A Review of Remote Sensing-Based Approaches for River Morphological Change Analysis using DSAS and Machine Learning Techniques

River morphology is very dynamic, and it continues to adapt to natural processes and human activities, with channel migration, bank erosion and deposition occurring at varying rates; these processes can create major difficulties in sustainable river management. The efficiency in monitoring such morphological changes over large spatial and temporal scales has been made possible due to advances in satellite remote sensing. The Digital Shoreline Analysis System (DSAS) is one of several approaches that have been widely used to measure rates of riverbank changes through statistical calculations, including Net Shoreline Movement (NSM), End Point Rate (EPR), and Linear Regression Rate (LRR). The reliability of DSAS-based analysis, however, is very dependent on the accuracy of the bankline that is extracted from the satellite image. The recent advancement of machine learning techniques, especially those based on random forest (RF) and support vector machine (SVM), has shown better accuracy in land–water classification and feature extraction in challenging riverine environments. The combination of machine learning and DSAS provides increased robustness in the delineation of riverbanks and the evaluation of morphological changes that have occurred. The present paper provides a detailed review of remote sensing–based approaches for morphological assessment of natural streams with a particular emphasis on the simultaneous use of DSAS and machine learning techniques. The study reviews pertinent published sources, discusses their strengths and weaknesses and outlines current trends and gaps in the literature. A conceptual framework is suggested to combine the integration of DSAS, RF and SVM to support future studies in order to enhance the accuracy and applicability of river morphology monitoring in the context of sustainable management of river basins.