Based on the observed data on runoff plots of Anjiagou watershed in 2014-2016 in Dingxi City, Gansu Province, the effects of vegetation types and rainfall regimes on slope erosion and sediment yield were studied. The results showed that:the erosive rainfall could be classified into three patterns:pattern A was an aggregation of rainfall events with short duration and medium to high intensity, pattern B with moderate rainfall and low intensity, and pattern C with heavy rainfall and low intensity. The rainfall frequency of runoff and sediment generation was pattern B > pattern C > pattern A. The runoff coefficients of pine forest (6.915%~9.379%), wheat field (5.838%~9.034%) and alfalfa land (6.610%~9.671%) were 1.7~2.5 times than that of ice grassland (2.724%~5.246%) and 2~3 times of seabuckthorn forest (2.296%~3.863%) respectively. The runoff coefficient of pattern A was greater than those of pattern B and C. Pattern B and C were the main rainfall patterns of runoff generation. Annual average amount of soil loss was as followed:wheat field (1.478~3.478 t/hm²)>alfalfa land (0.558~2.079 t/hm²)>pine forest (0.459~0.887 t/hm²)>ice grassland (0.097~0.253 t/hm²)>seabuckthorn forest (0.012~0.038 t/hm²). The amount of soil loss of wheat fields was 3~5 times than that of pine forest, and 14 times of ice grassland, reached 123, 130 and 92 times of that of seabuckthorn forest at the slope gradient of 10°, 15° and 20°, respectively. The contribution rate of pattern C to sediment yield of pine forest was the largest, pattern B was the second, and pattern A was the smallest. The contribution rates of pattern A to sediment yield of seabuckthorn forest reached 79%~86%. So high intensity and short duration rainfall was the main rainfall pattern causing soil loss in seabuckthorn forest. The contribution rates of pattern A to sediment yield increased with the increases of slope gradients in wheat field and alfalfa land. At the slope gradient of 20°, rainfall with medium to high rain intensities was the main rainfall pattern causing soil loss. There was a very significantly positive linear correlation (P<0.01) between the runoff depth and the amount of soil loss in wheat field, alfalfa land, pine forest and ice grassland, and the increment speed of soil loss was:wheat fields > alfalfa land > pine forest > ice grassland. At the same time, the increment speed of sediment yield was accelerated as the slope gradients increased. These findings could provide theoretical basis for building the soil erosion prediction model and important practical significance to scientifically guide the allocation of soil and water conservation measures on slope land.