Local and associated regional factors commonly drive changes in ecosystem services (ES). Yet, some current studies focus on considering local drivers based on statistical methods, which may ignore the impacts of drivers in associated regions such as upstream-downstream, and consequently lead to deviation in attribution analysis and management of variations in local ES. To overcome this challenge, this study proposes an attribution analysis method of changes in ecosystem services based on emergy and partial differentiation equation (PDE), and applies it at River Basin scale to quantitively measure cross-regional causation. Taking the Yellow River Basin (YRB), China, as a case, this study identifies and assesses the specific contribution of drivers to the variations in ES from 2000 to 2020. The results show that the ES in the entire YRB increased 30%, and the ES in the upper, middle and lower reaches increased 30%, 36% and 3% respectively. All the assessed ecosystem services increased in the past two decades, among which three services including air purification (182%), climate regulation (145%) and water purification (79%) have the largest increase ratio. Natural drivers R (such as evapotranspiration, precipitation, etc.), human drivers S (land use change), significance driver τ (the significance degree of human attention to ES improvement) commonly and disproportionately drive the ES dynamics. The contribution ratios of the three above drivers to the increase in ES are 0.7:8:1, 1:9:0.8, 0.2:7:1.6 and (−7):11:4 for the entire basin, the upper, middle and lower reaches respectively. The decrease in upstream water use results in midstream lake and wetland ES increase 1.44E+18 and 7.49E+17sej during 2000–2020, accounting for 55% and 97% of lake and wetland ES in 2020 respectively, which decrease the decline ratio of midstream lake and wetland ES caused by midstream local land use transfer from water bodies to built-up land or agricultural land. This is a significant method development of the proposed attribution method to quantitatively solving causation in upstream-midstream-downstream of a River Basin. Meanwhile, this study can provide guidance for collaborative and integrated management and conservation of ecosystem services in the upper, middle and lower reaches of watersheds or other related regions.