The National Hurricane Center (NHC) Next Generation Wind Speed Probability (WSP) Model

The National Hurricane Center (NHC) legacy Wind Speed Probability Model (WSP) estimates the probability of exceeding 34, 50 and 64 kt wind speeds from 1000 realizations that are centered around the official forecast. The WSP model is run on a 50 km grid and the probabilities generated from the model are point specific. For track and intensity, the realizations are based on the previous 5-year error distributions of the NHC and Central Pacific Hurricane Center (CPHC) forecasts. The errors are randomly generated using a Monte Carlo model, with only a weak dependence on the track spread of real time dynamical model forecasts. The wind radii associated with the realizations are based on climatological errors of a simple radii model that includes the official forecast initial radii as input, but not the official wind radii forecasts. Furthermore, the wind probabilities are calculated from the wind radii for each member, but the full 2-dimensional wind field is not calculated. Land surface effects are included in a very simple way, with a constant adjustment factor for inland points, and elevation effects not taken into account.

The next generation WSP model (referred to as WSP 2.0) is based on 1000 realizations of complete two-dimensional wind fields derived from an empirical modified rankine vortex algorithm referred to as the WTCM. The WTCM includes detailed land surface information in a surface layer model to reduce the wind speeds over land.. In addition to the legacy wind speed probability data for the thresholds mentioned above, WSP 2.0 will provide the probabilities of exceeding major hurricane force winds (96 kts) as well as wind speed exceedance grids/maps/graphics for several probability levels (namely, the 5%, 10%, 25%, 50%, 75%, 90%, and 95%). A preliminary version of the the model using the same track, intensity, and radii perturbations as the legacy model has been run in real time in experimental mode during the 2022 and 2023 seasons and they are accessible in CIRA’s website: https://rammb-data.cira.colostate.edu/tc_realtime/. The test runs shown on this website are based on the realizations run at 20 km resolution, another significant improvement over the legacy model.

Additional enhancements being explored and developed for WSP 2.0 include: 1) testing model configurations at 10 km and 5 km resolutions; 2) new methods to generate radii perturbations replacing climatology-based perturbations in the legacy model with one that perturbs the official wind radii forecasts and; 3) testing new ways of incorporating real time dynamical ensemble models information. This includes replacing the climatology-based error distributions to generate the track and intensities with a new machine learning based methodology. The development team is also coordinating this work with the Probabilistic Storm Surge Model (P-Surge) developers in an effort to unify and bring consistency to the wind forcing in both WSP 2.0 and P-Surge. Work is also underway to add elevation adjustments and a gust parameterization to the WTCM wind model and these could be added to a future version of WSP 2.0. Once the initial model development is completed, a full validation and calibration will be conducted, which will be important for downstream applications.

This presentation will detail the model design, a summary of the main findings during the first two years of in-house experimentation, and the path to operations. Preliminary results show that point probabilities over land are much smaller than in the legacy model (expected). Also, the land surface database used in the WTCM wind model is 1 km resolution but the 1000 realizations being tested so far are on a 20 km grid. This results in aliasing issues that are being addressed by testing a combination of higher resolution runs with de-aliasing post processing techniques. Finally, there is a need to recalibrate downstream applications that are based on the WSP model, such as the generation of wind-based watches and warning, public and marine forecasts, and risk communication products and guidelines. The transition plan is to continue to run the model experimentally for the next two years while development, validation, and calibration continues with an aim to implement it operationally in 2026 at the earliest.