Search for a command to run...
The COMPASS project aims to bridge the gap between attribution of single-driver extremes and more complex extremes, by producing a methodological framework for climate and impact attribution of such extremes for a range of hazard types. The aim of this report ‘Deliverable 2.7: Report on recommendations for attribution methods suitable for compound events with damaging impacts ’ is to provide recommendations for attribution methods suitable for compound events with damaging impacts, using the additional knowledge gained from the use cases and other work carried out in COMPASS project. The report finds four common themes for compound attribution methods. Firstly, the strong benefit of impact attribution for complex extremes, where many variables can be combined to produce one set of impacts. Hence bypassing many of the statistical challenges faced when going from the attribution of a single variable to multiple variables, but it is also beneficial given the difference in attribution results found in many of the use cases between when impacts are modelled and when just the hazard is modelled. Secondly, the COMPASS use cases highlight the benefit of modelling multiple hazard variables compared to just a single hazard, with the inclusion of the compounding nature of the event shown to amplify impacts in many of the use cases. Thirdly, there are challenges around the impact attribution of compound drought-heatwaves, with many different impacts all requiring different hazard models that in many cases are not openly available. This was less of a problem for compound flood-based events. The fourth theme was the importance of including exposure and vulnerability assessments and their role in how changes in the hazard due to climate change are being translated into societal impacts. Recommendations are provided for attribution methods applied to events with damaging impacts for each of the four categories of complex extremes; multivariate, temporally compounding, spatially compounding and preconditioned (Zscheischler et al., 2020) for different event types. These are based on the feedback and experience of the use case authors, the common themes identified and the transferability and ease of implementation of the methods to different event types and regions. These are general recommendations for the most suitable method; however, to improve confidence in the results a wide range of methods should ideally be used for the event in question. Furthermore, different attribution methods answer different questions about the event in question. Therefore, we encourage stakeholder engagement at an early stage in order to guide the choice of method to produce the most useful output.