Search the RAMMB website
(Sponsored by the Joint Hurricane Testbed)
Under previous JHT support, the Monte Carlo wind speed probability model (hereafter MC model) was developed to estimate the probability that any given location will experience 34, 50, or 64 kt winds over a given forecast period. In 2006 the MC model was successfully transitioned to operations at the National Hurricane Center (NHC), replacing the Hurricane Strike probability program. Subsequent JHT support has led to several improvements of the MC model, including the development of a verification program, the addition of forecast-specific track uncertainty through the inclusion of the Goerss Predicted Consensus Error parameter, and the development of the stand-alone Hurricane Landfall Probability Application (HuLPA). Other minor corrections made since 2006 include reducing the calculation time step, fixing inconsistencies in the wind radii, improving the intensity perturbations over land, and code optimization.
As the developers and primary users continue to gain experience with the MC model new opportunities to improve the model are identified. This proposal seeks to complete a number of upgrades to the current MC model, many of which are based on NHC feedback over the past few hurricane seasons. Proposed improvements to the current MC model include replacing the linear forecast interpolation scheme with a more precise spline fit scheme, applying a bias correction to the model track error statistics to provide consistency between NHC’s uncertainty products, and applying a bias correction to the radii-CLIPER used by the MC model to improve the accuracy of the wind speed probabilities for exceptionally small or large (e.g. 2012’s Hurricane Sandy) tropical cyclones. Additionally, several additions to the MC model are proposed such as estimates of the arrival and departure times of 34/50/64 kt winds, an integrated GPCE parameter, and wind speed probabilities beyond 5 days (proposed to 7 days). Finally, this proposal seeks to consolidate the error statistic generation code, which is currently in several pieces, into a single streamlined version that will reduce the time needed to update the MC model statistics each year.
In year 1, the following upgrades will be introduced:
An experimental version of the MC model is currently running at CIRA/CSU. During the season, plots of the current MC model, experimental MC model, and difference wind speed probabilities will be displayed in near real-time (links below, Atlantic only).