Statistical tropical cyclone intensity forecast improvements using GOES and aircraft reconnaissance data

Over the past few years some modest improvement in operational tropical cyclone intensity forecast skill has occurred due to the Statistical Hurricane Intensity Prediction Scheme (SHIPS) and improvements in the NCEP version of the GFDL hurricane model. The SHIPS model primarily relies on empirical relationships between intensity change and parameters that describe the storm environment, such as vertical wind shear and ocean properties. However, there is little input to SHIPS related to the structure of the storm inner core. In this paper, a new intensity forecast method is described that combines the usual SHIPS input with parameters determined from radial profiles of GOES infrared data and aircraft reconnaissance flight level winds within 200 km of the storm center. Two methods for the prediction are compared. The first uses the same linear regression methodology utilized by SHIPS. In the second method, the storm intensity changes are modeled using a first-order differential equation adapted from studies of population growth. The evolution equation includes two forcing terms. The first is a growth term, which is related to inner core processes that can be modified by environmental interactions. The second term dampens the growth rate as a storm approaches it maximum potential intensity (MPI), which is a function of the thermodynamic environment of the storm. This formulation separates the thermodynamic and dynamical effects, and constraints the storm evolution. Under a reasonable set of assumptions, all of the parameters in the evolution equation are known except the dynamical growth rate term, which is estimated statistically from the GOES, aircraft and storm environmental variables. This model is termed the GOES and Recon Intensity Prediction (GRIP) model. Results of independent tests from the 2005 Atlantic hurricane season will be presented.