River discharge measurements have proven invaluable to monitor the global water cycle, assess flood risk, and guide water resource management. However, there is a delay and overall decline in the availability of gauging data and stations are highly unevenly distributed globally. While not a substitute for river discharge measurement, remote sensing is a cost-effective technology to acquire information on river dynamics. The general approach has been to relate satellite observation to discharge measured in situ, which prevents its use for ungauged rivers. Alternatively, hydrological models are now available that can be used to estimate river discharge globally. While subject to greater errors and biases than measurements, model estimates of river discharge do expand the options for applying satellite-based discharge monitoring in ungauged rivers. Our aim was to test this approach. We used gridded surface water extent information from two sources: (1) Global Flood Detection System (GFDS) passive microwave data; and (2) MODIS optical data. The data were available for the common period of 2000–2014. The hydrological model used was the World-Wide Water (W3) model version 2, providing river discharge from 1980 to 2014. We designed and compared two methods to relate simulated storage and discharge to MODIS and GFDS surface water extent fraction for developing satellite gauging reaches (SGRs), and applied the best performing method to construct SGRs across the Amazon Basin. River discharge estimates from MODIS SGRs, GFDS SGRs, and the W3 model were evaluated with in situ river discharge measurements. The results showed SGRs can be successfully established over a large area using MODIS and GFDS water extent and modelled discharge, and used to estimate river discharge at both gauged and ungauged sites.