Determining the coastal erosion vulnearbility of Australia using Bayesian networks

Mr Yongjing Mao1, Dr Daniel Harris1

1The University Of Queensland,

Abstract:

Evaluating coastal vulnerability on different spatial-temporal scales is critical for effective coastal management. Local-scale coastal response to environmental and anthropogenic impacts has been frequently assessed with numerical models, however, large-scale evaluation typically relies on data-driven methods such as Discrete Bayesian networks (BNs). Here, we used a new method, the Hybrid Bayesian network, to assess large-scale (continental scale) coastal response to environmental changes. Both Discrete and Hybrid BNs were developed, tested and compared using Digital Earth Australia datasets, and coastal erosion vulnerability was evaluated. These BNs used forcing parameters (e.g. waves, tide, sediment sink/source, and sea level rise) and geologic constraints (e.g. geomorphology, backshore profile and surfzone slope) to predict shoreline retreat rate. Validation showed that Hybrid BNs, which provide a more realistic assessment of the range of shoreline retreat rate, outperform in predicting continuous variables, when compared with Discrete BNs. Discrete and Hybrid BNs provide consistent qualitative findings for the coastal vulnerability of Australia. Among forcing parameters, the sediment sink/source was found to be the most informative variable to indicate coastal vulnerability, followed by tide, relative sea level rise and wave processes. In the scenario of an increased sea level rise rate, tropical tidal flats were predicted as the most at risk in Australia. We find that BNs can reflect impact of different factors on coastal evolution, and predict future shoreline change by exploring historical data. The performance of these models can be further improved when more and better datasets become available.

 


Biography:

As a PhD candidate in UQ, I am working on evaluating coastal vulnerability on large scales (e.g. continental or global)  using data driven models. With remote sensing methods, I create global scale datasets that describe historical coastal evolution and indicate the change in the future. Meanwhile, I am exploring these newly-created and previously-existing datasets with data driven models to determine the coastal vulnerability in both data-rich and data-poor regions.