This study presents a reliable methodology for monitoring streamflow in a dynamic river of the Alps prone to bathymetric changes using non-contact instruments. The method relies on water level and surface velocity radar monitoring, discharge measurements by Large-Scale Particle Image Velocimetry (LSPIV), and topographic surveys. A single proportional relation, resistant to bathymetric changes, is established between maximum surface velocity (Vs,max) and bulk velocity (Umean). Different methods are used to build this relation: (i) an empirical approach calibrated with the LSPIV measurements; (ii) the Isovel model; (iii) the Q-Commander software developed by the Sommer company. The applicability of the method is tested over a 2.5-year dataset. Compared to the empirical approach, both models, which require minimal input data, predict well the Vs,max-Umean relation. The location of the maximum surface velocity, which reveals to be resistant to bathymetric changes, is also well predicted by these models. Discharge is calculated at a time step of 10 min by multiplying the bulk velocity and the wetted area. The results are compared to the discharge series at the historical station located 2.5 km further upstream, which has a stage-discharge rating curve. Good agreement is observed when surface velocity is above 0.7 m/s, but accuracy decreases for lower velocities. A simplified uncertainty analysis estimates a 20% relative error on discharge calculated with the presented method.