3A.8 Relationships between tropical cyclone track and intensity model initialization adjustments and early model forecast errors

Monday, 16 April 2012: 3:15 PM
Champions AB (Sawgrass Marriott)
Richard D. Knabb, The Weather Channel Companies, Atlanta, GA

Output fields from the latest individual runs of operational, fully dynamical models used in tropical cyclone forecasting are not available before human-generated, official forecasts for that cycle must be released. Since they cannot use these “late” model runs, official forecasts utilize “early” model output from the previous forecast cycle. Track and intensity forecasts from these early models are derived by adjusting the short-term (6- or 12-hour) model forecast to the observed position and intensity at the start of the next forecast cycle, and then applying each adjustment to the remainder of each model's track and/or intensity forecast.

Human forecasters often consider, among many other factors, how each model initializes the tropical cyclone. The adjustments made to create the early model track and intensity forecasts partially quantify model initialization errors, and they could perhaps provide some clues to forecasters about the potential accuracy of each early model prediction beyond the initial time of the new forecast cycle.

This paper examines the extent to which the adjustments made, relative to the observed position and intensity, to create the early model track and intensity forecasts are related to the subsequent errors in those early model predictions. The focus is on the benefit to short-term (12- to 24-hour) forecasts, but the relationships with forecast errors out to five days will also be examined. The behavior of several models relative to track forecasts will be described, but intensity forecasts from the fully dynamical regional hurricane models will similarly be considered. The goal is to provide human forecasters will a little more “guidance on guidance” at the time they make their track and intensity forecasts, hopefully improving slightly upon the consensus (average) of the most reliable models, especially at lead times within the watch/warning time frame of 48 hours or less.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner