Background

Europe’s most damaging weather rarely occurs as a single hazard. Winter windstorms can coincide with heavy rainfall, severe events can cluster within weeks, and a single weather system can affect multiple countries. These “compound extremes” can cause financial losses far greater than individual hazards, yet we have limited understanding of how uncertain these losses are in a changing climate.

New climate ensembles and machine-learning methods make it possible to explore rare and interconnected extremes, but uncertainty enters at every stage: from climate models and natural variability, through rare-event statistics, to translating hazards into financial losses. A key challenge is to identify which uncertainties dominate future loss estimates, how they propagate through the climate-to-loss chain, and which can be constrained using observations and improved modelling. This research will support more robust estimates of climate-related financial risk.

PhD project

How uncertain are future financial losses from compound wind and rainfall extremes, what drives this uncertainty, and which uncertainties can be reduced?

This PhD will develop a framework to understand how uncertainty propagates from climate projections and regional weather extremes to financial losses. The aim is not simply to estimate how compound events will change, but to determine why future loss estimates are uncertain, how different sources interact, and which matter most for financial risk.

The student will investigate uncertainty in the frequency, intensity and temporal clustering of compound wind and rainfall events across Western Europe, focusing on poorly sampled joint tails where small differences in hazard dependence and return periods can produce large differences in estimated losses. Large climate-model ensembles, multivariate extreme-value statistics and machine-learning emulators will be combined to explore rare events at the scale needed to quantify these uncertainties.

A key innovation will be to decompose uncertainty across the climate-to-loss chain, distinguishing contributions from climate-model structure and parameters, internal variability, future scenarios, statistical estimation of rare compound events, spatial and temporal aggregation, and the translation of hazards into financial losses. Sensitivity and uncertainty-propagation methods will identify which sources dominate decision-relevant loss metrics.

The project will also assess where uncertainty can be constrained using observations and historical events, distinguishing reducible uncertainties arising from limitations in models or data from fundamentally irreducible uncertainty. Working with Aon Impact Forecasting (co-supervisor Aidan Brocklehurst), the student will test these advances across the UK, France, Germany, Belgium and the Netherlands, developing transferable methods for other compound hazards and regions. Results will support robust catastrophe-risk estimates and climate risk decisions across Europe.

Applicant Profile

We welcome students with a strong background in physics, meteorology, mathematics, statistics, and/or any related experience in quantitative disciplines who are excited by applying scientific and computational methods to real-world climate risk and decision-making.

Strong Python programming skills are highly desirable and will be valuable throughout the project. Experience with statistics, machine learning, climate science, or catastrophic modeling would be advantageous, but no prior knowledge is required. Training will be provided in the relevant climate, statistical, and risk-modelling concepts.

The project would particularly suit a student who enjoys working across disciplines and is interested in understanding not only what future climate risks might be, but also why estimates of those risks are uncertain and how scientific evidence can improve them. The project will not involve developing or running climate models; instead, the student will work primarily with publicly available climate-model outputs, applying advanced statistical, computational and machine-learning approaches to understand uncertainties in compound weather extremes and their impacts.

Other Information

Dr Shipra Jain – personal research page: https://climatejain.github.io/

UCL Department of Risk and Disaster Reduction: https://www.ucl.ac.uk/risk-disaster-reduction/

UCL Department of Statistical Science: https://www.ucl.ac.uk/mathematical-physical-sciences/statistics

Zscheischler et al. (2020), A typology of compound weather and climate events, Nature Reviews Earth & Environment – the standard framing for this research area: https://www.nature.com/articles/s43017-020-0060-z

Aon, Catastrophe Model Insight – what catastrophe models are and how the sector uses them: https://www.aon.com/en/capabilities/reinsurance/catastrophe-model-insight

Aon, Climate and Catastrophe Insight (annual report) – the loss record this project seeks to explain: https://www.aon.com/en/insights/reports/climate-and-catastrophe-report