Attribution of Extreme Temperature Events for the Western US using a superensemble of Regional Climate Model Simulations

Observed trends in extreme heat constitute a possible signal of human-influenced changes in global climate. The California and Nevada heat wave of 2006, as well as the record-setting heat wave of late June 2013 in the Southwest were both deadly and costly for the local population. Events such as the severe 2011 Texas heat wave and drought have been shown in recent studies to be more likely due to anthropogenic climate change. The present work focuses on attribution of these events by employing a superensemble of regional climate model simulations from the climateprediction.net (CPDN) experiment, which allows for robust statistical analysis of climate change impacts over the Western US. Through the use of public volunteered distributed computing, the project provides an ensemble size large enough to examine the tails of the distribution of climate variables so critical for reducing the uncertainty for historically rare events. Specifically, the present study compares the decade of the 2000s and the 1960s, in which anthropogenic greenhouse gases were not believed to have had as large an influence. This analysis yields a rich dataset of indicator extreme event variables such as the number of days above 35°C in a month, which enable a more direct assessment of heat wave occurrences. Investigation of return periods for the number of days above 35°C threshold over the Desert Southwest domain suggests heat waves such as those in 2006 and 2013 to be close to 10 times more likely during the 2000s compared to the 1960s. Other variables studied include the number of days over 40°C and the highest daily temperature in a month, along with monthly averages of maximum and minimum temperature. The average monthly minimum temperature, a crucial variable for healthcare concerns during heat waves, also displays a marked rise in magnitude and a strong positive shift in frequency of warm summer months. Results from this experiment also highlight the influence of increasing number of simulations on confidence intervals, which significantly reduces the uncertainty of both the change in magnitude of a given event and its corresponding return period.