Spatial structure of weights in the Local Ensemble Transform Kalman Filter: A case with an intermediate AGCM
The Local Ensemble Transform Kalman Filter (LETKF) updates the ensemble mean and perturbations by a linear combination of background ensemble members. The LETKF computes the linear-combination weights at every grid point by assimilating observations within a localization cut-off radius. This study aims to investigate the spatial structure of weights in the LETKF using an intermediate global atmospheric model known as the SPEEDY model.
We examined sensitivities of the spatial structure of weights to the ensemble size, localization scale and observing network. A larger localization scale results in spatially smoother patterns of weights. Finer structures appear in densely observed regions. We also show that the weight interpolation technique of Yang et al. (2009, QJRMS) works well with the SPEEDY-LETKF. The weight interpolation is more beneficial for experiments with longer localization scales with which the weights have spatially smoother patterns. This poster will include the most recent progress up to the time of the symposium.
Dr. Shunji Kotsuki RIKEN Center for Computational Science