In order to represent flood and inundation events with a distributed rainfall-runoff-inundation model, it is necessary to appropriately model river cross-sections along all rivers including small rivers. However, in most cases, cross-section data of small rivers are not available. In this study, we propose a new method to estimate cross-sections of small rivers. Through investigation of conventional estimation method by power law regression with downstream data, we revealed that the conventional method may underestimate crosssections of small rivers. Alternatively, we propose to use a linear regression with a certain intercept, so that we can estimate the lower bound of cross-section of small rivers for preventing underestimation. Finally, by comparing the results of RRI model with different cross-section settings (real section, conventional estimation method, and our new estimation method), we examined the applicability of our new method, and demonstrated that the new method reproduces a similar result with the case based on the real sections.