The workpackage focusses on temperature extremes, but also analyzes other events, such as the dust bowl drought (Seager et al.,
2008). Observed extreme events are selected from different situations of the large-scale circulation, representing the locations that experienced peak early 20th century warming, including Eastern North America and Europe. Using the weather@home experiment (see Pall et al., 2011), a team in Oxford, led by Hegerl and co-supervised by Allen, performs a very large ensemble of simulations to identify the probability distribution of extremes given early 20th century conditions (A), and the probability distribution if the contribution by greenhouse gas increases is removed from the observed sea surface temperatures (B).

These simulations are run on a distributed computing network operated in Oxford, which makes use of computer time voluntarily donated. Ensemble A will be augmented with a, much smaller, ensemble of simulations with a new high-resolution atmospheric model, HadGEM3@N144L38 (C; Hewitt et al., 2011) in order to reliably identify the response of atmospheric circulation to sea surface temperatures, and evaluate the weather@home model. The frequency of circulation states that lead to temperature extremes are determined, along with the temperatures that occur during those regimes. The frequency of the selected extremes are compared between simulations with and without greenhouse gas forcing (A,B). This technique has so far only been used for attributing changes in the risk of recent events, such as UK flooding and the Russian heatwave (Pall et al., 2011; Otto et al., 2012). The results determine if greenhouse gases increases have contributed to the risk of temperature extremes as early as in the early 20th century. Careful analysis of observed changes determines the statistics of climate extremes, and of the circulation states that lead to them. The challenge will be to integrate results from the two modelling setups (A,B, vs. C), and to rigorously compare it to observations. The benefit will be crucial scientific underpinning of event attribution (see Stott et al., 2004), and much improved understanding of the causes of early to mid- 20th century temperature extremes.

This WP addresses question iii), and contributing to ii).