[6-4]

A meso Hybrid EnKF-4DVAR system based on the JMA nonhydrostatic model

K. Ito (University of the Ryukyus), M. Kunii, T. Kawabata, and K. Saito (Meteorological Research Institute)

This paper evaluates a potential of a meso hybrid ensemble Kalman filter (EnKF) and four-dimensional variational (4DVAR) data assimilation system for predicting severe weather events from the deterministic point of view. This hybrid system (HYBRID) is an adjoint-based 4DVAR system using a background error covariance matrix constructed from perturbations in a local ensemble transformed Kalman filter (LETKF) data assimilation system, both of which are based on the JMA non-hydrostatic model. An assimilation of pseudo single-observation located in the north of a tropical cyclone (TC) yields an analysis increment suited to dynamics of TC in HYBRID, while an analysis increment does not follow dynamics of TC with a 4DVAR system using so-called NMC method. Real-time data assimilation experiments exhibit that HYBRID provides a better initial condition than NMC-based 4DVAR for the intensity and track forecast of tropical cyclone (TC) Roki on September 2011 and the prediction of local heavy rainfall in Kyushu, Japan, on July 2012. In these cases, HYBRID provides a better initial condition than LETKF for a short-term prediction of high impact weather, while LETKF provides better initial condition for a relatively long-term prediction.