Post-processing of extreme precipitation projections in southern Quebec

The government’s INFO-Crue project, under the responsibility of the Ministère de l’Environnement et de la Lutte contre les changements climatiques, de la Faune et des Parcs (MELCCFP), aims to prepare Quebec in terms of flood risks. One of the tools used for this purpose is a new edition of the Hydroclimatic Atlas of Southern Québec, whose major new feature in terms of climate is the inclusion of regional models of the NA-CORDEX and ClimEx ensembles in order to better represent projected changes to flooding caused by extreme precipitation, in particular. However, the quantile post-treatment method used at Ouranos is not optimized to adequately correct precipitation distribution tails.

The aim of project was therefore to develop a technique for parametric post-treatment of extreme precipitation in climate models by operationalizing work already underway at Ouranos and Polytechnique Montréal.

• Produce grid-based information to better account for extreme precipitation observed at stations

• Develop and implement a parametric methodology for post-processing distribution tails

• Produce the climate scenarios for the 2020 edition of the Hydroclimatic Atlas of Southern Québec

• Produce grid-based information to better account for extreme precipitation observed at stations

• Develop and implement a parametric methodology for post-processing distribution tails

• Produce the climate scenarios for the 2020 edition of the Hydroclimatic Atlas of Southern Québec

A set of 180 climate scenarios was created for Quebec. These scenarios incorporate a new parametric post-treatment technique developed to correct the distribution tails of simulated precipitation events using a generalized Pareto distribution (GPD). This new technique is quite flexible and fits into the quantile post-processing chain used operationally at Ouranos (Figure 1).

Figure 1: Schematization of the mixed post-treatment method for precipitation. The green boxes represent additions to the quantile post-processing method used operationally at Ouranos (Roy et al. 2020, article in development).

The results demonstrate that the new mixed post-treatment technique can better maintain the distribution of the tails of precipitation events simulated by climate models, while correcting for potential biases in a more robust manner. By default, this parametric post-processing technique uses the GEV parameters produced by Jalbert, Perreault and Genest (2019). However, it is also possible to directly use grid-interpolated observed data, or even to correct grid-interpolated observed data using GEVs before integrating it into the post-processing chain. The latter option was used in this project and turned out to be less robust than the direct use of the GEV parameters because it is sensitive to artifacts in the observed grid. However, it outperforms quantile-quantile post-processing.

Marco Braun, Ouranos