Event Attribution: A fast Method based on Surrogate Fields.

Abstract:

For event attribution we need probability distributions of the quantity under consideration for both current conditions and pre-industrial conditions. These distributions will typically be calculated from large ensembles produced by a climate model and require large computational resources. Here we discuss an alternative based on a simple algorithm to produce ensembles of surrogate fields based on observations or a few climate model experiments.

This method can be seen as a field extension of bootstrapping as it produces surrogate fields with the same spatial and temporal structure (as measured with lagged cross-correlations) as the target field. We have previously used this method in tests of paleo-reconstruction methods and in a study of the significance of warm records.

In this study the method will be used for event detection. We will in particular be interested in recent European extreme warm events and cold spells. The method allow us to address issues such as the relative effects of the two aspects of climate change: the change in the mean and the change in the variance. We will also address the selection problem. Some studies consider a single geographical site where an extreme event has been observed and investigate the probabilities for such an event at that particular site. However, considering the probability for a similar event in an extended area will give different probabilities and may result in different conclusions regarding the attribution.