Production of new hydroclimatic simulations for southern Quebec using Raven: An examination of the uncertainty related to hydrological modelling in the context of climate change

The use of the Raven framework will help building several semi-distributed models appropriate to the watersheds of southern Quebec. It will allow the exploration of the different algorithms to simulate snow melting and evapotranspiration.

Project details
Scientific program
2020-2025 programming
Theme(s) and priority(s)
Extreme Events
Start and duration
April 2021• 30 months
Project Status
In closing
Linked project
Support for INFO-Crue
Principal(s) investigator(s)
François Anctil
Université Laval
Biljana Music
Aida Jabbari
University Laval


Following the spring floods in 2017, the Ministère de l’Environnement et de la Lutte contre les changements climatiques (MELCC) initiated the INFO-Crue project, aimed in particular at delimiting flood areas based on new mapping that takes climate change into account. The operational hydrological modelling platform used by the MELCC for this mapping is based on a single hydrological model (HydroTel), which has a fairly rigid structure, limiting its ability to describe uncertainty well. 

This project explores a flexible hydrological modelling platform, Raven, to create several variants of a semi-distributed model (HBV-EC; Fig. 1), with complexity similar to the model currently used to develop the Hydroclimatic Atlas of Southern Quebec (HydroTel). These different models can then be used to generate ensembles of hydrological simulations to characterize the uncertainty associated with hydrological modelling.

Figure 1 Raven

Fig. 1. 48 variants of the HBV-EC model, combining various formulations of potential evapotranspiration, snowfall and snow melt.


The objective of this study was to implement the HBV-EC model emulated by Raven on the southwest Saint Lawrence hydrographic region and to propose a multi-model methodology allowing for an accurate description of the uncertainty associated with hydrological modelling.


This project is part of the INFO-Crue initiative set up by the MELCCFP.


  • Calibration and validation of the HBV-EC model for the southwest Saint Lawrence hydrographic region

  • Use of several modules offered by Raven to represent a single hydrological process in different ways, and quantification of structural uncertainty

  • Application of a subsampling approach to create a sub-ensemble that is less demanding in terms of calculation time and to possibly improve overall performance;

  • Perturbation of the input data by introducing different noise levels to the precipitation and temperature time series in order to quantify the uncertainty associated with meteorological forcing

  • Examination of the uncertainty associated with the calibration process used


In this project, forty-eight variants of the HBV-EC model were used to produce hydrological simulations in historical and future climates. The simulations were performed in the Raven modelling system. The work focuses primarily on the 2000-2020 time period, since flow observations are required to verify the quality of the description of uncertainty. 

In addition, the comparison between the simulated and observed flows shows that the ensemble of 48 models appropriately represents the hydrographs observed at the various hydrometric stations. The application of the sub-sampling approach resulted in the selection of an ensemble of 12 models whose performance is comparable and sometimes superior to that of the 48 models. Figure 2 shows a summary of the work for the Rivière au Saumon watershed. The left panel illustrates the dispersion described by the ensemble of all 48 hydrological models driven by deterministic climate forcing, without taking uncertainty into account. The right panel illustrates the dispersion described by 12 selected hydrological models, driven by probabilistic climate forcing that takes  uncertainty into account. The visual inspection of both panels confirms that a procedure based on 12 hydrological models and noisy inputs allows for an accurate description of uncertainty. Adapted verification metrics were used to confirm this finding. The analyses also showed that this ensemble of 12 models leads to better simulations than using the HBV-EC model alone (Fig. 3).

Figure 2 raven

Fig. 2. Comparison between the annual mean flow observed at the Rivière au Saumon station (black line) for the 2000-2020 time period and the ensembles of flows simulated by (a) 48 models and (b) 12 models selected with atmospheric forcing. The median flow is shown by the red line, the 5th to 95th percentile of the ensemble is shown in pale blue, and the 25th to 75th percentile is shown in dark blue.

Figure 3 raven

Fig. 3. Comparison between a single model and the ensemble of the 12 models selected, for the Famine River watershed: (a) performance (root mean squared error) and spread, and (b) reliability diagram.

Benefits for adaptation

Benefits for adaptation

Implementation of HBV-EC modelling emulated by Raven for improved projection of hydrometeorological risks

Better understanding and description of the uncertainty associated with hydrological modelling

The exploration of robust solutions for flood risk management by making hydrological simulations available to the public in order to contribute to river models and improve the mapping of flood zones


Other participants

  • Direction de l’expertise hydrique (MELCCFP)

Related projects



button back to top