Nonpoint source pollution (NPS) has been known as runoff or diffuse pollution. It is still difficult to real-time monitor and simulate NPS, due to the significantly complex influencing factors. Therefore, the assessment theory of landscape ecological risk can be expected to identify the regional pollution risk from the perspective of landscape pattern and ecological process. In this study, NPS risk assessment was established using the “source-sink“ landscape pattern theory. The first step was to identify “source-sink“ landscape. Cultivated, residential, transportation and commercial land were identified “source landscape", while woodland, green land, water area and wetland were identified “sink“ landscape. The second step was the landscape spatial load comparison index (LCI) correction. The output coefficient and empirical model were combined to determine the “source-sink” landscape pollutant output/interception coefficient. Then, the slope, altitude and distance from river were selected as the pollutions migration factors to correct LCI values. The final step was to evaluate the NPS pollution risk. The risk was calculated by LCI, slope factor and cost distance. The risks were divided into three grades, according to the natural breakpoint. Taking Shenzhen River-bay Basin as an example, the risk of NPS pollution was assessed using the “source-sink“ landscape pattern theory. The numerical and spatial correlation analysis was conducted with the field monitoring pollutant data to verify the accuracy of the improved model. The research results showed that: (1) No.7, 9, 10, 12 and 13 watersheds were the high risk of NPS in Shenzhen River-Bay watershed in the whole Buji River basin and the lower reaches of Futian River basin. The low-risk areas were except No. 1 and 6 watersheds in the north of study area and No. 16, 17 and 18 watersheds in the east of study area. The rest 8 watersheds were all medium risk areas. NPS risk level was closely related to the “source-sink” landscape area ratio and landscape spatial pollutants load index, but the topographic factor shared the little influence on the basin. The residential land and transportation land showed the largest contribution rate to NPS risk, while the forest land and water area had outstanding effect on pollutant retention. (2) Both average concentrations of total nitrogen (TN) and total phosphorus (TP) in Shenzhen River-Bay basin were fully met the IV standard level of surface water quality. The average concentration exceedance rate of TN and TP were 12.99% and 11.89%, respectively, in rainy season, and 8.61% and 9.44% in the non-rainy reason. The maximum concentrations of pollutions occurred in the rainy season. The pollutants spatial analysis showed that Shenzhen River basin pollution risk was more serious than Shenzhen Bay basin, while the downstream was than in the upstream, and the higher in the coastal area than in the non-coastal area. The pollutants monitoring data showed that the high-risk area of NPS were No.5, 7, 9, 10, 12, 13 watersheds. Except No.5 watershed, the high-risk watersheds were consistent between monitoring and assessment. (3) The NPS risk assessment index using the “source-sink” landscape theory shared a significant correlation with the actual monitoring risk of TN and TP, indicating a highly spatial aggregation. The NPS risk assessment index can also provide the scientific and accurate reference to assess the ecological risk of watershed in regional scale.