Modeling Past and Future Extreme Weather Correlations across North America

Abstract

Extreme weather events, or nonseasonal and unexpected weather phenomena such as tornadoes and droughts, have increased in frequency (Kayaga et al., 2020) and intensity (Seneviratne et al., 2012) as a result of climate change (Cowan et al., 2020). However, most studies on extreme weather focus on the causation (Rouault et al., 2002; Duke et al., 2017), and costs (Beillouin et al., 2020; Giordono et al., 2020), of such events, not their correlation. This study conducted correlation analyses on historical precipitation and temperature data using the Global Historical Climate Network-Daily group of stations (Menne et al., 2012). Extreme values, or days where recorded data exceeded the yearly 90th percentile threshold in heat or precipitation (below 10th percentile threshold for cold) were tested for correlation coefficients of above 0.75 within uniform geographic grids of 3.82 degrees in length in the U.S. and Canada. The relationship between historical yearly averages and their extreme value counterparts was then extrapolated into projections for extreme temperature and precipitation values through 2100 under different emissions scenarios put forth by the Intergovernmental Panel on Climate Change (IPCC Data Distribution Centre, 2018a). 10.96% of Canadian grids revealed correlation between extreme precipitation and extreme temperature values, while only 2.78% of U.S. grids found the same. 34.25% of grids in the U.S. and 14.62% of grids in Canada found correlations between extreme temperature values only, while 5.5% of U.S. grids found correlations between extreme precipitation values only compared to 12.5% of Canadian stations. Percentage of correlating grids mostly increased when correlation tests were restricted from year-round to values of warm summer months only. Local weather systems were found to play substantial roles in the U.S., particularly on extreme precipitation correlations. Overall, both countries found correlations clustered near ocean coastlines, large bodies of water, mountain sides, and some urban communities. Stations finding extreme precipitation correlations showed less spatial variability than stations finding extreme temperature correlations, and stations at higher-latitudes tended to show less correlation. Future extreme values were projected to deviate the most from their mean value counterparts under emissions scenarios with no or little climate change mitigation by 2100. Extreme days across North America reached nearly 50°C for heat, -70°C for cold, and 475 mm for annual precipitation on average by year 2100 under the most intense emissions scenario. Extreme weather response policies will prove critical to combating the human, ecological, and economic effects of multi-community extreme weather events as spatial correlation and extreme weather magnitudes intensify in the future. This study puts forth a few recommendations of risk management strategies in order to inform policy proposals.