Opportunities and Challenges For Probabalistic Extended Range (7-30 Day) Forecasts of Tropical Cyclone Events

An opportunity exists to take tropical cyclone predictions to the next level of extended range (7-30 days) by using advanced dynamical models, and thus make a large step toward a seamless prediction from one day to seasonal forecasts. Recent studies (Elsberry et al. 2010, 2011; Tsai et al. 2012) have evaluated the European Center for Medium-range Weather Forecasts (ECMWF) 32-day ensemble predictions of tropical cyclone-like vortices in the western North Pacific on time scales of 7-30 days. Rather than the typical ensemble-based display of the probability that the tropical cyclone would pass within 120 km of a point within the forecast interval, similar ensemble member vortex tracks are matched to form “ensemble storm” tracks. A weighted-mean motion vector technique is applied that gives the largest weight to those member vortex 12-h vectors that originate closest to the endpoint of the prior 12-h track position. Two ensemble storms that approach within 180 n mi are merged. An objective verification method has been developed to search for all ensemble storms that have similar tracks to each Joint Typhoon Warning Center (JTWC) official track, and a quality measure is assigned based on membership functions for the shortest distance, average distance, distance at formation time, and distance at ending time. Thus, the focus is on the overall track similarity. Tsai et al. (2012) calculated contingency tables of hits, misses, false alarms, and correct negatives to summarize the performance of the ECMWF 32-day ensemble forecasts during the 2009 and 2010 seasons in the western North Pacific. Nearly all of the JTWC storms were predicted with a few misses that were short-lived tropical depressions or weak tropical storms. However, a large number of false alarms (ensemble storm tracks that could not be matched with the JTWC tracks within the specified space and time thresholds) were also revealed by the objective verification. Studies are in progress to understand, detect, and eliminate/minimize these false alarms. A spline technique has been applied to the ensemble member vortex tracks to produce more smoothly varying motion vectors. A cluster analysis of ~9000 fitted ensemble member tracks resulted in eight track clusters, which have been simplified to westward, northwestward, northeastward, and recurving tracks. A new option is to only allow ensemble member vortices of the same track cluster type to be included in an ensemble storm. The time evolution of percentages of the track cluster types in track bifurcation scenarios is also provided to assist forecasters in assessing the most likely option. A method of objectively combining tracks that are nearby in space and time, and thus most likely represent the same tropical cyclone, is expected to increase the reliability and decrease the number of false alarms. The new version of the technique will be tested during the 2012 season with the focus on the Atlantic during the NASA HS3 field experiment. Preliminary results will be presented at the Special Symposium. Several World Meteorological Organizations are cooperating to facilitate the participation of various research groups, tropical cyclone forecasters, and numerical weather prediction centers in addressing both the opportunities and the challenges for extended-range forecasts of tropical cyclone events.