Find in Library

Study region: The Yellow River located in the north of China, the most hyper-concentrated sediment-laden rivers in the world, is featured by complicated spatiotemporal variation of river patterns. Study focus: River patterns contribute to comprehending the spatial morphological evolution of rivers associated with hydrological processes. However, it is still a challenge to effectively discriminate different river patterns due to the dynamics and variability of rivers. A solution is proposed to discriminate river patterns based on Rough Set theory via a holistic synthesis of multi-dimensional morphological information. First, a hierarchical structure integrating the boundary and the interior was suggested to maximally describe the morphological feature of river patterns. Rough Set theory was applied to select the main feature factors from multiple geometric indexes and landscape pattern indexes. Finally, river pattern discriminant rules were generated based on the reduced feature subsets. New hydrological insights for the region: Results demonstrate that 81 discriminant rules efficiently express the morphological feature of four river patterns. Compared with several competitive machine learning methods, river pattern discriminant rules display good performance, with the advantages of interpretability, simple modelling, and fewer training samples. Furthermore, the established discriminant rules of river patterns have achieved an accuracy of 98.63% in the application of the Lower Yellow River, which is expected to analyse the spatiotemporal evolution and geomorphological processes of different river patterns.