We simulated hydrological and biogeochemical responses to logging in a forested watershed to determine the vulnerability and/or resiliency of the forest ecosystems in the Lake Shumarinai Basin in northern Hokkaido, Japan. We used a biogeochemical model (PnET-CN) and a rainfall-runoff model (HYCYMODEL) to predict ecosystem responses. The PnET-CN model simulated well the observed NO3 (-) concentrations in streamwater, particularly at high concentrations during snowmelt; however, the model could not simulate small increases in NO3 (-) during the summer. By considering hydrological processes within the watershed and combining the model with the HYCYMODEL (PnET + HYCYMODEL), the seasonality of streamwater NO3 (-) concentrations was better simulated. Using these models, the long-term effects of logging were simulated for coniferous, deciduous, and mixed forests. NO3 (-) concentrations in streamwater increased in response to the logging disturbance in both coniferous and deciduous forests. In the coniferous forest, NO3 (-) concentrations reached a maximum 10 years after logging, and high concentrations persisted for 30 years. In contrast, NO3 (-) concentrations in the deciduous forest reached a maximum within 3-4 years and recovered to pre-disturbance levels after 15 years. We also used the models to determine the effects of different sizes and types (coniferous, deciduous, and mixed forest) of logging areas on Lake Shumarinai. The model results indicated that large areas of cutting require more than 100 years for complete lake recovery. Whereas the annual discharge to the lake minimally increased, the annual NO3 (-) load greatly increased. Our simulation results elucidate the vulnerability and resiliency of forest ecosystems and provide valuable information for ecosystem management.