The need for calibration limits the direct application of rainfall-runoff (RR) models in ungauged basins. In this paper, we present a conceptual framework for calibrating RR models using river hydraulic variables (river width or water surface elevation) that are observable by remote sensing. By integrating a RR model with at-a-station hydraulic geometry, using power functions to describe the cross-sectional hydraulic relationship at the basin outlet, the simulated river width or water surface elevation become the output of the integrated model. The objective of the calibration is then shifted to minimizing the difference between the simulated values and the satellite measurements of the hydraulic variables. Correspondingly, the calibration process is carried out by tuning the RR model and power function parameters simultaneously. The HYdrological MODel (HYMOD) RR model and Nondominated Sorting Genetic Algorithm II (NSGAII) multi-objective optimization scheme were selected based on the characteristics of satellite observations. A proof-of-concept experiment was carried out for the Pakse gauging station in the Mekong Basin, southwest Laos. Discharge estimates with acceptable accuracy were obtained by calibration, using ground measurements of either river width or water surface elevation at the Pakse gauging station, under designed average and low satellite observation frequencies. From the results of the experiment, a higher observational frequency was found to be preferable for making more reliable estimations. Using ground measurements with the possible error of satellite observations as calibration data, the maximum uncertainty was less than 20% of the mean daily discharge at Pakse station. This conceptual framework can provide a new tool for improving river discharge estimation in large ungauged basins. Copyright © 2011 John Wiley & Sons, Ltd.