Statistical adjustment of dynamical tropical cyclone model track predictions.

Abstract

A technique of statistically adjusting dynamical forecasts of tropical cyclone motion was tested. All tests were performed with operationally-analyzed data from the U.S. Navy Fleet Numerical Weather Central (FNWC). Three sets of regression equations were developed to modify forecasts of typhoon tracks. The first set of equations was based only on forward integration of the FNWC Tropical Cyclone Model (TCM) for 28 cases in 1975-76. An independent sample of cases from 1977-78 indicated that the first equation set was based on too small a sample size, especially considering the anomalous nature of the 1975-76 storm tracks. A second equation set based only on forward integration of the TCM was derived from 61 storm track forecasts from 1975-78. Results from the experiments with these equations indicate that systematic data and model errors can be used to statistically adjust forecast storm tracks. The second equation set based on forward integration showed improvement over the unmodified model predictions at all forecast times. A third equation set based on forward and backward integration of the TCM explained the greatest amount of variance of all the equation sets. In a dependent test of these equations using 31 of the 1977-78 cases, the U.S. Navy 7th Fleet error goal of 100 and 150 nautical miles at 48 and 72 hours was nearly met.