Improvement in Tropical Cyclone Forecasts by Multiple Linear Regression Equation Adjustment of Analog Tracks.
Abstract
A statistical technique is tested for adjusting the analog storm tracks used in the Northwest Pacific Analog Tropical Cyclone Forecasting Program (TYAN78). In the present version the tracks are rotated based on linear extrapolation of the deviations of the past 12-h motion for the analog and the current storm. The technique proposed here uses multiple linear regression equations to adjust the 12-h through 72-h analog positions. Predictors include translation speeds along the analog track from the -36-h to 72-h positions plus the comparison of the past 12-, 24- and 36-h translated positions of the analog with the current storm. By simply averaging the regression adjusted positions, the vector forecast errors were smaller than the TYAN78 errors. Both the TYAN78 and regression adjusted forecast errors were less than the official JTWC forecast errors. The improvement in the regression adjusted forecasts was in a smaller standard deviation relative to the mean position. For example, in the dependent sample the mean errors at 72 h for the official forecast, TYAN78 forecast and regression adjusted forecast were 438, 321, and 288 nautical miles, respectively. The corresponding standard deviations were 198, 151, 162 nautical miles. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1980
- Accession Number
- ADA092139
Entities
People
- Kenneth Allen Peterson
Organizations
- Naval Postgraduate School