Investigating the Influence of Non-Gaussian Temperature Distribution Tails on Future Changes in Extreme Warm Threshold Exceedances

Changes in extreme temperature threshold exceedances under warming may be manifested in more complex ways when the underlying temperature probability density function (PDF) exhibits non-Gaussian tails compared to if the distribution were normal. For example, a greater increase in the number of days exceeding an extreme warm temperature threshold is demonstrated for PDFs with shorter-than-Gaussian warm side tails under the simplest prototype for future warming, a uniform warm shift. Leveraging this simple warm shift prototype as a way to measure non-Gaussian short warm tails, we demonstrate that such special PDF cases are commonly found globally in reanalysis, often in spatially coherent regions. Because of the notable implications that these special cases of PDF shape carry, it is important that climate models are able to capture non-Gaussian features with reasonable fidelity to boost confidence in their projections of future changes in extremes. Evaluation suggests that many GCMs are capable of capturing regions of notable short tails, although the model suite exhibits considerable intra-ensemble variability in skill. Next, the degree to which model skill at capturing regions of short warm side tails influences projections of future changes in exceedances under warming is assessed. In most cases, models that capture the non-Gaussian short tails with high skill also project greater than Gaussian exceedances of warm extremes, supporting the postulate that realistic simulation of non-Gaussian short tails in the current climate is important for projecting future changes in extremes with confidence. It is further demonstrated that shorter-than-Gaussian warm tails result in earlier emergence in and greater magnitude of mid-century increases in extreme warm threshold exceedances compared with regions exhibiting near Gaussian PDF tails. A preliminary investigation of the driving mechanisms behind regions of non-Gaussian short tails is presented as a step towards assessment of model skill at the process level and a more meteorologically based interpretation of projections of future change.