Geographic distribution of hurricane model track forecast errors

Forecast models are an important tool in the forecasting of tropical cyclone movement within the North Atlantic basin. The performance of an earlier generation of models has shown sensitivity to geographical location of tropical cyclones in the north Atlantic basin (Neumann and Pelisser, 1981). This study is designed to determine if the impact of cyclone location continues to be a factor in the performance of the current set of forecast models and, if so, to investigate key factors responsible for the geographic dependency of the models. In this study, we examine the performance of statistical (CLIPER & NHC90), simplified dynamic (BAMD, BAMS, & VBAR), and three-dimensional primitive equation (GFDL & AVN) models within each of four quadrants within the Atlantic Basin divided by 25° N latitude and 55° W longitude. In addition, we compare the AVN to other global scale models (NOGAPS & UKMET) to determine the effect of boundary conditions on our results. Results show that the GFDL model produces the best model forecasts in all quadrants. In contrast, the AVN model produces the least variation in performance between quadrants, but displays no skill (model performance relative to the CLIPER) in forecasting storm tracks in the southern quadrants. The remaining models show varying degrees of skill as well as different levels of bias between one quadrant and another. The East-West and North-South biases can be attributed to various factors including, but not necessarily limited to: 1) model horizontal and vertical resolution, 2) inadequate model physics for varying environmental conditions (e.g. tropics vs. mid-latitudes), 3) changes in input data quality and coverage (e.g. near the North American continent vs. over the open ocean), and 4) initial vortex bogusing techniques. The relative importance of each factor in producing biases differs from model to model. Results from this study will provide valuable information to both operational hurricane forecasters and numerical modelers to improve hurricane track forecasts.