Wednesday, 9 January 2019
Hall 4 (Phoenix Convention Center - West and North Buildings)
A Monte Carlo probability (MCP) model, which is similar to that have been used in the National Hurricane Center (NHC) official forecast, is evaluated for its feasibility to predict landfalling hurricanes and compared with ensemble forecasts. The MCP model can produce the prediction of hurricane track and the wind estimations. The probability of wind speed at individual locations during hurricane forecast period can also be generated. First, in this study, the MCP model is applied to predict landfalling hurricanes with the GFS tropical cyclone forecast error statistics. Hurricanes Irma and Harvey (2017) are used as study cases. The GFS forecasts and history forecast errors are used to generated probability map for the 0-120h forecast with a time interval of 12h with 1000 samples of the track, intensity and structure realizations near and during hurricane landfall periods. Compared with best track data and GFS deterministic forecasts, outcomes from the MCP tend to more agree with GFS forecasts with small biases, while track errors are presented in both GFS and MCP forecasts. The analysis on track realizations from MCP reveals that track errors always lead by the random errors. The comparison between mean intensity realizations and best-track intensity indicates the wind speed forecasts are reasonable during both Irma and Harvey’s landfalls, despite the complexity of the track of the Hurricane Harvey. Comparing with GFS final analysis (FNL) data, the GFS-derived MCP results in the accurate prediction of the hurricane-affected areas. Results from MCP are also compared with GFS ensemble forecasts. It is found that the GFS ensemble forecast results are limited by their coarser resolution and small sampling size. Additional evaluations with more hurricane cases, the use of multiple ensemble forecasts from different models, and the combination of the MCP with ensemble forecasts are in progress. Results will be presented during the conference.
![](images/flag.png)
![](images/blue_ribbon.png)