
Spatial and Temporal Characteristics of Temperature-Precipitation Covariability across Canada in a Changing Climate
Abstract
Changes in the regional characteristics of temperature and precipitation can intensify the occurrence and severity of extreme events such as rain-on-snow induced flooding, droughts and wildfires. Analyzing these climate variables in isolation without considering their interdependencies might result in severe underestimation of their combined effects. In this study, copula functions are used to describe the joint behaviour of temperature and precipitation across Canada over a historical period of 1910-present using observations and further till the end of the 21st century using three large ensembles of regional climate models. Moreover, given the lack of observation data over Canada, gridded datasets are also evaluated under both univariate and bivariate settings. The importance of preserving the dependence structure is shown through a hydrologic model forced with multivariate bias-corrected data. Climate projections are evaluated against observations using a hierarchical Bayesian framework followed by calculation of extreme climate indices over four future warming scenarios corresponding to +1.5, +2.0, +3.0 and +4.0°C mean temperature rise above the pre-industrial period. Finally, an ensemble pooling approach is applied to calculate non-stationary return periods of compound extreme events.
Results show clear signs of accelerated warming and wetting over northern Canada and strong evidence of hot and dry conditions in Prairie Provinces while non-stationary analyses reveal shifts towards warm and wet climate conditions for the rest of southern Canada. Results from the comparison of multiple gridded datasets show that while they represent temperature and precipitation well, their performance in simulating the joint behaviour is relatively weak. Hydrological modelling results indicate that the multivariate bias correction of the input datasets can improve streamflow simulations particularly extreme events compared to the univariate approach. Climate projections show an increase in warm spells in the future accompanied by an increase in extreme precipitation as well for most regions in Canada. The importance of considering the dependence of temperature and precipitation extremes in calculating joint return periods reveals potential future changes in the frequency of compound extremes. Overall, this study provides a comprehensive characterization of the joint behaviour of temperature and precipitation over Canada under a changing climate.