Changes in land use have disrupted natural nutrient cycles, leading to declines in water quality in freshwater systems worldwide. Restoration approaches that consider the interface between terrestrial and aquatic systems are thus key to build resilience in water ecosystems. However, despite billions of dollars spent in water quality monitoring programs, decision makers are confronted with inadequate information to prioritize landscape restoration that benefits freshwater systems. Here, we present an adaptable framework to integrate remote sensing, hydrological, land use and socio ecological data to prioritize the restoration of catchments that affect lake water quality at global scales. This framework was applied to different case studies across North America, Oceania and South America. For each region, we used novel large-scale datasets to identify areas with the highest sources of nutrient runoff and predict loads of TP and TN entering lakes. These models were combined with zero-dimensional models, which are easy to implement, to mechanistically predict lake-wide TP and TN concentrations based on inflow nutrients, lake residence time and other physical characteristics. We used this information to prioritize the selection of sites where restoration could reduce the impact of sources associated with land cover change and other nutrient sources (e.g. atmospheric deposition, farm fertilizers, manure) under different land use scenarios. Our results show that models based on large scale datasets had a high model estimation performance (R2 > 0.75) of total loads and yields sources. We also show how the framework can be applied at multiple spatial scales, from single lakes to entire catchments and regions, and how uncertainty in potential restoration outcomes increases in areas with poor water quality monitoring data. By applying our framework to prioritise for lake restoration, we show the importance of accounting for links between terrestrial and freshwater ecosystems to maximize benefits of restoration across the land to water interface, particularly in data poor regions.