Analyzing and predicting the extratropical transition of tropical cyclones

Extratropical transition (ET) is a process that a tropical cyclone can undergo as it moves to higher latitudes over colder seas and interacts with the midlatitude regime. The resulting extratropical cyclone can be powerful with heavy precipitation to the left of track and a large extent of gale-force-winds to the right of track. These large asymmetries produce significant hazards to coastal settlements and maritime activities that range from heavy precipitation, large waves, high winds, and even fire weather. The whole transition process typically takes 18-48 hours to be completed with two basic possible end results, either dissipation or intensification of the combined system. These two classes of tropical cyclones behave very similarly during the early stages of extratropical transition. Thus, the end result of this transition is very hard to predict accurately even with full-physics atmospheric prediction systems such as the Navy's Operational Global Assimilation and Prediction System. Although there are many factors involved in the process, research has shown that the reintensification of the extratropical cyclone results primarily from its interaction with a midlatitude trough. Thus, the end result of extratropical transition may depend more on the phasing between the tropical cyclone and the trough it moves into rather than the details of the tropical cyclone structure.

Here we present results from a multi-stage, pattern-recognition technique to predict the simplest of ET outcomes: whether an ETing system will reintensify or dissipate after ET. To date the system uses NOGAPS analyses as input to a statistical technique that separates the discriminating large scale factors associated with reintensifying or dissipating ET cases. Using these data we have achieved a best performance of 88% detections with 27% false alarm rate on test storms from 2003 and 2004 with a limited training set from 1997 - 2002. It is of interest to analyze how the technique determines whether a particular tropical cyclone will reintensify or not by reconstructing the actual physical fields produced by the technique. Results show that the reconstructed patterns are physically meaningful, and by examining the overall prediction performances of multiple variables at multiple levels in the atmosphere, we are able to determine which physical parameters are of major importance to ET. In addition, we are looking at creative ways to increase the training set so that more useful classes of ET can be discriminated by this system. These classes include: dissipation, weak reintensification; moderate reintensification; strong reintensification; and timing to minimum pressure (e.g., 24 hours, 48 hours, > 72 hours).