Inconvenient: Canada Summer Daily High Temperature Trends, 1900-2023

SUMMARY

1. Over the period 1900-2023, the average summer (JJA) daily high temperatures across the six southernmost large provinces of Canada (British Columbia, Alberta, Saskatchewan, Manitoba, Ontario, and Quebec) show no trend.
2. The average of the 3 hottest days’ in each month month show a slight downward trend, while the 3 coolest days’ average temperature per month shows a slight upward trend.
3. Recent years have generally averaged as warm as was experienced in the 1920s to 1940s, with 7 of the 10 hottest summers occurring in the 1930s.
4. Results for the 6 provinces separately are also presented.

Introduction

Given media reports, it is likely that most Canadians think they have been experiencing unprecedented summer warmth in the last couple of decades. But this isn’t true.

Below I present analyses of daily high temperatures (Tmax) from all available stations in the 6 southernmost large provinces, based upon the daily Global Historical Climate Network (GHCNd) dataset. These are the 6 provinces that border the Lower 48, and contain 86% of Canada’s population.

I simply averaged together the relevant statistics (monthly average Tmax, average of the warmest 3 days in each month, and average of the coolest 3 days in each month) from all available stations. Each station had to have at least 90% of the days in a month reporting data for that month to be included in the analysis.

Since stations come and go over the years, and since there are some large terrain elevation variations in western Canada, I performed an elevation correction to these Tmax metrics, in all provinces, using the departure of each year’s station-average elevation from the all-year (1900-2023) station average elevation, using a lapse rate of 6.5 deg. C per km. Corrections for average changes in station-average latitude were not done, which might be necessary in the winter since there are large North-South gradients in air temperature then. Such corrections in the summer would likely be small, but I can revisit that nuance at a later time.