To achieve advanced watershed ecological management, policy-makers have struggled to predict ecological impacts for a long time. As a process of ecosystem analysis, ecological risk assessment (ERA) has been widely adopted to analyse the possibility of adverse ecological effects. ERA has developed from considering only a few indicators in an area to multiple sources and receptors in large-scope regions. However, the transfer of risk in large-scope regions caused by internal interaction has not been deeply analysed, especially in regions with complex internal interaction structures. This would lead to extensive management, where watershed-level policies may not be fit for some subregions, thus leading to limited management efficiency. In this study, we integrate an Input-Output (IO) model into the Relative Risk Model (RRM), and propose an IO model based ERA (IO-ERA) methodology, which would reveal the intensity of ecological risk caused by local sources, direct water flows and indirect transfers. An IO-ERA is conducted in Taihu Lake watershed as a case study, in which we could demonstrate that IO-ERA is capable of providing advanced insights of risk analysis in large-scope regions. The outcome of IO-ERA would support watershed administration to transfer from single standard regulation to diverse, dynamic and lean ecological risk management.