Non-point source pollution is one of the important sources of water pollution, revealing the spatio-temporal distribution characteristics of non-point source pollution load and screening and deployment of best management practices (BMPs) is vital for the efficient treatment of water pollution. In this study, the spatial distribution characteristics of total nitrogen (TN) and total phosphorus (TP) load were quantitatively simulated using the SWAT model in the upper Beiyun River in 2019, and evaluate the effect of different BMPs on TN and TP reduction in critical source areas of non-point source pollution. The results showed that: (1) In 2019, the TN and TP loads generated in the upper Beiyun River were 119336.50 kg and 4473.92 kg respectively,showing a spatial distribution characteristics of high in the southeast and low in the northwest, and the TN and TP load mainly come from urban land, cultivated land, and orchard. (2) A total of 8 sub-watersheds were identified as critical source areas, accounting for 5.82% of the total watershed area, and the TN and TP loads generated accounted for 21.49% and 29.61% of the total watershed area respectively. (3) The reduction rates of TN and TP in 1/5 area ratio filter strip were the highest, 42.03% and 46.02% respectively; The reduction rates of TN and TP in 2km grassed waterway were the highest, which were 13.59% and 30.09% respectively; As the agricultural land area within the key source area is small (7.77%), the pollutant reduction effect of the chemical fertilizer reduction measures is low. These study results can provide scientific reference for quantitative identification of non-point source pollution load and water pollution control.