The previously developed PADDY-Large model was improved by the coupling of geospatial information about watershed properties (Digital National Land Information and digital cultivated soil maps of Japan). The improved model was used for simulating the spatiotemporal variations of pesticide concentrations in rice paddies located in two river basins. We also developed a method for clarifying land use in sub-basins and for allocation of paddy fields to the nearest 50-m-long river segment for the model calculation. We validated the model by monitoring paddy pesticide concentrations in river water in basins under rice cultivation in southern Ibaraki Prefecture, Japan, and comparing the measured concentrations with concentrations simulated by the model. The measured concentrations of the herbicide mefenacet in river water peaked 1–2 weeks after transplanting (mid-May) and then decreased rapidly in early June. The period during which mefenacet was detected corresponded to the application timing of mefenacet. The improved PADDY-Large model accurately simulated changes in measured concentrations of mefenacet at the sampling points. The spatial distribution of mefenacet concentrations was related to the distribution of paddy fields. River segments with relatively large areas of paddy fields in upstream areas of the sub-basins had relatively high mefenacet concentrations.