10 year (1960?69) sample of observations in western North Pacific tropical cyclones over open ocean was used to derive statistical regression equations to forecast the maximum wind speed for 24, 48 and 72 h periods. Stratification of the dependent data by latitude bands, by months, and by maximum intensity were tested with both five-predictor and ten-predictor equations. An independent sample of tropical cyclones (July, August and September of 1955?59) was used to test the derived regression equations. Verification was in terms of the relative forecast error according to the acceptability criteria set by the Joint Typhoon Warning Center, Guam. Equations derived from a combined dependent sample of July, August and September storms stratified into two bands (one north, one south of 20°N) produced forecasts which were equivalent or superior to equations derived from storms stratified by months or in three 10° bands from 5°N to 35°N. These two five-predictor equations for the 24 h period were also superior to equations derived for storms within the classes ≤65, 66?100 and ≥101 kt. Although not tested with a homogeneous set of forecasts using operational data rather than post-season data, the objective forecast technique appears to give results comparable to or better than recent official intensity forecasts, especially for the 72 h interval.