New Methods for Incorporating Situation-specific Track Uncertainty into the Monte Carlo Wind Speed Probability Model

The Monte Carlo wind speed probability model (MC model) is the base algorithm used to generate the 34-, 50-, and 64-kt Wind Speed Probabilities (WSP) product at the National Hurricane Center (NHC). The MC model, first developed under the Joint Hurricane Testbed, incorporates past track, intensity, and radii forecast errors to generate a set of 1000 forecast ensembles (i.e., realizations) based on the official NHC forecast. This original methodology only used climatological error distributions and did not include any uncertainty information from numerical models.

Several years after implementation at the NHC, the MC model was updated to incorporate information about numerical model track uncertainty using the Goerss Predicted Consensus Error (GPCE). The GPCE-version of the MC model divides past NHC official track forecast errors into terciles based on the GPCE value. GPCE is a function of numerical forecast model consensus spread, and hence the GPCE-version of the MC model was the first attempt to incorporate track uncertainty information from numerical models. The GPCE-version of the MC model has proven to be more skillful than the original MC model and was implemented in operations at NHC in 2010.

Although the GPCE-version of the MC model incorporates numerical model uncertainty, it does so in an indirect way that is still heavily constrained by past official track forecast errors. We will present two new methods for incorporating numerical model track uncertainty in the MC model in a more direct manner through the use of numerical model track ensemble data. Since the WSP product is a public forecast product issued by NHC, the MC model output needs to be consistent with the official NHC track and intensity forecast. Both methods presented here adhere to this constraint while still directly using track data for a number of global ensembles. Preliminary results show that each of these methods have the potential to be somewhat more skillful than the GPCE-version of the MC model when verified over a multi-year sample. For certain cases where the ensemble spread is large and asymmetric with respect to the official forecast track, these methods improve the skill of the WSPs by as much as 50%. These verification results, along with a description of each method’s strengths and weaknesses, will be presented.

Disclaimer:

The views, opinions, and findings contained in this article are those of the authors and should not be construed as an official National Oceanic and Atmospheric Administration (NOAA) or U.S. Government position, policy, or decision.