An improved method for sand wave morphology discrimination in rivers by combining a flow resistance law and support vector machines

Abstract

A parameterized expression for sand wave morphology in rivers is established using a flow resistance law while accounting for sediment incipient velocity. A distinct relation is drawn between the proposed characteristic parameters and the sand wave morphology based on flume data. Support vector machines (SVMs) are then used to separate the boundaries of the sand wave morphology due to the high classification accuracy of SVMs. The boundary line data from each sand wave morphology is extracted and fitted to establish a discriminant standard, which is then successfully validated using experimental and quantifiable data. Also, based on the foregoing methodoly, it is further discovered that the short-term significant fluctuation of sand wave morphology is closely correlated with significant channel changes in rivers with a high width-depth ratio, using Yellow River Estuary as an example.