Research

An analysis of trends in the magnitude and frequency of extreme rainfall in the UK, for durations 1 to 16 days. A spatial extreme value model is fitted to improve the detection of any signal of non-stationarity, test the explanatory power of climatic covariates and explore the spatial variability of trends. While most individual rain gauges show no significant trend, there is evidence of non-stationarity across the dataset as a whole.

This study offers practical recommendations for the application of different NbS interventions in hydraulic modelling and outlines the corresponding parameter ranges to consider. To demonstrate the real-world utility of these recommendations, a case study of the River Chew catchment (England) is presented, detailing pre- and post-intervention configurations wherein multiple candidate areas are converted to wetland and woodland using a HEC-RAS 2D model.

Risk managers are having to plan and make decisions for an uncertain future. This case study of the application of an adaptation pathway approach at Hurst Spit provides an important example of how to embrace uncertainty in decision-making. The approaches and model developed are transferrable across geographic locations, industries and sectors.

Shingle beaches often act as primary soft defences against coastal flooding. Unhealthy beaches can reduce protection, prompting designers to increase beach volume to improve overtopping performance. However, no tools currently exist to assess wave overtopping on seawalls fronted by dynamic shingle beaches, creating design uncertainty. To address this, the University of Plymouth and JBA Consulting are conducting a PhD project combining physical and numerical studies to develop overtopping and force formulations for design applications.

Accurate shoreline‑change prediction is essential for coastal management under climate variability. This study assesses deep learning methods, using LSTM and CNN‑LSTM models with direct input and autoregressive approaches. All models captured temporal change, with autoregressive methods performing best. Noise reduction through seasonal decomposition and filtering improved accuracy, especially with 52‑week data. Lower‑resolution datasets reduced performance, requiring at least fortnightly data. Ensemble modelling produced effective quasi‑probabilistic forecasts, highlighting the value of AR architectures and high‑quality data.

Convection rainfall is a major driver of hydrological extremes. This article highlights the value of convection‑permitting modelling (CPM) for managing flood risk and urban pollution from sewer systems. It presents CPM applications for flash‑flood guidance in Sierra Leone and for a UK rainfall‑perturbation tool helping water companies plan for future conditions. Both examples emphasise co‑design with end users.