Fast-track extreme event attribution: General methods and techniques to determine the dynamic contribution to an event.
Abstract
Extreme weather event attribution has become an accepted part of the atmospheric sciences with numerous methods having been put forward over the last decade. We have recently established a new framework which allows for event attribution in quasi-real-time. Here we present the methodology with which we can assess the fraction of attributable risk (FAR) of a severe weather event due to an external driver (Haustein et al. 2016). The method builds on a large ensemble of atmosphere-only GCM simulations forced by seasonal forecast SSTs (actual conditions) that are contrasted with ensembles forced by counterfactual SSTs (natural conditions). Having an associated 30 year actual and natural climatology in place, we are able to put the current event into a climatological context and determine the dynamic contribution that lead to the event as opposed to the thermodynamic contribution which would have made such an event more likely regardless of the synoptic situation. As a second independent method (also applicable in near-real-time), we apply pattern correlation to separate thermodynamic and dynamic contributions. Finally, using reanalysis data, we test whether our attributed dynamic contribution is also detectable in the observations. Despite the high monthly variability, ENSO related teleconnection patterns can be detected fairly robustly as we will demonstrate with a recent example during El Nino. The more consistent the 3 methods are, the more robust our results will be. We note that the choice of time scale matters a lot when determining the dynamic contribution as well as estimating the FAR (Uhe et al. 2016). The weather@home ensemble prediction approach is accompanied by two more methods based on observational data and the CMIP5 ensemble. If the FAR across 3 methods is consistent, we have reason to trust our central attribution statement. Two recent examples will be shown in order to demonstrate the feasibility (van Oldenborgh et al. 2016a/2016b), complemented by new results from South Asia where we also investigate the effects of anthropogenic aerosols.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2016
- Bibcode:
- 2016AGUFMGC13G1263O
- Keywords:
-
- 0520 Data analysis: algorithms and implementation;
- COMPUTATIONAL GEOPHYSICSDE: 0555 Neural networks;
- fuzzy logic;
- machine learning;
- COMPUTATIONAL GEOPHYSICSDE: 1699 General or miscellaneous;
- GLOBAL CHANGEDE: 1986 Statistical methods: Inferential;
- INFORMATICS