Turning fragmented climate data into decision tools: who is most at risk, and how sure are we?

About the project

About the Project Background Climate change does not put everyone in the same danger. Whether a heatwave, flood or drought becomes a disaster depends on where the hazard lands and on who and what is exposed. Answering “where is risk worst, and for whom?” means pulling together data that was never meant to be combined: climate projections, satellite records, census and economic statistics, health data, and text from policy documents and news. Each is patchy, measured differently, and uncertain. Researchers use data science (machine learning, text mining, data fusion) to turn these fragments into comparable risk indicators, and decision science (multi-criteria analysis, preference elicitation) to weigh them into a picture that reflects what people care about. National risk portals and composite indices exist, but they tend to bury their assumptions, treat uncertain inputs as exact, and give users little to interrogate. The open problem is making climate risk assessment transparent, uncertainty-aware and usable. PhD project This PhD builds both the methods and a working decision tool to answer a deceptively simple question: as the climate changes, who is most at risk, and how confident can we be? The student picks one sector and question early on, for example urban heat, agricultural drought, flood exposure, climate-linked migration or infrastructure resilience, and co-develops the project from there. Three linked challenges run through it. Data: find and pull together open sources that do not agree on units, resolution or coverage, and use web scraping, text mining, geospatial processing and data fusion to convert them into quantified risk indicators, carrying the uncertainty in each rather than discarding it. Integration: apply multi-criteria decision analysis and preference elicitation to combine indicators into a composite risk score, and stress-test how the ranking of high-risk places shifts under different weightings and under input uncertainty, so the result is honest about what it does and does not know. Tool-building: design an interactive geospatial dashboard that lets a non-technical user see why a place scores as it does, change assumptions, and compare scenarios. The project connects two UNRISK themes, data science and decisions/communication, and is co-supervised by Dr Andrea Taylor, whose work on how people perceive and act on climate and weather risk will shape the elicitation and the dashboard. There is scope to work with an operational or policy partner. The student leaves with reusable methods for uncertainty-aware risk aggregation, an open-source decision-support tool, and evidence on how weighting and uncertainty change which places count as most at risk, plus skills spanning data engineering, decision modelling and stakeholder work. Applicant Profile Applicants with a strong quantitative background, in statistics, data science, engineering, computer science or similar, who want to point those skills at climate and environmental risk. You should be happy programming (for example Python or R), wrangling and joining messy data from many sources, dealing with missing values, and building interactive tools/dashboards. No prior climate or decision science is expected; curiosity about working across disciplines, and about getting tools into the hands of decision makers, matters more. Training needs are planned with you in the first year.

Research areas