为准确揭示密云水库流域径流变化的物理成因，基于降水、径流非平稳特征诊断结果，采用SWAT（soil and water assessment tool）水文模型，基于Budyko理论的弹性系数法和水量平衡法揭示该流域径流变化原因；对比验证各模型归因结果差异，进一步探究归因方法和基准期选择对流域径流变化归因结果的影响。结果表明：基于Budyko理论的水量平衡方法更适用于水利工程众多、跨流域调水活动频繁的密云水库流域径流变化归因计算；利用该方法进行分析，当基准期长度达到16年及以上时，流域径流变化归因分析结果逐渐趋于稳定，而基准期选取过短会高估人类活动对径流的减水效应；此外，基准期选取靠近研究初始时段时，计算得到人类活动的减水效应相对较大。综合考虑上述各因素的影响，认为人类活动是密云水库流域径流变化的主导因素；在变化期Ⅰ，其减水效应在?128.32% ~ ?119.56%波动，在变化期Ⅱ，其对径流的削减作用为?75.58% ~ ?70.70%。研究结果有助于提升对密云水库流域径流复杂演变规律和物理成因的科学认识。

For the non-stationary hydrological time series, it is of great importance to explore its formation with attribution analysis to understand the hydrological process and solve the problem of water resources. Under the influence of both climate change and human activity, hydrological process in the Miyun Reservoir basin shows significantly non-stationary characteristics, while there are great differences in the results of attribution in this basin. In order to accurately reveal the causes of runoff changes in the Miyun Reservoir basin, the whole period (1960-2019) was divided into three parts: 1960-1979 for the baseline period, 1980-1998 for the change period Ⅰ, and 1999-2019 for the change period Ⅱ based on the detection result of the non-stationary characteristics of precipitation-runoff. The SWAT method, the elastic coefficient method, and the water balance method based on Budyko theory were used for attribution, and then model suitable for attribution in the Miyun Reservoir basin was selected according to the attribution results and basin characteristics. Based on the selected model, length and position of baseline period were changed to further explore the influence of baseline period on the attribution result.It was shown that attributions of elastic coefficient method and water balance method based on Budyko were similar, while the attribution of SWAT method was quite different. In the change period Ⅰ, contribution of human activities to runoff changes show as SWAT method < elastic coefficient method < water balance method. And in change period Ⅱ, results of the two methods on Budyko both showed that human activity was the main reason for the decrease of runoff, while result of SWAT showed that the impact of climate change on runoff far exceeded the impact of human activities, and became the dominant factor affecting the runoff. Cross-validation of multi-method attributions showed that the water balance method based on Budyko theory was more suitable for the attribution of runoff changes due to the construction of hydraulic projects and frequent inter-basin water transfer in the Miyun Reservoir basin. Based on the selected model, it was found that the results of attribution tended to be stable as the length of the baseline period reached 16 years. Reduction effect of human activities on runoff would be overestimated while the baseline period was shorter than 16 years. And the calculated water reduction effect of human activities was larger when the baseline period was closer to the initial position of the whole period. Considering both the effects of model and baseline period, it was concluded that human activities were the dominant factor in the runoff changes in Miyun Reservoir basin; its reduction effect on runoff fluctuated from ?128.32% to ?119.56% during the change period Ⅰ, and was about ?75.58% to ?70.70% during the change period Ⅱ. These results are helpful to improve the scientific understanding of the complex changes of the runoff and its attribution in Miyun Reservoir basin.