Topography exerts a major control on land surface processes through its influence on atmospheric forcing, soil and vegetation properties, network topology and drainage area. Land surface spatial structure that captures spatial heterogeneity influenced by topography is expected to improve representation of land surface processes in land surface models. For example, land surface modeling using subbasins instead of regular grids as computational units has demonstrated improved scalability of simulated runoff and streamflow processes. A new land surface spatial structure is being developed by further dividing subbasins into subgrid units based on elevation, topographic slope and aspect to take advantage of the emergent patterns and scaling properties of atmospheric, hydrologic, and vegetation processes in land surface models. In this study, two methods (local and global) developed to derive topography-based subgrid land units are explored. We applied the two methods over the topographically contrasting regions of the Northwestern United States to evaluate the capability to capture topographic patterns. Results show that the local method is able to capture topographic patterns better than the global method. The local method is being applied to develop a global dataset of subgrid land units for the Community Land Model (CLM4.5).