Estuarine Back-barrier Shoreline and Beach Sandline Change Model Skill and Predicted Probabilities: Event-driven beach sandline change

The Barrier Island and Estuarine Wetland Physical Change Assessment was created to calibrate and test probability models of barrier island estuarine shoreline (backshore) and beach sandline change for study areas in Virginia, Maryland, and New Jersey. The models examined the influence of hydrologic and physical variables related to long-term and storm-derived overwash and back-barrier shoreline change. Input variables were constructed into a Bayesian Network (BN) using Netica, a computer program created by NORSYS Software Corporation that allows users to work with belief networks and influence diagrams. Each model is tested on its ability to predict changes in long-term and event-driven (i.e., Hurricane Sandy-induced) backshore and sandline change based on learned correlations from the input variables across the domain. Using the input hydrodynamic and geomorphic data, the BN is constrained to produce a prediction of an updated conditional probability of backshore or sandline change at each location. To evaluate the ability of the BN to reproduce the observations used to train the model, the skill, log likelihood ratio and probability predictions were utilized. These data are the probability and skill metrics for the event-driven beach sandline change model.

本数据集为障壁岛与河口湿地物理变化评估（Barrier Island and Estuarine Wetland Physical Change Assessment），旨在为弗吉尼亚州、马里兰州及新泽西州的研究区域校准并测试障壁岛河口岸线（后滨）与海滩沙线变化的概率模型。该模型探究了与长期及风暴引发的越流、障壁后岸线变化相关的水文与物理变量的影响。输入变量通过NORSYS软件公司开发的计算机程序Netica构建为贝叶斯网络（Bayesian Network, BN），该程序支持用户构建信念网络与影响图。基于研究域内输入变量间习得的相关性，各模型均针对其预测长期及事件驱动（即飓风桑迪引发的）后滨与沙线变化的能力进行测试。依托输入的水动力与地貌数据，贝叶斯网络（BN）经约束后可生成各位置后滨或沙线变化的更新条件概率预测结果。为评估贝叶斯网络重现模型训练所用观测数据的能力，本研究采用了技能评分、对数似然比与概率预测三类指标。本数据集包含事件驱动型海滩沙线变化模型的概率与技能度量指标。