Regional Flood Models and Digital Elevation Model (DEM) Uncertainty

Abstract

Flooding is a natural hazard that affects millions of people throughout the world every year. Hydrodynamic models are a key tool in delineating current and future flood hazard, and provide a key resource for decision makers to reduce flood risk. However, hydrodynamic models need data to drive them, with many parts of the world not having high-quality data at a high-resolution. These areas are considered data-sparse. Data-scarcity is partially characterised by a lack of high-resolution (<30m) topographic data, with this elevation data previously shown to be a key control on the propagation of a flood. Therefore, hydrodynamic models at the intermediate scale (270-1000m) are needed to estimate flood hazard given the lack of high-resolution data, computational resources and a Monte Carlo approach to estimate uncertainties in predictions. This thesis presents three results chapters assessing the ability of an intermediate scale hydrodynamic model to estimate flooding in a large river delta, before going on to establish the effects of uncertain topographic information of flood predictions and connectivity between river channels and floodplains. In the first chapter, an intermediate scale hydrodynamic model of the Mekong Delta is built, with results showing that a model at this scale has a good level of skill and thus a similar approach could be used to estimate flooding in other data-sparse river deltas. However, it is shown that uncertainty in topography from the global digital elevation models (DEMs) had a large influence on flood predictions. This finding inspired the subsequent chapters. Chapter 2 characterised the spatial error structure of floodplains in two global DEMs, and used these relationships to simulate plausible versions of the DEM. By using DEM ensembles, probabilistic flood hazard maps could be produced, with these maps avoiding the spurious precision compared to flood maps that use a single deterministic DEM. Chapter 3 further explored the influence of DEMs by developing a novel method to quantify river-floodplain connectivity across scales and DEM products. Results demonstrated that the DEM product had more influence on river-floodplain connectivity than scale, with the quantification of river-floodplain connectivity shown to be a useful indicator of the appropriateness of a DEM to be used in a hydrodynamic model. This thesis has subsequently enhanced our understanding of the skill of hydrodynamic models at the intermediate scale to model flooding in large data-sparse river deltas, as well as improving our understanding of the impact uncertain topography has on flood predictions by promoting the treatment of topography as a probabilistic entity as opposed to a deterministic one as it currently is.

Date of Award	25 Jun 2019
Original language	English
Awarding Institution	University of Bristol
Supervisor	Jeff Neal (Supervisor) & Paul D Bates (Supervisor)