The development of radar altimetry over rivers along the past 20 years has shown a strong potential to provide hydrologist with valuable information on river water level dynamics. The development of new sensors and satellite mission concepts such as spatial interferometry or temporal interferometry opens midterm (10-­20 years) perspectives for the spatialized measurement of river surface variables such as width W, water level Z, surface slope Is and surface velocity Vs. Although not as accurate as in situ measurements can be, satellite measurements would ensure exhaustive and homogeneous global coverages, and provide repetitive and near real time information on these variables. Therefore a key question arises for hydrologists : assuming the availability of satellite measured river surface variables, with known uncertainty levels, would it be possible to estimate river discharge without any in situ measurement, and what would be the resulting uncertainty on discharge estimate? We developed a method to estimate river bottom parameters (river bottom elevation Zb, bottom slope Ib, velocity profile coefficient α and Manning coefficient n) from a time series of synchronous surface variable measurements (W(ti), Z(ti), Is(ti), Vs(ti)) (ti i=1..N) realized on a given river section at different stages along the hydrological cycle. The method relies on the forcing of equality (or minimization of deviation) between two expressions of the river discharge : velocity integration on the section and Manning head loss equation. Various criteria have been developed based on the quadratic difference between these two expressions, and minimization techniques have been tested and optimized to estimate the river bottom parameters. The method has been implemented both on simulated data (without noise or with added measurement noise), and on real data (Amazon river). Additionnaly, the robustness of the method to surface variable uncertainty has been explored. A simplified version of the method, with fixed value of the Manning coefficient,results in a 8% discharge estimation with a 25% standard deviation over the Amazon dataset (12 stations). The full method, while giving relevant estimates of river bottom parameters and river discharge on exact simulated data and on some in situ gauging stations, leads to inaccurate results on most of in situ gauging stations as well as on simulated data with significant measurement noise. Current works are dedicated to improve its robustness.