Shunji Kotsuki;
Steven J. Greybush;
Takemasa Miyoshi
RIKEN Advanced Institute for Computational Science, Kobe, Japan
The 8th EnKF Data Assimilation Workshop
On the assimilation order of the serial ensemble Kalman filter: A study with the Lorenz-96 model
Poster:
201805_enkf_kotsuki_02.pdf
We usually assume that the assimilation order of the serial ensemble Kalman filter (EnKF) has no significant impact on analysis accuracy. However, Nerger (2015) derived that different assimilation orders result in different analysis states and error covariances if covariance localization is applied in the observation space. This study explores if we can optimize the assimilation order for better analysis accuracy. We examine several assimilation orders with the serial ensemble square root filter using the Lorenz-96 40-variable model. The results show that the small difference due to different assimilation orders could eventually result in a significant difference in analysis accuracy. The analysis is improved significantly by ordering observations from worse to better impacts using the ensemble forecast sensitivity to observations (EFSO), which estimates how much each observation reduces or increases the forecast error. The error reduction by the serial assimilation process of the EFSO-based ordering is similar to that by the experimentally-found best sampled assimilation order.
We usually assume that the assimilation order of the serial ensemble Kalman filter (EnKF) has no significant impact on analysis accuracy. However, Nerger (2015) derived that different assimilation orders result in different analysis states and error covariances if covariance localization is applied in the observation space. This study explores if we can optimize the assimilation order for better analysis accuracy. We examine several assimilation orders with the serial ensemble square root filter using the Lorenz-96 40-variable model. The results show that the small difference due to different assimilation orders could eventually result in a significant difference in analysis accuracy. The analysis is improved significantly by ordering observations from worse to better impacts using the ensemble forecast sensitivity to observations (EFSO), which estimates how much each observation reduces or increases the forecast error. The error reduction by the serial assimilation process of the EFSO-based ordering is similar to that by the experimentally-found best sampled assimilation order.
Thanks to McGill's Atmospheric and Oceanic Sciences department for hosting this web-site.