Impacts of severe land use changes on the hydrology of snow dominated catchments in southern Quebec

• Hydrological models’ response to RCM-simulated land use changes were evaluated. • Extreme deforestations shows higher flood peaks in snow-dominated catchments. • The choice of hydrological model structure (i.e., more sensitivity to snow melt/accumulation vs to evapotranspiration) impacts level of sensitivity to land use. The present study evaluates the hydrological response to severe land use changes using the distributed physically-based hydrological model HYDROTEL over 89 catchments in Eastern North America. Additionally, two catchments were selected to assess the impacts on past extreme flood events. Streamflow simulations were generated using three regional climate simulations from the Canadian Regional Climate Model version 5 (CRCM5) for the 1986–2015 period. The simulations were based on current land use (EVAL) and two extreme land use scenarios with only grass (GRASS) and only forest (FOREST) land use. The FOREST and GRASS simulations include the feedback of land-use to climate. The Multivariate Bias Correction (MBCn) algorithm was applied to correct biases in EVAL simulation, with the same coefficients used for GRASS and FOREST simulations to ensure a consistent climate baseline and maintain the spread of the simulations. HYDROTEL’s land-use inputs were adjusted to match those from the CRCM5 simulations. The results indicated significant changes in both the timing and magnitude of spring peak flows associated with snowmelt: the FOREST scenario showed earlier, reduced peak flows, while the GRASS scenario led to delayed, larger peaks. Additionally, the catchments were categorized by size-large (>1000 km 2 ), medium (500–1000 km 2 ), and small (<500 km 2 )-to evaluate size-related impacts. A model comparison was conducted between HYDROTEL and the physically-based distributed model WaSiM on two diverse catchments, revealing that HYDROTEL demonstrated higher sensitivity to snowmelt and accumulation processes (because of its multi-land use snow model structure), while WaSiM showed greater sensitivity to evapotranspiration (due to its Richards-equation-based soil module). Finally, it was found that the RCM-simulated land use-climate feedback is considerably more important than the hydrological model land-use change effect on the spring flood modification.