Improvement of GRAPES_3Dvar with a New Multi-Scale Filtering and Its Application in Heavy Rain Forecasting
To improve the effect of the numerical simulation and forecasting of heavy rains, it is very important to introduce the meso- and small-scale information in the assimilation analysis of initial values. For enhancing the introduction of meso- and small-scale information in the regional GRAPES 3Dvar system, climatic background error sample was used in this paper to estimate the level covariate correlation scale of variable error, and then statistical result fitting was performed with recursive filter of the feature scales of three different levels, thus replacing the original single-scaled recursive filtering. The new scheme was used to assimilate and forecast the rainstorms in the Jianghan Plain during 1-2 June, 2015, and the research results showed that power spectrum attenuation in the new scheme is slower. Through single point test and field analysis, we found that the new assimilation scheme introduces more meso-α scale information. In the report of the rainstorm, it was found that, with the adoption of the new scheme, the moisture field, divergence field and vorticity field are much closer to the observation values when measured in the analysis field and forecasting field. So the precipitation forecasting skill is improved obviously. By analyzing the energy spectrum, it was learned that the new scheme could reflect more meso-α scale information and the new scheme has positive effect on the forecasting of rainstorms in the Jianghan Plain area. In meso-α scale, there are some favorable factors for rainfall, such as lower convergence, upper divergence and increased humidity. Based on individual cases of rainstorm, batch experiments for 16 days were completed, and the result showed that the new scheme could improve precipitation forecasting skill, which is consistent with the results of cases study.