Plant, N. G., USGS, St. Petersburg, USA, email@example.com
Sallenger, A. ., USGS, St. Petersburg, USA, firstname.lastname@example.org
Howd, P. ., USGS, St. Petersburg, USA, email@example.com
Stockdon, H. ., USGS, St. Petersburg, USA, firstname.lastname@example.org
Holland, K. T., NRL, Stennis Space Center, USA, email@example.com
BAYESIAN-PREDICTION APPROACH APPLIED TO COASTAL MORPHODYNAMICS
The processes that drive coastal evolution (i.e., morphodynamics) are extremely sensitive to small changes in bathymetry and topography. This is particularly true for barrier island response to large storms. In this case, key morphodynamic variables include dune height, dune width, beach slope, inlet widths and inlet depths. Even if we had perfect models for predicting flows due to waves and winds and correspondingly perfect models for predicting sediment transport, uncertainties in the initial geomorphic variables would likely lead to large uncertainties in the predicted evolution. Because, existing flow and sediment-transport models are not perfect, there are large uncertainties in all aspects of morphodynamic-prediction. A probabilistic approach is required to cope with the uncertainties associated with morphodynamic prediction. We use a Bayesian network learn the predictable part of coastal evolution and to track prediction uncertainties. The approach assimilates both model predictions and field observations. Predictions take the form of probability distributions for a reduced set of key variables (e.g., dune height and width) that describe barrier island response to storms. The probabilistic approach yields forecast uncertainties for the key morphologic variables. The predictions will be tested against field observations.
Presentation is given by student: No