Finding the appropriate spatial resolution in modeling is a serious challenge at the beginning of every modeling project. The paper presents a methodology to adjust the spatial geography to the resolution of a network. Based on the quadtree algorithm, raster cells are generated that are dynamic in size. Smaller raster cells are used in urban areas and larger raster cells are used in low-density, rural areas. Trip tables of a travel demand model for the State of Georgia are disaggregated to this new zone system of raster cells, and assignment results validate significantly better than when using the original zone system.