Main content area

A new approach for river network classification based on the beta distribution of tributary junction angles

Jung, Kichul, Shin, Ju-Young, Park, Daeryong
Journal of hydrology 2019 v.572 pp. 66-74
drainage, rivers, statistics, support vector machines
Different river networks in nature have distinct features depending on regional constraints; these affect the development of the river network and form different characteristics. The differences in drainage networks help in distinguishing between various network types and understanding natural processes. The aim of this study is to develop a new and simple classification method for determining various river network types based on a critical parameter of river networks, the tributary junction angle. For the analysis, fifty river networks are predefined as five network types (dendritic, parallel, pinnate, rectangular, and trellis networks). The tributary junction angles are calculated for each drainage network type, and then the beta distribution is employed to identify distributional characteristics of the junction angles. Parameter estimates of the beta distribution are used to classify different river networks because the estimates provide classified features of natural processes. Support Vector Machines are then utilized to determine the network classification with the parameter estimates. Furthermore, the results are validated against classification using a different approach with commonly used statistics. The proposed method can clearly distinguish between different network types, except for the rectangular and trellis types. In addition, parameter estimates of the beta distribution indicate differences in drainage network types more clearly than the statistics of the angles. Overall, the parameter estimates of beta distribution have high potential for application to the classification of river networks.