Determining the Shoreline Retreat Rate of Australia Using Discrete and Hybrid Bayesian Networks. Issue 6 (16th June 2021)

• Record Type: Journal Article
• Authors: Mao, Yongjing; Harris, Daniel L.; Callaghan, David P.; Phinn, Stuart
• Is Part Of: Journal of geophysical research. Volume 126:Issue 6(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-06-16
• Subjects: Bayesian network -- shoreline retreat -- sea level rise; Geomorphology -- Periodicals; 551.3
• DOI: 10.1029/2021JF006112
• Languages: English
• ISSNs: 2169-9003

Abstract

Abstract: Evaluating shoreline retreat rate (SRR) on different spatial‐temporal scales is critical for effective coastal management. Large‐scale evaluations typically rely on data‐driven methods such as Discrete Bayesian networks (BNs). However, these BNs require discretization of continuous variables which can lead to information loss. Here, we propose a new method, the Hybrid BN to incorporate continuous variables without discretization. Both Discrete and Hybrid BNs were developed and compared to evaluate large‐scale (continental scale) SRR in Australia, using Digital Earth Australia data set. These BNs used forcing parameters (e.g., waves, tide, sediment sink/source, and sea level rise [SLR]) and geomorphic settings (e.g., geomorphology, backshore profile, and surfzone slope) to predict SRR. Validation of the BNs showed that Hybrid BNs, which provide a more realistic assessment of the range of SRR, outperform in predicting continuous variables, when compared with Discrete BNs. However, Discrete and Hybrid BNs provide consistent qualitative findings for the SRR of Australia. Among forcing parameters, the sediment sink/source was found to be the most informative variable to indicate the shoreline retreat, followed by tide, SLR rate, and wave processes. In the scenario of an increased SLR rate, tropical tidal flats were predicted as the most at risk coasts in Australia. We found that BNs can reflect the 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 data sets become available. Key Points: Hybrid Bayesian networks outperform discrete ones in predicting continuous shoreline retreat The distance to inlets is the most significant in determining shoreline retreat rate, followed by tidal impacts and sea level rise (SLR) In Australia, tropical tidal flats are the most sensitive coasts to the increase in SLR rate