Application of Wind-Pressure Relationships to Global Model Output for Tropical Cyclone Intensity Forecasting

Deterministic global dynamical models are one of the primary tools used by forecasters at the National Hurricane Center. Improvement in track forecasts from the National Hurricane Center over the past 20 years can in large part be attributed to improvements in global dynamical models, particularly the National Centers for Environmental Prediction Global Forecast System (GFS), and the European Centre for Medium-Range Weather Forecasts (ECMWF) global model. These models, and their ensembles, are the primary guidance available to forecasters for tropical cyclone genesis and track forecasts. However, global models have not demonstrated skill in forecasting the maximum wind speed (intensity). While global model intensity forecasts are generally not explicitly considered by forecasters, the surface pressure field is sometimes used subjectively as a forecasting aid. For instance, if a global model forecasts a tropical cyclone to deepen, a forecaster may be more inclined to forecast strengthening. This study seeks to determine if global model pressure fields can be used directly to make tropical cyclone intensity forecasts.

Focusing on the GFS, surface pressure forecasts were converted to wind forecasts through the use of previously developed tropical cyclone wind-pressure relationships. The Hurricane Forecast Improvement Project stand-alone GFDL vortex tracker was obtained and modified to track the minimum pressure, outermost closed isobar, and several other variables. The tracker was run for several seasons of tropical cyclones in the Atlantic and eastern North Pacific basins. Minimum pressure forecasts were verified against the 2012 and 2013 best track. Despite having lower resolution, the GFS produced pressure forecasts that were comparable to, and often better than, the NCEP operational dynamical regional hurricane models. Maximum wind forecasts were then generated based on twelve different wind-pressure relationships. The Knaff-Zehr-Courtney (KZC, Courtney and Knaff 2009) was determined to be the best option based on verification of all the various relationships. Early results indicate that the KZC-based intensity has less error than the direct model wind forecast. Much of this improvement is due to a decreased negative bias. However, results also suggest that once the forecast is converted into “early” space, which accounts for bias in the first 6 hours of the forecast, the improvements are much smaller or negated entirely. Further research will be done to determine why the wind-pressure relationship does not translate to a better “early” forecast, and if adjustments can be made to create a better version of the forecast.