Development and Application of New Stream Geomorphic Stability Assessment Procedure in Urbanizing Watersheds Using Advanced Machine Learning

Abstract:

Identifying regime conditions from channel morphology parameters is an important but currently limited practice for characterizing alluvial rivers. Developing empirical equations is the most common method for establishing relationships between width (W), depth (H), slope (S) and shear stress () to flow resistance and sediment transportation equations. Such relationships that are seen in existing published sources for alluvial rivers are limited by the setting, size and range of the dataset used to develop the empirical equations and where good quality and comprehensive field data are not available. A database of channel morphology variables including W, H, S, flow (Q₂) and bed material median particle diameter (D₅₀) for 815 regime channels was compiled from existing publications for the development of a Group Method of Data Handling (GMDH) model. Model performance for predicting W, H, S and was tested by evaluating error statistics. Application of the model was tested by inputting regional flow and sediment data to predict channel conditions of well-known watercourses in Southern Ontario, Canada. The predicted results from the model accurately predicted regime channel conditions when compared with published Rapid Geomorphic Assessments (RGAs) and Qualitative Habitat Evaluation Index (QHEI) scores. These models can be used to establish channel conditions and inform the implementation of strategies to mitigate channel degradation, including Low Impact Developments (LIDs).