4-dimensional variational data assimilation study for tropical cyclone development in early stages.

摘要

Inaccurate initial conditions are responsible for most tropical cyclone forecast failures. An incorrect initial condition for a developing system may not lead to a prediction of its true development, while an incorrect initial condition for a non-developing system may produce an erroneous strong hurricane. Therefore, great efforts are needed to reduce the initial uncertainties in the analyses over the tropics.; In this dissertation, we adopted the 4-dimensional variational (4DVAR) data assimilation approach to improve the quality of the initial conditions. Time distributed observations were assimilated in a dynamically consistent manner to generate initial conditions under the optimal control theory. The Fifth-Generation NCAR/Penn State Mesoscale Model (MM5) and its adjoint model MM5ADJ were utilized in this study to conduct 4DVAR data assimilation experiments with different input data combinations. Two developing cases and two non-developing cases were chosen to test the effectiveness of the 4DVAR data assimilation methods.; The 4DVAR experiments with two consecutive operational analyses resulted in generally improved initial conditions. The significant improvements achieved in some experiments indicate that the 4DVAR data assimilation process should not be explained as a simple combination of the input data. The consistency constrained by the dynamical model requires an optimal selection of information from all of the input data, which usually leads to a better initial condition with reduced uncertainties. A set of asynoptic data, namely high density flight-level reconnaissance observations, was assimilated. The result demonstrates encouraging prediction improvement.; Two experiments with a bogus vortex specification were carried out. It seems that the technique of incorporating a bogus vortex in the initial condition is not appropriate for tropical disturbances in their early stages. In some complicated synoptic situations where several weather systems exist and strongly interact with each other, we have to specify all of them, not only the primary vortex, accurately, when the bogus technique is applied.; Diagnostic studies indicate that changes of the external upward forcing and the moisture content made by the 4DVAR data assimilations are most responsible for the better intensity predictions.