摘要：We present a new method that allows a separation of the attribution of human influence in extreme events into changes in atmospheric flows and changes in other processes. Assuming two data sets of model simulations or observations representing a natural, or 'counter-factual' climate, and the actual, or 'factual' climate, we show how flow analogs used across data sets can provide quantitative estimates of each contribution to the changes in probabilities of extreme events. We apply this method to the extreme January precipitation amounts in Southern UK such as were observed in the winter of 2013/2014. Using large ensembles of an atmospheric model forced by factual and counterfactual sea surface temperatures, we demonstrate that about a third of the increase in January precipitation amounts can be attributed to changes in weather circulation patterns and two thirds of the increase to thermodynamic changes. This method can be generalized to many classes of events and regions and provides, in the above case study, similar results to those obtained in Schaller et al (2016 Nat. Clim. Change 6627–34 ) who used a simple circulation index, describing only a local feature of the circulation, as in other methods using circulation indices (van Ulden and van Oldenborgh 2006 Atmos. Chem. Phys.6 863–81 ).
