A Bayesian Statistical Network Model of Bluff Retreat and a Littoral Sediment Budget Model for the Western Erie County Littoral Cell, Lake Erie
This project uses a combination of coastal and stratigraphic mapping at selected field sites; wave climate and sedimentologic data; GIS change-detection analysis; and Bayesian statistical modeling to improve understanding of the relationships between coastal processes, hazards, and sediment supply associated with bluff retreat on the Lake Erie coast. The project quantifies relationships between physical processes and landscape responses that may help improve community resiliency against a major coastal hazard on the Great Lakes. Seven field sites of ~1-2 km in length were selected for Bayesian modeling of bluff retreat that relied on statistical analysis of 2007 and 2015 lidar data and numerous environmental parameters to determine the principal causes of bluff retreat. Linking these detailed-study sites and extending from the OH-PA state line to the downdrift end of the WECLC at Presque Isle State Park, GIS/lidar mapping was used to quantify 2007-2015 bluff contributions to the littoral sediment budget.
Objective 1: Develop and apply a multivariate Bayesian statistical network model of bluff retreat to the western Erie County littoral cell (WECLC). The model will be a predictive tool for explaining recent-to-historical bluff retreat and patterns, and for simulating future retreat magnitudes and patterns over multi-decade periods through 2065.
Objective 2: Generate up-to-date GIS/lidar-derived estimates of bluff-sourced (a) total-sediment input (clays to boulders) to the WECLC that impacts both coastal water quality and beach resources; and (b) littoral sediment input (sand to boulders; or “sand+”) that represents the principal coarse sediment input to the WECLC and to the Presque Isle littoral cell immediately downdrift.
High-resolution continuous mapping of the lakefront bluffs was completed wherein crest elevations were obtained at <1 m intervals, and bluff-face topographic changes (from which sediment losses were derived) were mapped every ~1 m2, along the 33.5 km WECLC coast. These data represent a valuable addition to the regional geo-environmental knowledge base and allow improved understanding of bluff behavior. The project used Bayesian Network modeling to explain past bluff-retreat patterns and simulate future bluff-crest retreat through 2065. The importance of bluff retreat as a contributor of sediment to the littoral system during average lake-level conditions was established using change-detection analysis.
The optimal Bayesian Network model used the following eight inputs: SPR resiliency, long-term retreat rate, bluff face slope, beach prism width, toe elevation, top-shale elevation, bluff height, and wave impact hours. Fitting the final model correctly predicted the 2007-2015 retreat-rate bin for 95.4% of the transects (correct-classification rate). The predicted value was assumed to correctly match if the observed 2007-2015 retreat rate matched the bin with the largest predicted posterior probability. It was also assumed correct if the largest predicted posterior probability was tied among multiple bins (2 bins in 80 cases, 3 bins in 9 cases) and the observed 2007-2015 retreat rate was among those bins. When ties were excluded, the model predicted 71.5% of the short-term rates correctly. The mean predicted posterior probability of the observed 2007-2015 retreat rate was 84.1%. The model suggests that long-term retreat rate, bluff face slope, toe elevation and beach-prism volume together explain most of the predicted 2007-2015 crest-retreat rates. Groundwater flux within the model appears to have only a minor influence. Over the next 50 years, bluff-crest retreat was simulated to range from 1 to 15 m by site. That represents a range of crest retreat rates of 0.02 to 0.3 m/yr, within the range of values for historical bluff retreat.
The 8-year total-sediment and sand+ changes for the WECLC bluffs were net losses of 318,400 m3 and 105,850 m3, respectively. Lakefront bluffs in the Crooked Creek and Trout Run watersheds were the principal sediment-supply sources. Annualized, the bluffs supplied 39,800 m3/yr of total-sediment and 13,250 m3/yr of sand+ to the WECLC. Estimated sand+ yields were ~430 m3/bluff km/yr averaged across the six WECLC watersheds. The 13,250 m3/yr sand+ supply results in a small littoral-sediment transport rate by both ocean coast and Great Lakes standards: sediment input from streams is minimal. The sand+ sediment supply reflects change within a longer ~12-year period of relatively low and steady lake levels (1999-2011) followed by a 3-year slow transgression (2012-2015).