A Bayesian network model for bluff retreat on the southern Lake Erie coast, United States

Research output: Contribution to journalArticlepeer-review

2 Scopus citations


Coastal bluffs on the 73 km Pennsylvania mainland coast of Lake Erie consist of unconsolidated glacial through paleo-lacustrine Quaternary strata overlying Devonian shale bedrock. Erosion is a significant environmental hazard on this coast where the long-term average annual retreat rate at the bluff crest is ∼ 0.15 m/yr. As the primary geomorphic feature used to track land loss on high-relief coasts, understanding the processes that drive bluff-crest retreat and concomitant sediment and infrastructure losses remains a challenge in coastal-hazard management. This paper shows that bluff retreat on the Lake Erie coast of Pennsylvania can be successfully modeled using multivariate statistics that may have application to bluff settings across the Great Lakes basin and globally. A Bayesian Network (BN) model is developed using long-term historical data (1938–2007 training datasets) and validated against subsequent crest retreat mapped from recent lidar data. Seven field sites were sampled for environmental data that collectively covered ∼ 30 % of the 33.5 km length of the western Erie County littoral cell (WECLC). A BN with eight inputs correctly predicted bluff retreat rates for 95.4 % of transects and had a mean posterior predictive probability of 84.1 %. Scenarios show that greater crest-retreat rates were more likely to be associated with greater wave impact hours, less-resistant stratigraphy, narrower beaches, lower and steeper bluffs, a more energetic run-up regime, and a lower or absent bedrock-toe ledge. These diagnostic attributes should be useful across the WECLC to help identify, with finer spatial resolution, sectors with higher probabilities of having or developing significant erosion problems.

Original languageEnglish (US)
Pages (from-to)387-405
Number of pages19
JournalJournal of Great Lakes Research
Issue number2
StatePublished - Apr 2023

All Science Journal Classification (ASJC) codes

  • Ecology, Evolution, Behavior and Systematics
  • Aquatic Science
  • Ecology

Cite this