The Study Of Sample Optimization In The Ensemble-variation Hybrid Assimilation

The ensemble-variable hybrid assimilation method has been considered as an important development trend of data assimilation.In order to introduce flow-dependent background error covariance effectively and solve the difficulties caused by the huge amount of ensemble forecasts for business applications,the research tried to introduce the optimal selected historical forecasting samples and time-lagged samples as substitutions in the ensemble-variance hybrid data assimilation system to provide a solution for the practical application of the hybrid method.The paper explored the rationality of the selected historical samples at first.A serious of single observation tests based on different sample combinations were carried out.The precipitation scores and root-mean-square errors(RMSEs)of experiments based on different samples were verified and a heavy rainfall case was further analyzed and diagnosed.The main works are as follows:(1)The historical forecasting samples close to the weather conditions at the assimilation moment are selected effectively by using the empirical orthogonal functions to abstract the dynamical features from the samples and match them with the background.The single observation tests show that the wind increasement structure of selected historical samples experiment is boarder and similar to the increments of the same-size ensemble forecasting samples experiment in general.The sample-combined experiment has increment features of both two samples.The unreasonable increments are displayed in the experiments based on the limited ensemble forecasting samples.(2)The 19-day continuous cycling assimilation and forecasts indicate that samplecombined experiments perform close to the same-size ensemble forecasting sample experiment,but the former consumes only about 40% computing resources of the latter.Compared with the experiments based on the limited ensemble forecasting samples,the sample-combined experiment has better performance.It proves the historical forecasting samples could alleviate the problem of insufficient samples and improve the effects of hybrid assimilation and forecasts.The diagnosis of rainfall case also show that the sample-combined experiment improves the simulation of the vertical wind and water vapor in the center of heavy precipitation.Thus,the precipitation results with better intensity and location are obtained.(3)To further save the computational resources as well as keep the advantage of flow dependence,this paper explores the hybrid experiments with the combination of selected historical samples and time-lagged samples.The single observation tests indicate that using the combination of selected historical samples and time-lagged samples,the sample error caused by limited samples is mitigated and flow-dependent background error covariance could be introduced in hybrid assimilation system.Cycling assimilation and forecasts for a week show that the hybrid experiment based on combined samples has the smallest RMSEs,and performs better than other hybrid experiments which only use the time-lagged samples.Its precipitation score performs best overall,especially in the assessment of moderate and heavy level of the precipitation.The intensity and location of heavy precipitation center could be predicted well and the over-estimated problem is alleviated in the sample-combined hybrid experiment.