Realistic representation of spatial heterogeneity and lateral land surface processes within and between modeling units in earth system models is important because of their implications to surface energy and water exchanges. The traditional approach of using regular grids as computational units in land surface models may lead to inadequate representation of subgrid heterogeneity and lateral movements of water, energy, and carbon fluxes. Here a subbasin-based framework is introduced in the Community Land Model (CLM) by representing local processes in each subbasin as a pseudo grid matrix with no significant modifications to the existing CLM modeling structure. Lateral routing of water within and between subbasins is simulated with the recently-developed physically based routing model, Model for Scale Adaptive River Transport (MOSART). The relative merits on scalability (i.e., ability to perform consistently across spatial resolutions) on runoff generation and streamflow simulations compared to the grid-based modeling framework are investigated in two topographically and climatically contrasting regions of  the U.S. Northwest and Midwest by performing simulations at 0.125°, 0.25°, 0.5°, and 1° spatial resolutions. Results from systematic comparisons conducted using statistical metrics calculated between each coarse resolution and the corresponding 0.125° resolution simulations demonstrated superior scalability in simulating runoff and streamflow for the subbasin-based over the grid-based framework. The source of runoff scalability is found to be related to the scalability of major meteorological and land surface parameters of runoff generation. Scalability advantages in streamflow simulation are driven by a combination of improved consistency in runoff generation and the routing processes across spatial resolutions. Hence, this study demonstrates the importance of spatial structure for multi-scale modeling of hydrological processes.