Extreme rainfall and temporal loading: Towards more effective design storms

Pluvial flooding poses an increasing global risk to people and property, driven by urban expansion, infrastructure growth, and the intensification of rainfall associated with climate change. While tools to model pluvial flood hazard have also advanced, rainfall continues to be represented in these models using highly simplified forms. This is at odds with the physical complexity of rainfall events, which vary significantly in both space and time. While simplifications are inherent in modelling, there is limited understanding of how reducing the temporal complexity of rainfall affects hydrological outcomes. This gap in knowledge is especially concerning for pluvial flooding, where small-scale variations in rainfall can produce large differences in modelled impacts.

Simplification of rainfall events in flood modelling is generally achieved using design storms.
Design storms are synthetic profiles used to standardise the representation of extreme rainfall events for assessing flood hazard across different return periods. In these profiles a rainfall total is combined with a hyetograph, which dictates how rainfall is distributed over the course of the storm. Although design storms use locally specific estimates of total rainfall, in the UK the same standard hyetograph, with a highly peaked, symmetrical shape, is typically applied in all contexts, and the uncertainty introduced by this is rarely quantified. Looking ahead, intensification of convective storms in future climates may also bring changes in temporal rainfall patterns, further increasing the risk that current modelling practices will diverge from physical reality.

This thesis investigates the sensitivity of flood outcomes to temporal loading, and examines whether shifts in prevailing storm patterns under climate change can be detected using very high-resolution climate simulations. These analyses rely on event temporal loading metrics, categorical or numerical indicators calculated on detailed hyetographs or raw rainfall events. A 2-dimensional (2D), rain-on-grid flood model is applied to two small urban catchments in Leeds, prone to pluvial flooding. The sensitivity testing shows a clear response to temporal loading, with a late-peaking (‘back-loaded’) event causing up to a 25% increase in the flood affected area compared to a front-loaded event of the same size. To assess how these structures may change in future, very high-resolution climate simulations from the United Kingdom Climate Projections (UKCP) Local under Representative Concentration Pathway 8.5 are analysed. Little detectable change in the frequency of different temporal loading patterns is found in the future climate,
but regional differences do emerge, with central and southern England producing more highly asymmetric (both ‘front-’ and ‘back’-loaded’) events in both present and future simulations.

Through flood-modelling and climate-simulation analyses, this thesis identifies that the lack of a single, consistent definition of temporal loading, exemplified by the abundance of related metrics, undermines our ability to compare impacts or characterise rainfall events consistently. This finding motivates a comprehensive review and empirical analysis of existing temporal loading metrics. By organising these metrics around five conceptual dimensions, namely peakiness, rainfall mass asymmetry, peak-timing asymmetry, event concentration and intermittency, a structured framework for selecting metrics aligned with specific research questions and data constraints is established. The resulting framework allows future studies to link more specific aspects of temporal loading to flood hazard or to project how these aspects may evolve under climate change. Furthermore, by clarifying which metrics quantify the same features and which capture distinct characteristics, and by providing an open codebase for consistent metric implementation, this thesis addresses previous methodological transparency in this field, and provides a basis for more fair cross-comparison of studies.