Following the Québec (2011) and Alberta (2013) floods, private insurance companies have been invited to play an important role in covering flood risk in Canada, as it is already the case in the other G8 countries. In this context, we propose a hierarchical approach to generate reliable simulations of flood losses from the Chaudière River (Québec). Scenarios from a chain of physically-based models representing the climate, hydrology and hydraulicity are used to represent water discharge at a specific location on the river. Then, using the distribution of water discharge, a hydraulic model (HEC-RAS) calculates water levels (stage) in a 3D map along the river. Moreover, we employ civil engineering data to map the level of damage as a function of water level. That allows us to obtain the full distribution of losses at the policyholder level which is then used to price flood risk along that river. We also classify flood risk by territories and analyze various flood insurance contracts.