The local ensemble transform Kalman filter (LETKF) treats each grid point independently, without restrictions to the choice of the model grid system. In a two-way nested model configuration, a smaller domain model with finer grids is embedded within a larger domain model with coarser grids. Since the two-way nested time integration makes these two models consistent, the two models can be regarded as a single model with a heterogeneous grid system. In this study, a single LETKF system is applied to such a heterogeneous grid system using the Weather Research and Forecasting (WRF) model. Experiments are performed in the Northwestern Pacific, with a particular focus on Typhoon Sinlaku (2008). With this technique, resolving capability is enhanced effectively in the vicinity of Typhoon.