Global detection and analysis of surface water and river meander dynamics using cloud-based remote sensing

Abstract

Surface water features, such as rivers and lakes, are among the most dynamic features on Earth. Processes including river meandering, reservoir construction, and lake desiccation exhibit complex spatial and temporal dynamics influenced by climate, topography, land characteristics, and human activities. Monitoring and analysing the spatio-temporal variability of surface water extent requires long-term observations, and satellite remote sensing time series provide a potential solution to this need, offering decades of global coverage. However, the use and interpretation of such data have historically been constrained by computational limitations. This research aims to develop pixel-based algorithms that track surface water dynamics on a global scale. These methods enable investigations into a range of associated surface processes, including meander migration, a key focus of this thesis. Here, the cloud-based platform Google Earth Engine (GEE), which provides the necessary computational power to process vast amounts of satellite data efficiently, is used to underpin the development of two main new contributions. Firstly, GEE is used to develop the Water Transition Timing (WTT) dataset, which maps, for the first time, the timing of permanent surface water advance and recession events during the period 1984 to 2022 globally, encompassing both inland water bodies and coastal regions. Analysis of the WTT dataset shows that, since 2005, the expansion of surface water has outpaced the recession, largely driven by human activities in Asia, particularly dam construction, the growth of flood irrigation, and the indirect impacts of climate change, especially across the Tibetan Plateau. Secondly, a pixel-based algorithm is developed to detect subtle rates of erosion and sedimentation in meandering rivers across 4,923 river bends worldwide, an unprecedented scale of analysis achieved for the first time using satellite remote sensing. This algorithm facilitates the systematic investigation of a long-standing geomorphic question: whether river bend migration is driven predominantly driven by bank pull (whereby outer bank erosion precedes sediment deposition along the inner bank) or bar push (whereby initial inner bank sedimentation forces subsequent outer bank erosion) in real environments. The results indicate that bar push processes primarily occur in less erosive, slow-flowing ,rivers that have high boundary resistance through dense vegetation and a higher proportion of coarser boundary materials. In contrast, bank pull is more prevalent in more erosive fast flowing, non-vegetated, rivers with finer bank materials. This thesis has developed methods to detect temporal changes in the extent of surface water bodies using remote sensing and Google Earth Engine, thereby expanding the understanding of global surface water dynamics and meandering rivers. In addition to the algorithms and results, this research has produced publicly available datasets that can be used in the future by researchers to study the impacts of climate change on water resources, by government agencies to monitor newly formed artificial lakes, and by NGOs to identify communities affected by river erosion; ultimately enhancing water management at both local and global scales.

More information

Identifiers

ORCID for Gustavo Willy Nagel: orcid.org/0000-0002-8060-0772

ORCID for Steve Darby: orcid.org/0000-0001-8778-4394

ORCID for Julian Leyland: orcid.org/0000-0002-3419-9949

Catalogue record

Export record