Soil erosion poses a great threat to the sustainability of the ecological environment and the harmonious development of human well-being. The revised universal soil loss equation (RUSLE) was used to quantify soil erosion in the Three-River Headwaters region (TRH), Qinghai, China from 2000 to 2015. The possible effects of an ecosystem restoration project on soil erosion were explored against the background of climatic changes in the study area. The model was validated with on-ground observations and showed a satisfactory performance, with a multiple correlation coefficient of 0.62 from the linear regression between the estimations and observations. The soil erosion modulus in 2010–2015 increased 6.2%, but decreased 1.2% compared with those in the periods of 2000–2005 and 2005–2010, respectively. Based on the method of overlay analysis, the interannual change of the estimated soil erosion was dominated by climate (about 64%), specifically by precipitation, rather than by vegetation coverage (about 34%). Despite some uncertainties in the model and data, this study quantified the relative contribution of ecological restoration under global climatic change; meanwhile the complexity, labor-intensiveness and long-range character of ecological restoration projects have to be recognized. On-ground observations over the long-term, further parameterization, and data inputs with higher quality are necessary and essential for decreasing the uncertainties in the estimations.