Using Climate HRRRE Ensemble Perturbations for Improving GSI Hybrid 3D-EnVar Surface Analysis

The Gridpoint Statistical Interpolation (GSI) Data Assimilation System is the analysis component of many NOAA/NCEP numerical weather forecast systems. It can analyze many types of the observations with 3DVAR or hybrid ensemble variational (EnVar) analysis method. In GSI 3DVAR analysis, the background error covariance (BE) is setup to distribute the analysis increment homogeneously and isotropically, which can be problem for the analysis in the inhomogeneous and anisotropic environments, such as observations in a front zone or surface observations in complex terrain or near the coastal line area. The hybrid EnVar analysis introduces the follow-dependence features through ensemble perturbations, which greatly improved GSI analysis for large scale system. But GSI hybrid analysis with GFS EnKF ensemble forecast was found to have minor benefits to PBL and surface analysis in RAP/HRRR experiments. The possible reasons can be the coarse resolution of the GFS EnKF ensemble and rank deficince of the limited ensemble members.

GSD RAP/HRRR group is running a real-time HRRR ensemble (HRRRE) testing system, which can provide GSI hybrid analysis with high resolution ensemble forecast. But the number of ensemble is still very limited because of the limitation of the computer resources. To overcome the rank deficiency of HRRRE, we collected several months of the HRRRE forecast to generate a set of climate high resolution ensemble perturbations. From previous investigation, we found that many local constant features, such as terrain or land surface types, are well represented in those climate perturbations.

In this study, we will further study if surface analysis in complex terrain area can be improved through applying those climate perturbations in the GSI hybrid analysis. Retrospective experiment and pure analysis will be conducted and results from those tests will be discussed during the conference.