Non-point source pollution is a major water ecological environment problem that influences the water environment in the Haihe River Basin. Mechanisms for controlling non-point pollution by optimizing landscape patterns at the watershed scale has been a popular issue. In this study, the risk patterns of non-point source pollution was established using minimum cumulative resistance model, which combines landscape elements and external factors that influence landscape processes. Risk pattern verification was performed by correlation analysis between resistance values and measured pollutant data in the Haihe River Basin. Watershed non-point pollution risk was identified and simulation of landscape patterns were optimized based on the verified non-point source pollution risk patterns. The results showed that in the Haihe River Basin, more than 40% of the areas are at extremely high or high risk level, and are mostly concentrated in the southern plain area and mountain valleys along the river. The landscape pattern was optimized by setting the vegetation buffer zone along the riverside. The simulation analysis results indicatedthat in different sub-basins, strengtheningcontrols onhuman activities to reduce the pollutant output in theareas within 600-1600 meters on both sides of the river can significantly reduce the risk of non-point source pollution. In plain areas of the Haihe River Basin, the high risk levels of non-point source pollutioncan be reduced by 50% by setting the vegetation buffer zone to 300-400 meters and a 700-800 meter vegetation buffer zone can achieve the best effect. In mountainous areas with relatively low risk level, a 400-500 metervegetation buffer zone can reduce non-point source pollution risk most effectively. The results of this study provide a scientific reference for optimizing landscape patterns to control non-point source pollution risk in the Haihe River Basin.