Presenter: Erika Lentz (USGS)

Description:
Coastal landscapes evolve in response to sea-level rise (SLR) through a variety of geologic processes and ecologic feedbacks.  When SLR surpasses the rate at which these processes build elevation and drive lateral migration, inundation is likely.  A probabilistic modeling (Bayesian network) approach provides computationally efficient ways of generating broad scale (regional) assessments of first-order sea-level rise impacts (i.e. the likelihood of inundation vs. dynamic coastal response) over landscapes.  In addition to efficiency, an advantage to this approach is that predicted outcomes account for uncertainties related to future SLR projections, processes, feedbacks, as well as error associated with spatially extensive, continuous datasets.   Results show that 44% and 36% of low-lying, habitable land along the northeastern Atlantic coast of the United States is likely to dynamically respond to SLR by the 2050s and 2080s, respectively.  Sensitivity and skill evaluations show that error is greatest in the lowest elevation areas but that this error has little impact on prediction skill, as inherent correlations between landcover and elevation datasets are exploited to reduce uncertainty through Bayesian inference.  The approach remains flexible and easily updatable, so that as new data (e.g., SLR projections) and information (e.g., landcover transition thresholds) becomes available it can be ingested to update outcomes.  More broadly, we have found that model sensitivity in these first-order landscape change assessments is well-matched to larger coastal process uncertainties, for which more detailed process-based models provide important complements.  Outcomes have been incorporated in decision-support web utilities exploring both coastal resiliency and conservation, and alongside process-based models, this approach is being used to build and expand a national-scale portfolio of U.S. Geological Survey future coastal hazards products and data.