Short‐tailed temperature distributions over North America and implications for future changes in extremes

Abstract

Some regions of North America exhibit nonnormal temperature distributions. Shorter‐than‐Gaussian warm tails are a special subset of these cases, with potentially meaningful implications for future changes in extreme warm temperatures under anthropogenic global warming. Locations exhibiting shorter‐than‐Gaussian warm tails would experience a greater increase in extreme warm temperature exceedances than a location with a Gaussian or long warm‐side tail under a simple uniform warm shift in the distribution. Here we identify regions exhibiting such behavior over North America and demonstrate the effect of a simple warm shift on changes in extreme warm temperature exceedances. Some locations exceed the 95th percentile of the original distribution by greater than 40% of the time after this uniform shift. While the manner in which distributions change under global warming may be more complex than a simple shift, these results provide an observational baseline for climate model evaluation.

High‐Resolution Dynamical Downscaling Ensemble Projections of Future Extreme Temperature Distributions for the United States

“…The decreased variability noted above also occurred in the Northeast for the WGN and WH simulations, but the background warming for this region caused the median to shift enough that the historical 95th percentile still occurred close the 50th percentile. This result is in good agreement with Loikith and Neelin () where they concluded that shorter‐than‐Guassian high‐side tails have the potential to greatly increase in the frequency of extreme values if the shift in the PDF mean is large enough. For the Midwest and Northeast, the WGN and WH simulations experience a shorter‐than‐Guassian high‐side tail, but the background warming in the mean is large enough in the Northeast to still yielded greater increases in the TMAX extremes.…”

“…This is most evident in Northwest (Figure c) where the variability decreases by 15%–38% (17%–28%) for all five simulations when measuring the interquartile range for 2085–2094 under both scenarios (R8Y4) scenario. The proximately to the Pacific Ocean is likely a moderating factor in these PDF curves as the cool ocean water likely limit large increases to the warm side during the winter months (Loikith & Neelin, ). The Midwest experience similar morphology change in the PDF curves with a 16–29% decrease in IQR change under R8Y8.…”

“…To understand the robustness of the changes in extreme temperature from historical to future period, we conduct statistical significance tests on the means of the PDF curves. As Ruff and Neelin () and Loikith and Neelin () suggested, the PDF curves of daily temperature are not always normally distributed. In particular, Ruff and Neelin () found as the climate warms, individual station data projects modest or large departures from Guassianity.…”

Large‐Scale Atmospheric Control on Non‐Gaussian Tails of Midlatitude Temperature Distributions

“…Future work with this model will include land‐sea contrast and a simple treatment of vertical motion to enable comparisons with Northern Hemisphere data. Nevertheless, the results of this simple model that changes to both the variance and the skewness are inherent to global warming suggest caution in applying Gaussian statistics for predicting future extremes (e.g., Loikith & Neelin, ).…”

“…To understand long‐tail and short‐tail temperature distributions (e.g., Loikith & Neelin, ), we solve the advection‐diffusion equation on the sphere using the pseudo‐spectral method for temperature θ , with Newtonian relaxation to a prescribed equilibrium temperature profile: where κ is the hyper‐diffusion coefficient that results in a 0.1 day damping timescale on the smallest resolved spherical harmonic and τ is the thermal relaxation timescale. The model is run at T42 resolution.…”