Application of a Moving-Coordinate Prediction Model to East Coast Cyclones

Abstract

A statistically derived procedure for forecasting the 24-hr displacement. change in central pressure, and change in intensity of east coast cyclones is presented. The analysis is based on a moving coordinate system?i.e., predictor information is measured at points fixed with respect to the moving cyclone center rather than at points fixed with respect to the earth. Two sets of multiple linear regression prediction equations are derived from a sample of 352 cyclone cases. The first set is based solely on point values of sea-level pressure, 500-mb height, 1000?500-mb thickness, and 12-hr changes of these variables. The second set is based on point values plus a series of quantities derived from point values such as steering components, gradients, vorticity and thickness advection, and thermal wind. When applied to 106 independent cases, the point value equation yield a vector position root-mean-square (rms) error of 4.01 deg of latitude and a central pressure rms error of 8.70 mb. The complex equations fail to improve these scores significantly. Operational 24-hr forecasts made for 31 cases during the winter of 1959?1960 are compared with corresponding 18-hr forecasts prepared by the National Weather Analysis Center (NWAC). The statistical forecasts incur a vector rms error of 2.25 deg of latitude for position and a rms error of 8.14 mb for deepening; NWAC forecasts incur a rms error of 3.87 deg of latitude and a rms error of 7.25 mb for deepening.