A coupled modeling approach was proposed for quantitatively estimating the watershed non-point source load property response to future climate changes. The Lianjiang Watershed located in Anhui Province, China was set as case study and the fluxes and source apportionments of total nitrogen pollution in the water were modeled. The LARS-WG weather generator model was used to downscale the results of HADCM3 climate model to generate synthetic daily weather data under 3 typical greenhouse gas emission scenarios (A1B, A2, B1) in different future periods. The regional nutrient management model (ReNuMa) was employed to model the total nitrogen source apportionments of Lianjiang River and estimate the future pollution fluxes and contribution proportion under changed climate conditions, which were further compared to current status to estimate the response of watershed non-point source pollution to climate change. The results showed that future watershed nitrogen pollution fluxes would have an overall increasing trend that mainly focused in autumn and winter. Most additional pollution would come from farmland and natural land and enter into water through runoff and shallow ground flow process, which should be of great concern in local management.