Comparison of convection-permitting and coarser-resolution regional climate simulations for evaluating precipitation and streamflow in Southern Québec

Abstract

Understanding how climate model resolution affects the representation of heavy and intense hydrometeorological events is crucial for enhancing flood risk assessment. Most studies on the impacts of climate change have utilized Regional Climate Model (RCM) simulations with spatial resolutions coarser than 10 km. Deep convection parameterization is necessary to simulate convective events with RCMs, which reflect the averaged effects of convection. This may thus introduce uncertainty and can lead to underestimation of extreme rainfall. Recent work has demonstrated that high-resolution climate models (with grid spacing of 1-4 km) have the capability to improve simulations of sub-daily extreme precipitation compared with coarser-resolution regional climate models. In this work, we use three simulations of hourly Weather Research and Forecasting (WRF) model configuration (hereafter called CONUS, CORDEX 22 and CORDEX 44), all driven by the ERA-Interim reanalysis, at different spatial resolutions of 4 km, 25 km, and 50 km. We aim to compare precipitation and temperature patterns across simulations for three seasons (April-May, Summer, and Fall) at hourly temporal resolution over southern Québec, Canada. In terms of extreme precipitation analysis, we found that the WRF model with 4 km (CONUS simulation) produced the highest RX1-hour values (Maximum 1-hour Precipitation) during the summer. In contrast, the two CORDEX simulations show precipitation patterns that are comparably less intense. We further analysed how these differences extend into hydrological modelling. To do so, we simulated streamflow using a lumped hydrological model (GR5dt) at hourly time steps, across twelve southern Quebec watersheds (surface areas of 61–1550 km²). We used Satellite-derived precipitation product (IMERG) precipitation data and ERA5-Land temperature as our reference datasets for calibrating the hydrological model. The results show that during April-May, the higher temperatures from the CONUS simulation lead to earlier snowmelt and lower streamflow. Conversely, the two CORDEX simulations with colder temperatures accumulated larger snowpacks, resulting in delayed and more noticeable spring floods. The analysis of CONUS generally showed that simulated annual peak flows were underestimated, particularly during the summer, despite the model representing higher hourly precipitation intensities. This can be explained by the higher values of simulated potential evapotranspiration due to CONUS higher temperatures. However, a case study of a small watershed indicates that the CONUS simulation is capable of generating intense, short-duration hydrometeorological events that are not observed in the CORDEX22 and CORDEX44 simulations.

Date	20 Jul 2026
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Annie Poulin (Supervisor)