A Soft Record Analysis of Extreme Heat Across Australia

Extreme weather far beyond what has been experienced in recent memory can be especially dangerous and costly. Proactively identifying locations at high risk of experiencing unprecedented weather can assist with disaster preparedness. Such locations can be referred to as having soft records, meaning the most extreme event in observational records is not particularly severe compared to what is possible. In previous studies, the systematic identification of soft records over a large spatial domain involves applying extreme value analysis to gridded observational or reanalysis data. A limitation of these studies is the small sample size, which we propose can be addressed by adapting the UNprecedented Simulated Extremes using ENsembles (UNSEEN) approach that is commonly used to estimate event likelihood in the aftermath of isolated unprecedented events. The UNSEEN approach makes use of seasonal or decadal forecast/hindcast ensembles, which provide a large sample of plausible events over recent decades. To demonstrate the utility of applying the UNSEEN approach to a large spatial grid, we assessed record daily maximum temperatures across Australia using gridded observations and data from 10 different decadal forecasting systems. The observation‐based results highlighted broad areas of soft records in the south‐east of mainland Australia, extending north into south‐west and western Queensland. The UNSEEN‐based analysis also identified soft records in western Queensland, but not in the south‐east where the underlying positive trends in extreme temperature were far less severe in the models than in observations. We suggest that the use of large model ensembles (i.e., an UNSEEN‐based approach) can complement an observation‐based approach to identifying soft records over large gridded spatial domains.