USING A SELF-ORGANIZING MAP TO ANALYZE HYDRAULIC FEATURES OF VERMONT STREAMS TO GAIN INSIGHT INTO FLOOD RESPONSE

As climate change alters the frequency and intensity of flooding, water resource scientists and engineers need new, powerful tools that can be applied at broad scales to characterize stream reaches and their susceptibility to flooding. The hydraulic radius (R_H) of a stream reach is a relatively simple hydraulic metric that can be extracted from high-resolution digital elevation models to provide insight into a reach’s flow efficiency and potential for overbank flooding. The hydraulic radius is defined as the cross-sectional flow area divided by the wetted perimeter, and a plot of the derivative of hydraulic radius (R_H’) versus river stage provides information on floodplain-channel dynamics. We used an unsupervised machine learning algorithm to cluster features extracted from R_H’-stage plots for 3,057 stream reaches across the Lake Champlain Basin of Vermont. The Self-Organizing Map (SOM) is a powerful computational tool that has unique advantages for handling large data sets consisting of multivariate observations, varying data types, and nonparametric data. Reaches within a given were similar to each other in how they respond to flow events at varying stages, which represents valuable information for water resource scientists. Scientists engaged in flood prediction and flood mitigation can use this information to anticipate the flooding behavior of remote reaches, or other reaches that cannot been directly observed in the field. This study highlights the usefulness of the SOM for analyzing and interpreting data from complex, nonlinear hydrogeomorphic processes.