| As an efficient approach to improve the accuracy of numerical weather prediction (NWP), Four-dimensional Adjoint Variational Assimilation is focused on by many national meteorologists. In this paper the technique of adjoint codes which is a basic means of constructing adjoint model and the selection of weight, scaling and descend arithmetic are introduced amply. Then the MM5 Adjoint-model Assimilation System that is based on variational method with adjoint codes technique is introduced. Model terrain adjustment is carried out by means of the MM5 adjoint-model assimilation system (AMAS) using the conventional observation data and non-conventional data cloud-derived winds. Two different initially-estimated terrains (from stochastic intensification and weakening) are given and corrected by use of the AMAS, which indicates much the same assimilation for the terrains differing from each other. The modified terrains are more consistent with the observation data. For the terrain parameter of identical errors, the adjoint model with inclusion of non-conventional cloud-derived winds in will improve even more the AMAS modification to the terrain parameter. As shown in the subsequent sensitivity runs, an increase in the proportion of a certain level wind in different regions will adjust the terrains in effect. To verify effects of modified terrain on NWP, experiment is conducted on a large-scale rainstorm over the Changjiang-Huai basin on 23-24 July, 2002, indicating that the terrain modified by conventional data and especially the one adjusted by cloud-derived winds used as the MM5 surface feature will improve prediction than the unmodified terrain. At the same time, Ensemble Prediction technique is used in terrains modification. It was found that a stable prediction can be gotten and the chanciness of a single case can be avoided after the ensemble averaging terrain is used. A new feasible approach about the widely application of AMAS is offered in this paper.
|
PDF Full Text Request