Zwiers, Francis2023-08-142023-08-142020https://hdl.handle.net/10388/14877Page 28 of book of abstracts.Statistical extreme value theory (EVT) is a fundamental tool for characterizing climate extremes and understanding whether they are changing over time. Most operational frequency and intensity estimates are obtained by using EVT to analyze time series of annual maxima; for example, of short duration precipitation accumulations or some aspect of wind speed. A key implicit assumption in the application of EVT is “max-stability”; i.e., that the statistical behaviour of annual maxima is predictive of maxima calculated over multi-decadal or longer intervals. This assumption cannot be tested using available observational records, and it is rarely discussed in studies of extremes. Here we use a recent large ensemble simulation to assess whether max-stability holds for annual maxima of extreme precipitation. We find that annual maxima tend not to be max-stable in the model-simulated climate. We explore the implications of the lack of max-stability on the estimation of very long period return levels, and discuss reasons why the annual maxima of precipitation extremes may not be max-stable. We also demonstrate a possible solution that is based on an alternative statistical approach and that incorporates additional process-based information into the analysis. While our study focuses on precipitation simulated by a regional climate model, our findings have serious implications for the estimation of high return levels of many climate and weather elements from models and observations that may potentially impact engineering practice.enAttribution-NonCommercial-NoDerivs 2.5 CanadaWater researchClimate changeExtreme weatherPrecipitationClimate modelsLong period return level estimates of extreme precipitation: abstract