Investigations and Experiments of Variances Filtering Technology in the Ensemble Data Assimilation

Abstract

An experimental Ensemble Data Assimilation (YH‐EDA) system has been built with 10 ensemble members based on the operational YH4DVAR system. The system can provide flow‐dependent background‐error variances, which are superior to the operational ones both in structure and magnitude. However, the finite ensemble size implies a detrimental sampling noise for the variances estimation. To solve this problem, a spectral filtering technique is implemented to formulate a low‐passing filter. Taking into account the typical horizontal length scales of noise and signal, the filter can eliminate the sampling noise while extracting the signal of interest. In the ensemble variance filtering experimentations of the 9th typhoon “Jebi” in 2013, our results show that the 10‐members' filtered variances exhibit better performance than 30‐members's estimation. The successful implementation of the spectral filtering reduces the requirement of large ensemble size of Ensemble Data Assimilation system, which indicates that spectral filtering has become an important and necessary technology in EDA operational implementation.

Effect of Flow-Dependent Unbalanced Background Error Variances on Tropical Cyclone Forecasting

“…Bonavita, et al [39] showed that a relatively small set (e.g., 10 members) can adequately characterize the large-scale error structure of extratropical cyclones, and a larger set can model more refined features. The research of Liu, et al [40] found that the variance estimation using 10 members is very similar to the 30 members except for the relative reduction in noise, and there is no essential difference between the two. The number of EDA members in this article is also taken as 10.…”

The Impact of Length-Scale Variation When Diagnosing the Standard Deviations of Background Error in a 4D-Var System and Filtering Method Investigation

“…e YH4DVAR system is an operational deterministic 4D-Var system that uses the global spectral model as a constraint to impose a dynamic balance on the assimilation [17]. Considering that the sampling noise introduced by the randomization technique has a negative effect on the estimate, a spectral filtering method truncated at a fixed wavenumber in an operational setting solves the issue of sampling noise in the YH4DVAR system [18]. e spectral filtering method enables the removal of most of the sampling noise while preserving the standard deviations of interest.…”