Distributed hydrological models usually have cell-size dependence: model parameters must be readjusted when the digital elevation model (DEM) resolution is changed. As described herein, we propose a simple scaling algorithm forTOPMODEL, a conceptual rainfall-runoff model of small river basins. The topographic index (TI) plays a key role in TOPMODEL. The DEM resolution strongly affects TI. Therefore, we cannot use TI that has been estimated directly with low-resolution DEM. In the proposed model with coarse spatial resolutions, we use the cumulative probability distribution of TI of high-resolution DEM, which has power law scaling. For each aggregated grid, the cumulative probability distribution of TI is well approximated with a two-parameter function. It is used to evaluate the variable contributing area of surface flow. The proposed TOPMODEL, combined with the Muskingum-Cunge routing method and the aggregated coarse scale river network, which preserves the geomorphologic features of the original fine-scale one with sufficient accuracy, was applied to the Hayakawa River (a branch of the Fujikawa River). The discharge simulation result varies only slightly even if a very coarse scale is adopted. We also examine how simulation results change when the non-uniformity of various inputs is considered.