
Project Description
Long-runout geophysical flows, including debris flows and glacial lake outburst floods (GLOFs), can propagate rapidly over large distances and generate cascading impacts far from their source. Their evolution is controlled by complex interactions among topography, water, sediment availability and changing flow conditions, making both prediction and reconstruction challenging.
Physics-based models can reproduce these processes, but their computational cost limits rapid and large-scale application, while predictions are highly sensitive to poorly constrained initial and boundary conditions. Many of these uncertain parameters, however, may leave observable signatures in Earth observation, seismic and other indirect data.
This PhD will develop artificial-intelligence methods to reconstruct and predict the evolution and impacts of long-runout flows. The student will combine multi-temporal satellite imagery, digital elevation models, hydroclimatic information, seismic observations and mapped event impacts. Models will estimate flow pathways, velocity, inundation extent and potential exposure. Past events will be reconstructed from spatial and temporal evidence to identify conditions associated with exceptional mobility.
The research will investigate how information from well-observed events can improve prediction in regions with limited data. Potential directions include multimodal learning from Earth observation and seismic signals, reconstruction of different flow phases, prediction shortly after initiation, and transfer of models between debris flows, GLOFs and contrasting mountain environments. Particular attention will be given to uncertainty, incomplete observations, model explainability and physical plausibility.
The project will improve understanding of long-runout flow behaviour and develop AI-based approaches that allow for the prediction of flow evolution, propagation and impacts on exposed communities and infrastructure.
Project Specific Training
The student will develop skills in artificial intelligence for Earth observation, spatiotemporal prediction, multimodal learning, uncertainty analysis and explainable AI. They will learn to work with optical and radar satellite imagery, digital elevation models, seismic time series and environmental datasets, alongside scientific programming, cloud geospatial analysis and reproducible research. Through supervised computational work, they will reconstruct past flows, predict propagation and impacts, evaluate transferability across events, and assess the physical plausibility of AI outputs. Development will be supported through regular supervision, research-group participation, specialist courses, conferences, workshops and summer schools.
Potential Career Trajectory
The project will prepare the student for academic research in artificial intelligence, Earth observation, natural hazards, geospatial science and environmental forecasting. Its combination of satellite analysis, seismic-data interpretation, multimodal AI, uncertainty assessment and hazard modelling will also support careers in geological surveys, space agencies, environmental and engineering consultancies, catastrophe-risk modelling, civil protection, government and international disaster-risk organisations. The student will develop transferable skills in scientific programming, high-performance computing, reproducible research and the communication of complex predictions to technical and non-technical audiences. These capabilities are also applicable to wider data-science and technology-sector careers.
Access and Inclusion Statement
This is primarily a computational and data-driven project and does not require routine fieldwork, overseas travel or physically demanding activities. Core research can be completed using remotely accessible Earth observation, seismic, environmental and event datasets, supported by university computing facilities. Optional site visits, conferences or workshops will be encouraged and may involve travel, but remote participation or equivalent research activities can be arranged where required.
Reasonable adjustments may include flexible or hybrid working, adapted computing equipment and software, accessible training and meeting formats, additional preparation time, and flexibility around caring responsibilities or health-related needs. Any optional fieldwork would be planned collaboratively with the student, with accessibility, safety and individual requirements considered from the outset.
The balance between Earth observation analysis and AI modelling can be adapted to the student’s circumstances without compromising the central research objectives. The supervisory team will work with the student and relevant university services to identify appropriate adjustments and ensure equitable participation in research, training, conferences and other professional-development activities.
Project supervisor/s
Tom Matthews
Geography
King’s College London
tom.matthews@kcl.ac.uk
https://www.kcl.ac.uk/people/tom-matthews
Lorenzo Nava
Geography
King’s College London
lorenzo.nava@kcl.ac.uk
