23rd Conference on Weather Analysis and Forecasting/19th Conference on Numerical Weather Prediction


Tropical cyclone wind speed probabilities: Operational results and future applications

Michael J. Brennan, NOAA/NWS, Miami, FL; and D. P. Brown, R. D. Knabb, M. DeMaria, A. B. Schumacher, and J. A. Knaff

The National Hurricane Center (NHC) began issuing probabilistic wind speed guidance products for tropical cyclones (TCs) on an experimental basis in 2004, and these products became operational in 2006. Currently these products are generated from information (initial wind radii, and track/intensity forecasts) in the advisory packages issued by NHC, the Central Pacific Hurricane Center (CPHC) and the Joint Typhoon Warning Center (JTWC) A five-year climatology of intensity and track errors for each of these centers and climatological TC size distributions form the basis of the probability calculation.

A set of 1,000 plausible tracks intensities and wind radii (i.e., realizations) are created that vary around the official track and intensity forecast and the CLIPER-based wind radii using a Monte Carlo method for sampling historical official forecast errors and climatological size distributions. The wind swaths produced by this method are then used to determine the probability of experiencing of 34, 50, and 64-kt winds through the 120-h forecast time at a variety of point locations along the coast and inland, as well as on a hemispheric 0.5 x 0.5 degree grid. These products are designed to be used as a decision-making tool to evaluate the risk and uncertainty associated with tropical cyclone winds.

During the time period since these products were experimentally introduced, there have been numerous landfalling TCs in the United States that have provided an opportunity to evaluate the performance of the probability products. Basin-wide verification of these probability forecasts and their corresponding deterministic forecasts from 2006 and 2007 indicate that the probabilistic forecasts show increased skill in predicting the frequency of occurrence of the various wind thresholds over the deterministic forecasts from NHC, CPHC, and JTWC at all forecast lead times beyond 12 h. The verification results also indicate that the wind speed probabilities display a small bias, have high reliability, and are able to discriminate between events and non events.

A new method is being tested that allows the probabilities to vary based on the value of the Goerss Predicted Consensus Error (GPCE). Values of GPCE tend to be proportional to the spread of real-time dynamical model guidance, and represent a simple way to quantify the degree of uncertainty in the TC track forecast. Official track error distributions were sorted into three bins based on the value of GPCE, and it was found that the distribution of official track errors was much more varied when GPCE values were large. Based on these results, real-time values of GPCE are used to select a subset of the NHC track error distribution for the computation of the wind speed probabilities, allowing for a tightening (spreading) of the probability distribution when the GPCE values are small (large). Results from both the GPCE method and the operational Monte Carlo method will be compared for selected cases from the 2008 Atlantic hurricane season. Finally, the relationship of the wind speed probabilities to the placement of coastal hurricane watches and warnings will be presented.

wrf recording  Recorded presentation

Session 6B, Tropical Weather Part I
Tuesday, 2 June 2009, 4:00 PM-5:30 PM, Grand Ballroom West

Next paper

Browse or search entire meeting

AMS Home Page