Projected Trends of Wintertime North American Surface Mean and Extreme Temperatures over the Next Half-century in Two Generations of Canadian Earth System Models

Abstract

ABSTRACT Based on two single-model initial-condition 50-member ensembles of climate simulations conducted with two generations of Canadian Earth System Models (CanESM2 and its successor CanESM5), we analyze the ensemble mean and spread of the projected trends of wintertime North American surface air temperature (SAT) and extreme indices of cold (TX10) and warm (TX90) days over the next half-century (2021–2070) and explore the contribution of internal climate variability to these trends. The ensemble mean of future climate simulations forced by the high-emissions scenario Representative Concentration Pathway 8.5 (RCP8.5) in CanESM2 and the Shared Socioeconomic Pathway 8.5 (SSP5-8.5) in CanESM5 reveals a poleward intensified warming, high risk of severe warm days over the west coast of North America and northern Canada, and a weakening belt of extreme cold days extending from Alaska to the northeastern United States. The warming trend is stronger in CanESM5 than in CanESM2, likely because of higher climate sensitivity and slightly higher CO2 emissions in CanESM5. Large ensemble spreads are apparent in the SAT trend and in the historical simulations and future projections of extreme temperatures, especially for the TX10 index. Individual realizations differ from the ensemble mean in both spatial pattern and magnitude of the projected trends. The signal-to-noise ratio reveals strong signals of the SAT and TX90 trends primarily over the west coast of North America and northern Canada, along with relatively strong signals of the TX10 trend over most of the central to eastern parts of North America in CanESM2 and western Canada and the southwestern and eastern United States in CanESM5. The components of the mean and extreme temperature trends generated by internal climate variability exhibit large-scale spatial coherences and are comparable to the externally anthropogenic-forced components of the trends, mostly in the central parts of North America. Overall, similar ensemble mean patterns of North American mean and extreme temperature trends are evident in the two models; CanESM5 tends to be less uncertain in projecting those trends than CanESM2.