Sensitivity analysis of statistical-dynamic model by adjoint method

摘要：

The present paper is concerned with the sensitivity of the prediction distance of a statistical-dynamic model with respect to initial positions and initial velocities of tropical cyclone (TC) centers and forces by the adjoint method. First, initial positions, initial velocities and whole forces were retrieved through the variational assimilation method for target TCs, and then the functional derivatives of the prediction distance of the model were calculated. It turns out that the prediction distance depends on initial latitudes, initial velocities of TC centers, and whole forces. But initial longitudes of TC centers, i.e., the variation of longitudes of TC centers has no impact on the prediction distance, the variation of latitudes of TC centers has slight impact, and variation of initial velocities of TC centers and of whole forces has big impact. Especially, the functional derivatives of the prediction distance of the model with respect to whole forces exhibit periodically distributed characteristics, and tend to zero. The results are useful for the improvement of TC track prediction accuracy.

作者: Jie Xiang Xi-jun Zhang Yi Luan

作者单位: Institute of Meteorology, PLA University of Science and Engineering, Nanjing 211101, China;Key Labor Institute of Meteorology, PLA University of Science and Engineering, Nanjing 211101, China;

会议类型: 国际会议

会议名称: The 5th International Conference on Nonlinear Mechanics(第五届国际非线性力学会议)

会议地点: 上海

会议语种:英文

页码: 1563-1567

在线出版日期: 2007-06-11（万方平台首次上网日期，不代表论文的发表时间）