To account for most of the important sources and to minimize the impact of uncertainties of boundary conditions, a rather large domain is usually needed in Eulerian air quality modeling. For selected areas such as in the vicinities of large point sources or urban centers, smaller grid sizes are needed to achieve acceptable resolution. Consequently, a combination of different grid sizes is desirable for an optimal balance in resolution and computational cost. Pleim et al. had established a one‐way nesting process in which information from the coarse grid model is passed to the fine grid subdomain model, but not in the opposite direction. A new two‐way nesting process based on the concept of error correction is proposed. This process can be self‐consistent compared to the two grid systems. We have made comparison between the two nesting processes using the Bott advection scheme in numerical experiments and evaluated their performance in air quality modeling. Wherever the two‐way nesting process is used, we have relatively better results compared to the one‐way nested subdomain model.