Using a Bayesian network to predict barrier island geomorphologic characteristics

Abstract

Quantifying geomorphic variability of coastal environments is important for understanding and describing the vulnerability of coastal topography, infrastructure, and ecosystems to future storms and sea level rise. Here we use a Bayesian network (BN) to test the importance of multiple interactions between barrier island geomorphic variables. This approach models complex interactions and handles uncertainty, which is intrinsic to future sea level rise, storminess, or anthropogenic processes (e.g., beach nourishment and other forms of coastal management). The BN was developed and tested at Assateague Island, Maryland/Virginia, USA, a barrier island with sufficient geomorphic and temporal variability to evaluate our approach. We tested the ability to predict dune height, beach width, and beach height variables using inputs that included longer-term, larger-scale, or external variables (historical shoreline change rates, distances to inlets, barrier width, mean barrier elevation, and anthropogenic modification). Data sets from three different years spanning nearly a decade sampled substantial temporal variability and serve as a proxy for analysis of future conditions. We show that distinct geomorphic conditions are associated with different long-term shoreline change rates and that the most skillful predictions of dune height, beach width, and beach height depend on including multiple input variables simultaneously. The predictive relationships are robust to variations in the amount of input data and to variations in model complexity. The resulting model can be used to evaluate scenarios related to coastal management plans and/or future scenarios where shoreline change rates may differ from those observed historically.

Smartphone-Based Distributed Data Collection Enables Rapid Assessment of Shorebird Habitat Suitability

“…These considerations notwithstanding, the overall accuracy of the geolocation data was sufficient as input into our multi-variate models (e.g., [13, 14]), which are generally robust to uncertain data.…”

“…These included fields for nest Site ID, Geomorphic Setting, Substrate Type, Vegetation Type and Vegetation Density. The selections for the biogeomorphic characterization are based on standard classifications (e.g., [12]) and previous work [13, 14]. …”

“…The science problem underlying the iPlover app was defined by previous work that characterizes and predicts landscape change (e.g., [14, 15]) and piping plover habitat utilization [13]. These approaches used Bayesian networks (BN), which implement fundamental principles of prediction and data assimilation which are well-suited to uncertain observations that can occur in the input data (e.g., geolocation) and subjective classifications (e.g., biogeomorphic characterization) used here.…”

UAS-SfM for Coastal Research: Geomorphic Feature Extraction and Land Cover Classification from High-Resolution Elevation and Optical Imagery

“…Profile-based feature extraction techniques have been applied to create consistent datasets of coastal change for uses such as vulnerability assessment and coastal modeling [34,57], which are dependent on the availability of elevation data with sufficient temporal resolution and data density to reflect current conditions. The profile-based shoreline position extraction technique employed here was originally designed for elevation point clouds produced by NASA ATM, which typically produced 15-20 elevation points suitable for use in analysis within a 2-m wide swath [21].…”

Statistical modeling of the long-range dependent structure of barrier island framework geology and surface geomorphology

“…Gutierrez et al (2015) used a Bayesian network to predict barrier island 123 geomorphic characteristics and argue that statistical models are useful for refining predictions of 124 locations where particular hazards may exist. These examples demonstrate the benefit of using 125 statistical models as quantitative tools for interpreting coastal processes at multiple spatial and 126 temporal scales (Hapke et al, 2016).…”