Phase space-based evaluation of numerical forecasts of cyclone structure evolution

Evaluations of operational model simulations of tropical cyclone evolution have traditionally been restricted to assessment of the accuracy of the track and intensity forecasts. While these storm characteristics are clearly important for identifying the regions at most threat from destructive winds, the area of damaging winds, precipitation, and wave activity are typically much greater in extent and their distribution rely on the synoptic structure of the evolving storm. In this paper we propose a novel method for assessing operational forecasts that utilizes cluster analysis to assess forecasts of storm structure evolution. Structure forecasts are shown to depart from the verifying analysis as forecast lead time increases. Biases in these forecasts differ based upon the initialization procedure of the model being assessed.