Training School on Statistical Modelling of Compound Events
Compound weather and climate events refer to the combination of multiple climate drivers that contributes to societal or environmental risk. They can pose serious threats to natural systems and human societies. Modelling compound events requires knowledge on advanced statistical methods. The goal of this training school organized by the COST Action DAMOCLES is to train the next generation of compound event researchers. The school will provide an introduction into various statistical approaches to study compound events. There will be ample time to work on scientific projects in small groups to apply what has been learned in the classes and to socialize with the other participants and lecturers.
Where: Lake Como School of Advanced Studies (Italy)
For whom: PhD students and early postdocs (maximum 2 years after completion of PhD)
How to apply: Send motivation letter (1 page), CV, preference for student project (1st, 2nd, 3rd choice, see below) and at least one reference (e.g. PhD advisor) as one pdf file to email@example.com.
Application deadline: 10 June 2019
Financial support: 20 stipends between 900 and 1400€ are available from DAMOCLES for student from participating counrtries (depending on the travel distance)
Accomodation: 2- and 3-bedrooms have been reserved at In Riva Al Lago Como.
Organizing committee: Jakob Zscheischler, Carlo de Michele, Aglaé Jézéquel, Philippe Naveau
- Introduction into concept of compound event
- Copula theory
- Multivariate extreme value theory
- Importance sampling
- Processes behind compound events (droughts and heatwaves; compound flooding)
Prof. Fabrizio Durante (University of Salento, Italy)
Prof. Sebastian Engelke (University of Geneva, Switzerland)
Prof. Bart van den Hurk (Deltares, Netherlands)
Prof. Douglas Maraun (University of Graz, Austria)
Prof. Carlo de Michele (Politecnico di Milano, Italy)
Dr. Philippe Naveau (LSCE, France)
Prof. Gianfausto Salvadori (University of Salento, Italy)
Prof. Sonia Seneviratne (ETH Zurich, Switzerland)
Dr. Pascal Yiou (LSCE, France)
Dr. Jakob Zscheischler (University of Bern, Switzerland)
Participants of the Traning School are requested to choose from one of the following 4 student projects. During the school there will be ample time for the four groups to work on their projects
Student project 1
Project title: Identifying drivers of extreme impacts
Supervision: Karin van der Wiel (KNMI) & Jakob Zscheischler (University of Bern)
Extreme impacts are often related to multiple compounding conditions in the weather and climate system. For instance, unfortunate combinations of temperature and precipitation can lead to crop failure or vegetation mortality. Identifying which combination of weather conditions lead to and extreme impacts is challenging and often made even more difficult do to a small sample size in observations. In this project, we will work with very long impact model runs (crop model, vegetation model) and explore approaches to identify multivariate climate conditions that are associated with extreme impacts.
Student project 2
Project title: Importance sampling for compound events
Supervision: Aglaé Jézéquel (LMD) & Pascal Yiou (LSCE)
The goal is to adapt an analogue-based importance sampling algorithm (developed by Pascal Yiou and Aglaé Jézéquel) to a multivariate compound event (currently it is used to simulate single variable extremes such as very extreme heatwaves and evaluate their probability of occurrence). This could for example be useful to simulate very extreme compound events, which may have a higher occurrence probability with climate change. Similar to a weather generator, simulated compound events could be used as an input for an impact model. The project will also compare the return periods obtained by importance sampling with other approaches, e.g. based on multivariate extreme value theory.
Student project 3
Project title: Model evaluation of bivariate relationships with copulas
Supervision: Emanuele Bevacqua (University of Reading) & Carlo de Michele (Uni di Milano)
Combinations of hot conditions and high/low relative humidity can lead to extreme heat stress for the human body or exacerbate the fire risk. To develop reliable heat stress and fire risk assessments, it is crucial to evaluate how climate models represent relative humidity and temperature, as well as their interplay, i.e. statistical dependence. In this project, we will carry out this evaluation and quantify the contribution of the model biases in temperature, relative humidity, and - especially - their dependence, to the final bias in heat stress and fire risk indices. The results will highlight the regions where a careful development of present and future heat stress and fire risk assessments is required.
Student project 4
Benefits and limitations of statistical models for assessing compound flooding
Supervision: Elisa Ragno (TU Delft) & Bart van den Hurk (Deltares)
The goal of the project is to investigate the benefits and limitations of statistical methods to model dependencies between hydro-meteorological data. Specifically, the project will retrace the work presented in van den Hurk et al. (2015) in which the authors investigate the effect of the co-occurrence of precipitation events and storm surges on inland water level using a set of dynamical models. These models will be replaced by a statistical model calibrated at the specific site. The comparison between the results from the two approaches will provide the basis to discuss opportunities and limitations of the application of statistical methods to investigate compound events. Moreover, the project will place emphasis on the importance of compound events for flooding risk assessment.