Flood dynamics derived from video remote sensing

Masafu, Christopher Kusimba (2024) Flood dynamics derived from video remote sensing. PhD thesis, University of Glasgow.

Abstract

Flooding is by far the most pervasive natural hazard, with the human impacts of floods expected to worsen in the coming decades due to climate change. Hydraulic models are a key tool for understanding flood dynamics and play a pivotal role in unravelling the processes that occur during a flood event, including inundation flow patterns and velocities. In the realm of river basin dynamics, video remote sensing is emerging as a transformative tool that can offer insights into flow dynamics and thus, together with other remotely sensed data, has the potential to be deployed to estimate discharge. Moreover, the integration of video remote sensing data with hydraulic models offers a pivotal opportunity to enhance the predictive capacity of these models.

Hydraulic models are traditionally built with accurate terrain, flow and bathymetric data and are often calibrated and validated using observed data to obtain meaningful and actionable model predictions. Data for accurately calibrating and validating hydraulic models are not always available, leaving the assessment of the predictive capabilities of some models deployed in flood risk management in question. Recent advances in remote sensing have heralded the availability of vast video datasets of high resolution. The parallel evolution of computing capabilities, coupled with advancements in artificial intelligence are enabling the processing of data at unprecedented scales and complexities, allowing us to glean meaningful insights into datasets that can be integrated with hydraulic models. The aims of the research presented in this thesis were twofold. The first aim was to evaluate and explore the potential applications of video from air- and space-borne platforms to comprehensively calibrate and validate two-dimensional hydraulic models. The second aim was to estimate river discharge using satellite video combined with high resolution topographic data. In the first of three empirical chapters, non-intrusive image velocimetry techniques were employed to estimate river surface velocities in a rural catchment. For the first time, a 2D hydraulicvmodel was fully calibrated and validated using velocities derived from Unpiloted Aerial Vehicle (UAV) image velocimetry approaches. This highlighted the value of these data in mitigating the limitations associated with traditional data sources used in parameterizing two-dimensional hydraulic models. This finding inspired the subsequent chapter where river surface velocities, derived using Large Scale Particle Image Velocimetry (LSPIV), and flood extents, derived using deep neural network-based segmentation, were extracted from satellite video and used to rigorously assess the skill of a two-dimensional hydraulic model. Harnessing the ability of deep neural networks to learn complex features and deliver accurate and contextually informed flood segmentation, the potential value of satellite video for validating two dimensional hydraulic model simulations is exhibited. In the final empirical chapter, the convergence of satellite video imagery and high-resolution topographical data bridges the gap between visual observations and quantitative measurements by enabling the direct extraction of velocities from video imagery, which is used to estimate river discharge. Overall, this thesis demonstrates the significant potential of emerging video-based remote sensing datasets and offers approaches for integrating these data into hydraulic modelling and discharge estimation practice. The incorporation of LSPIV techniques into flood modelling workflows signifies a methodological progression, especially in areas lacking robust data collection infrastructure. Satellite video remote sensing heralds a major step forward in our ability to observe river dynamics in real time, with potentially significant implications in the domain of flood modelling science.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > T Technology (General); T Technology > TC Hydraulic engineering. Ocean engineering
• Colleges/Schools: College of Science and Engineering > School of Geographical and Earth Sciences
• Supervisor's Name: Williams, Professor Richard and Hurst, Dr. Martin D.
• Date of Award: 2024
• Depositing User: Theses Team
• Unique ID: glathesis:2024-84036
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Jan 2024 12:25
• Last Modified: 19 Jan 2024 14:44
• Thesis DOI: 10.5525/gla.thesis.84036
• URI: https://theses.gla.ac.uk/id/eprint/84036

Related URLs

• Article DOI
• Article DOI
• Article DOI
• Data DOI
• Data DOI
• Data DOI