An Examination of the Differences in Tropical Cyclone Pressure-Wind Relationships Among Observations, Model Analyses, and Model Forecasts

Since the pressure-wind relationship within a tropical cyclone is typically somewhat unique to that tropical cyclone (one aspect that can be considered a fingerprint), evaluating forecast models through their pressure-wind relationships provides a more insightful and more strict evaluation of their performance (rather than that done by track or intensity alone). While much research has been done on the development of pressure-wind relationships and how accurately they represent real tropical cyclones, little research has been done on pressure-wind relationships within modeled tropical cyclones, and the evaluation of tropical cyclone intensity forecasts by examining the pressure-wind relationships. For example, an unrealistic intensity “jump” has been often observed in multiple models from the initial conditions to the first forecast hour output by the model. Such sudden structural and intensity changes suggest that the analysis field of a numerical model has a different phase space of existence than that produced by numerical integration of the equations in that model. It is hoped that by examining the pressure-wind relationship revealed in the models, one can better quantify weaknesses and biases in the physics of certain high resolution hurricane models.

In this study, only the pressure-wind relationships and model forecasts of northern Atlantic Basin tropical cyclones will be used – partly for reasons of scope, but also because the Atlantic basin is currently the only basin that is routinely flown by aircraft reconnaissance. Model analysis (initialization) and forecast data for the GFDL, SHIPS, HWRF, and GFS from the Automated Tropical Cyclone Forecast (ATCF) archives is utilized, and the models' intensity (measured by both pressure and wind) forecasts are compared to both the National Hurricane Center's “best track” reanalyzed data and aircraft reconnaissance data from the actual tropical cyclones. The pressure-wind relationships of the 6-hourly model forecasts are compared to the observed MSLP and maximum wind speeds (the tropical cyclone's actual pressure-wind relationship, which of course has its own uncertainty). This project also provides analysis on the weaknesses of pressure-wind relationships in their approximations of the real atmosphere, and why the approximations may be in error. A subset of storms where models performed particularly poorly is also analyzed by comparing a storm's reanalyzed “best track” from the National Hurricane Center with the SHIFOR intensity forecast to allow for differentiation between scenarios that were simply “difficult” to forecast for any model, and scenarios where a certain model performed particularly poorly. Finally, possible explanations for the poor performance are provided with the hope that discovery of errors and biases in the current models will aid improvement of the physics of models still in development.