This thesis explores the use of machine learning and uncertainty quantification methods to improve prediction and forecasting for extreme weather events, with a focus on storm surges,
coastal inundation and road flooding. Machine learning-based approaches, such as Artificial Neural Networks combined with Bayesian model selection and Monte Carlo simulations, are
employed to enhance the accuracy of storm surge forecasts by propagating input uncertainty and providing confidence intervals. The methodology is applied to a case study in Millport, Scotland, demonstrating improved predictive performance and computational efficiency, with Pearson correlation coefficient of 0.942 for 24-hour surge forecasts. In the context of coastal inundation, a framework is presented that incorporates aleatoric and epistemic uncertainties,
with operational validation during Storm Ciara in the Firth of Clyde, showing its effectiveness in addressing complex coastal flood risks. Additionally, the thesis addresses the fragility of Scotland's trunk road network to disruption from precipitation events, and particularly the development of empirical fragility curves to quantify the vulnerability of transportation infrastructure. The analysis, based on data from Transport Scotland, SEPA, and NIMROD, provides insights into the potential impacts of extreme weather on critical infrastructure, with a focus on uncertainty at each stage of the forecasting process. The thesis concludes with reflections on the challenges and potential improvements to these methodologies for future climate resilience and infrastructure planning.
| Date of Award | 15 Sept 2025 |
|---|
| Original language | English |
|---|
| Awarding Institution | - University Of Strathclyde
|
|---|
| Sponsors | University of Strathclyde & EPSRC (Engineering and Physical Sciences Research Council) |
|---|
| Supervisor | Enrico Tubaldi (Supervisor) & Edoardo Patelli (Supervisor) |
|---|