A mathematical model was developed describing an onsite treatment of floodplain filtration for the improvement of river water quality. This technique consists of spraying the contaminated river water on the floodplain, resulting in filtration through the floodplain soil. The mathematical model's parameters were determined in the experiments and preliminary studies, and numerical techniques were used in the model analysis. The model was validated against experimental data, and a sensitivity analysis was performed in order to understand floodplain filtration. Results showed that the model predicted favorably the change in water quality during floodplain filtration, and that the floodplain filtration technique was an effective way of removing both BOD (Biochemical Oxygen Demand) and NO3 simultaneously from the river waters. It was found that most of the BOD was degraded within a few decimeters from the soil surface, and that there occurred significant oxygen infiltration from the atmosphere through the soil gas phase near the soil surface. It was also found that the N2O emission and the NO3 leaching increased with the increase in the influent NO3 level, and that as the filtration rate increased the performance of this technique improved without significant adverse effect. The N2O emission in the floodplain filtration for most of the river waters in Korea was expected not to exceed 0.5 mL/day/m2. Key words: floodplain filtration, BOD removal, denitrification, competitive Michaelis-Menten model.