Application of the Lagged-Averaged Technique to Tropical Cyclone Track Predictions
Abstract
The lagged-averaged forecast (LAF) technique applied to tropical cyclone track prediction is a weighted sum of recent forecasts that were started from initial conditions at various times lagging the start of the forecast period. The goal of this study is to reduce the track forecast error at t + 24 h. Two tests of the LAF approach are presented to demonstrate feasibility. The first test uses the nine CLIPER forecasts initiated at 24, 30, 36, 42, 48, 54, 60, 66, and 72 h prior to the common verifying time. In this test, the mean 24-h forecast error is reduced by 8% relative to the 24-h CLIPER forecast above. In the second test, the 'modified' LAF involves only the five CLIPER forecasts initiated at 24, 36, 48, 60, and 72 h prior to the verifying time. However, the 36-h through 72-h CLIPER forecasts are first modified using statistical regression equations that include predictors related to new track information since these forecasts were initiated. Significant reductions in the track forecast error result from these statistical adjustments. The modified LAF applied to an independent sample results in a reduction from 189 km to 124 km in the mean 24-h forecast error or a reduction of 34%. This is a significant improvement because the JTWC mean 24-h forecast error for the last four years is approximately 210 km. The standard deviations are significantly reduced from 118 km to 69 km. Because the combination of the modified CLIPER forecasts in the LAF technique results in a significant improvement in performance, it is recommended that this technique be applied operationally and also be tested with dynamical models. (jhd)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1989
- Accession Number
- ADA218389
Entities
People
- A. B. Bacon
Organizations
- Naval Postgraduate School