Using Feature-Based Methods to Blend Tropical Cyclone Official Forecasts in the National Blend of Models

The National Blend of Models (NBM, Gilbert et. al. 2016) is a nationally consistent and skillful suite of forecast guidance based on a blend of both National Weather Service (NWS) and non-NWS Numerical Weather Prediction (NWP) model data and post-processed model guidance. However, blending multiple model forecasts using bias correction and whole-grid weighted averages can be challenging for discontinuous, high amplitude features such as Tropical Cyclones (TCs). Additionally, it can be problematic to smoothly marry the blended Direct Model Output (DMO) inputs with the National Hurricane Center (NHC)’s gridded representation of the TC forecast advisory message (WTCM). These challenges can result in forecast oddities such as TCs with multiple eyes, washed out magnitudes, and unrealistically large areal extents, which negatively impacts NWS forecasters’ workflow by requiring manual editing within the Graphical Forecast Editor (GFE) to produce a reasonable starting point for gridded, graphical, and text forecasts.

This paper describes work currently underway to blend the 2-dimensional representation of the NHC official TC forecasts into the NBM. We adapted feature-based techniques to address these challenges, known as TC Feature Matching. The technique first identifies TC features from the models (including regional hurricane models) that are input into the NBM. A hybrid feature is then created by matching the TC features via a fuzzy-logic method across models and then by stacking them over the center of the WTCM TC feature, better preserving magnitudes and eliminating extraneous eye features. The WTCM is then blended on top of this feature matched background wind, with the final hybrid TC placed on the background NBM wind grid. The technique has shown promise creating coherent TC structures that preserve the WTCM center location and wind speed forecasts greater than 34 kt. For winds below tropical storm force, the technique preserves background winds greater than those of the WTCM, which can be reduced below 34 kt by its land surface algorithm, ensuring that winds associated with higher elevations in the background NBM wind field are preserved.

The TC Feature Matching technique was first implemented into the NBM version 4.1 in early 2023, and future work for NBM version 5.0 has been scoped and will include: land/water masking when shifting TC features from among the members of the NBM tropical wind; conducting the blending technique in u/v space to account for both wind speed and direction (currently Feature Matching technique is only conducted on the wind speed); preserving eye structure in the final feature matched field; applying the WTCM blending only when a storm is classified as tropical or subtropical system (WTCM does not apply to extra tropical cyclones); and adding in the future a companion TC Wind Gust field to the NBM from the WTCM. The Feature Matching algorithm and methodology, and evaluations of Feature Matching in real time TC events over the 2023 hurricane season will be described.