Journal papers 

Abstract 

The devastating 1968 flash flood in the River Chew, South-West of England, serves as a stark reminder of the unpredictable nature of such natural disasters and highlights the importance of natural hazard assessments. The uncertain and often incomplete historical data, and the limited field measurements at the time hindered our understanding of this event. By integrating historical evidence, including technical reports, newspapers, literature, and eyewitness accounts, with advanced hydraulic modelling (HECRAS 2D), this study reconstructs the 1968 flash flood. A sensitivity analysis of the computational methodologies in HEC-RAS, examining various governing equations and numerical methods, introduces an additional dimension to this research. The results verify a maximum flow rate of 165 m³/s at the Compton Dando hydrometric station, marking a 65% increase from the previous official estimate. This update aligns with over 90% of the historical flood marks observed. Findings suggest recalibrating hydrological models, revising risk assessments, and updating flood frequency analyses in the study area. This novel framework confronts the challenges of uncertain and incomplete historical records through a reverse engineering methodology to reconstruct missing peak discharges. The study also presents a new methodological blueprint that can be replicated for reconstructing historical flash flood events in various regions.

Abstract 

The necessity for high-resolution two-dimensional (2D) simulations in flood modelling often requires excessively long simulation times. This study evaluates the impact of various hardware configurations on Hydrologic Engineering Center-River Analysis System (HEC-RAS) 2D with particular emphasis on Central Processing Unit (CPU) speed, number of cores, Random Access Memory (RAM) capacity, addressing a critical gap in the optimisation of hardware setups for time-efficient simulations. These findings are invaluable for flood modellers and the HEC-RAS community, ultimately supporting more effective flood risk management and decision-making. Additionally, the study examines how different meshes, numerical solution methods, and solving equations perform within these hardware setups, aiming to examine the effects of computational techniques on overall simulation efficiency. Our investigations were carried out using both virtual machines on the Google Cloud Platform and a desktop PC. The findings indicate that optimal performance in HEC-RAS 2D simulations does not necessarily correlate with a higher number of cores or increased RAM. Instead, a well-adjusted configuration with 8 cores and 64 GB of RAM demonstrates superior efficiency. This result questions the usual assumptions about the need for greater computational power and emphasises the value of carefully optimising hardware for fast hydraulic modelling.