To evaluate the effectiveness of conservation tillage in mitigating soil erosion at a regional scale in the black soil zone of Northeast China, a SWAT model was developed for the Hulan River Basin in Heilongjiang Province. Eight simulation scenarios were designed to investigate the effects of none/reduced tillage seeding, straw returning, subsoiling and their combinations on rainfed cropland soil erosion. The results indicate that the SWAT model performs well in simulating runoff and sediment yield during both the calibration and validation periods. The Nash-Sutcliffe efficiency (NSE) values for runoff were 0.90 and 0.97, with percentage bias (PBIAS) values of 9.59% and -12.63% during the calibration and validation periods, respectively. The NSE values for sediment yield were 0.75 and 0.64, with PBIAS values of -0.63% and -15.43% during the calibration and validation periods, respectively. During the crop growing season from May to September, the soil erosion in rainfed cropland first increased and then decreased, with higher erosion in July and August and lower erosion in May. In terms of erosion control effectiveness, straw returning showed the best performance, followed by none/reduced tillage seeding, whereas subsoiling was less effective and even intensified soil erosion in April, May and September. Therefore, straw returning is recommended as the priority conservation tillage practice, followed by none/reduced tillage seeding. When multiple practices are integrated, the combination of straw returning and none/reduced tillage seeding is recommended.