Ongoing projects
eXtreme events: Artificial Intelligence for Detection and Attribution (XAIDA)
Climate is changing fast, with losses and damages experienced in every region and every sector. However the awareness of this fact remains limited. Of particular relevance is how climate change modifies and enhances extreme weather events. XAIDA, an EU-funded project, brings together the interdisciplinary expertise of a research consortium of 15 universities and research organizations that unites experts in machine learning, statistics and climate modeling.
Together we will design new methods and apply them to recent high-impact events to understand the role of climate change. Further, we will study if such events, or even more-intense ones, will occur in the future. We will collaborate with concerned stakeholders from different sectors to prepare risk assessment and adaptation strategies for extreme weather.
Identifying causal drivers of floods (CausalFlood)
Floods are among the most damaging natural hazards, causing thousands of deaths and millions of dollars in damages every year. Understanding which physical drivers cause floods is crucial for flood forecasting and managing flood risk under changing environmental conditions. CausalFlood will advance the development of a state-of-the-art causal inference method for the application to flood drivers. The method will be applied to well-calibrated high-resolution hydrological simulations across different catchments in Europe to identify key flood drivers.
Detecting compound climate features of extreme impacts (COMPOUNDX)
COMPOUNDX exploits multiple available data archives of climate and impact model output as well as observation-based datasets to (i) develop statistical tools to identify compound climate features of extreme impacts based on machine learning, (ii) apply the developed tools in the real-world observations, and (iii) evaluate process-based models in their ability to simulate compound features. The impacts considered in the project are crop failure, carbon cycle extremes, and floods.
Completed projects
Compound events in the ocean
Extreme events shape the structure of biological systems and affect the biogeochemical functions and services they provide for society in a fundamental manner. There is overwhelming evidence that ocean extreme events, such as marine heatwaves, will increase in frequency, duration and intensity under future global warming, pushing marine organisms, fisheries and ecosystems beyond the limits of their resilience. Of particular concern are compound events, which correspond to events with multiple concurrent or consecutive drivers (e.g. marine heatwaves co-occur with very low nutrient levels) resulting in extreme consequences for marine ecosystems. This PhD project analyses compound events in the ocean in observations and models and quantifies uncertainties in model projects.
Risk assessment of critical ecological thresholds in Amazonia and Cerrado
The hydrological cycle is changing across the tropics due to interactions between global climate change and deforestation. The impacts on natural systems can be large, persistent, and may have major consequences for the carbon cycle and atmospheric CO2 concentrations. Many of the largest impacts on forests, savannas, agricultural systems result from the occurrence of extreme weather events. Recent weather extreme events have already caused (a) fundamental changes in the structure of tropical forests, (b) widespread losses in agriculture output, and (c) catastrophic forest and savanna fires. In this project, we use statistical models to quantify the multiple climatic drivers of large impacts to tropical forests, savannas, agriculture, and disturbance regimes. We assess how climate change may alter the likelihood of compounding, catastrophic weather events to occur in the near future across Amazonia and Cerrado.
Machine learning for detecting compound climate drivers of extreme impacts
In this project, we use of a dynamic global vegetation model (LPX-Bern) to explore the ability of state-of-the-art machine learning techniques to identify climate and weather features that cause extreme reduction in vegetation productivity. LPX-Bern is forced with very long simulations from a weather generator to generate large amounts of impact data. We then use machine learning approaches such as Convolutional Neural Networks to identify climate conditions that are associated with extremely low vegetation productivity through classification.
New metrics for constraining multiple drivers of hazard and compound hazards
The impacts of climate extremes are often particularly severe when several hazards occur at the same time. Currently it is not clear whether current climate models can capture major changes in risk associated with climate-related hazards. Existing modelling approaches used to assess risk may therefore lead to serious mal-adaptation. This project develops new metrics to evaluate climate models with respect to multivariate relationships, extremes, and compound hazards and uses these metrics to constrain model ensembles with observations to improve projections of hazards and compound hazards. The project focuses on the compound hazards a) drought and heat, b) wind and precipitation extremes, as well as on the hazards c) human heat stress and d) fire risk.