Improvement of tropical cyclone track forecasting using a model-constrained 3D-Var data assimilation scheme

In this study, a new 3DVar method is proposed by adding a stable function to the cost function of the available 3DVar as a constraint, similar to the method put forward by Huang and Wu, 2001. By this method, minimizing distance between observations and model variables and time variance of model variables makes the optimized initial conditions satisfy the constraints of full dynamics and physics in the numerical model instead of some simple constrains in general 3DVar methods. The forward and adjoint models used in this method are as same as those in 4DVar method but only are integrated one time step to calculate time variance. Because of only using observations at one time slice and being constrained by the model, it is called model-constrained 3DVar (MC-3DVar), which is implement based on MM5 4DVar system in this study.

Using this new 3DVar, AMSU-A retrieved air temperature are assimilated for simulations of 32 tropic cyclone (TC) cases. The results show a big decrease in track forecast error. Meanwhile, one case study of assimilating AMSU-A temperature, QuikSCAT sea level winds, and cloud drift winds gives dramatic track error decreases. All the results show that the MC-3DVar technique can be used to improve numerical TC forecast and the inclusion of more satellite data into it can give better forecasts.