Identifying Typhoon Targeted Observations Sensitive Areas Using the Gradient Definition Based Method

Increasing additional observation in the typhoon targeted observation sensitive area can help providing more accurate initial field for numerical models, further to improve the typhoon forecast skill. The critical problem is how to identify the typhoon targeted observation sensitive area. Conditional nonlinear optimal perturbation (CNOP) has been proved to be an effective method. Generally, the CNOP is solved using adjoint-based method, which needs to utilize the adjoint models of the numerical models. However, the adjoint models for some numerical models have not been developed or only for some modules. The gradient definition based method is an adjoint-free method, which has been applied to solve the CNOP of Zebiak-Cane (ZC) model with 1080-dimensional solution space to study the optimal precursors of El Nino-Southern Oscillation (ENSO) event. It is very easy to realize, but the time efficiency will go down dramatically along with the rapidly increasing dimensions. In this paper, the gradient definition based method is applied to solve the CNOP of MM5 model with more than 105-dimensional solution space to identify the typhoon targeted observation sensitive area. Compared to the adjoint-based method, the identified sensitive area and the benefits of the CNOPs are very similar for typhoon Matmo (2014) and Fitow (2013), and higher time efficiency can be achieved. Furthermore, the OSSEs’ results show that the sensitive area identified can be used to improve the forecast skill of typhoon Matmo and Fitow to different extent.