Tropical Cyclone Track Predictability and the Adjoint Method of Data Assimilation.

摘要

This thesis explores a new method to improve hurricane track forecasts. This is done by modifying the model's initial conditions using the adjoint method developed by Talagrand and Courtier (1987). The idea is to run the model forward using the governing equation, and then run the model backwards using the adjoint equation. The result of the forward integration is the distance function, and the result of the backward integration is the gradient of the distance function, where the distance function is a scalar measure of the distance between the observed and model hurricane track. The gradient of the distance function between the observed and model hurricane track. The gradient of the distance function is used in a minimization scheme that modifies the initial conditions. These new initial conditions produce a model track closer to the observed track. Like Talagrand and Courtier, we derive the adjoint method using the spectral nondivergent vorticity equation. However, to eliminate computational error, here we use the Adams-Bashforth time integration scheme instead of the leapfrog method. Experiments were run using the nondivergent barotropic model to indicate how the adjoint method can improve hurricane track forecasts. (KR)