Next seminar: 

26 March 2026 - Léo Belzile (HEC Montréal)

(14:00 UTC) = (10:00 New York) = (15:00 Paris) = (21:00 Shanghai) = (0:00, March 27 Sydney)

Improved Return Level Estimation Combining Climate Model Output and Historical Records

Abstract:  The distributions of extremes of temperature and precipitation are evolving with climate change, which complicates risk assessment. Climate models provide state-of-the-art information about potential trends, but they are often miscalibrated for extremes, which makes them unsuitable for risk analysis as is. On the other hand, historical records suffer from sparse spatial and temporal coverage, meaning extrapolation in the future is fraught with challenges. We propose using multivariate spatial Gaussian process priors for modelling generalized extreme value parameters of both climate model output and station data, allowing for temporal trends that capture effects of climate change. This allows us to provide predictions throughout the domain and to perform bias correction of risk measures. To showcase our methodology, we model annual maximum temperature at multiple weather stations in Quebec and Ontario, along with high resolution data from the CRCM5-CMIP6 regional climate model, and use these to perform downscaling of return levels at unobserved sites over the two provinces. This work is joint with Alexandra M. Schmidt, Ruihang Shen and Nada Sari.

Please check your local time! 

The calendar below has been created in UTC (Coordinated Universal Time). 

The timezone displayed depends on the settings of your device. When synchronized with your google calendar, it will be your local time.