Channel Activity Remote Sensing Retrieval Model: A Case Study of the Lower Yellow River

Abstract

Meandering channel migration is a widespread phenomenon in rivers all around the world. Channel activity, which reflects the rate of change of a meandering channel, is calculated by averaging lateral channel migration. Channel migration can create new channels and abandon old ones, with effects on the natural environment. Floods can even lead to excessive rates of channel migration, which can threaten cities or farmland. Remote sensing can detect the spatial and temporal dynamic characteristics of the river channel, taking into account both spatial and temporal resolution, and can help in planning for the safety of the river channel in advance. Previous studies on river channels have suffered from a low accuracy of data, low level of automation, and subjectivity. To overcome these limitations, we propose a channel activity remote sensing retrieval model (CARSM) in this paper. CARSM extracts water using the modified normalized difference water index (MNDWI) combined with Otsu’s method on the Google Earth Engine (GEE) platform, then extracts the channel centerlines via water mask maps using RivWidthCloud, and finally calculates channel activity based on the geometric relationship of the channel centerlines. With more objective extraction results, CARSM can guarantee more than 95% accuracy of channel activity and its high degree of automation can save a lot of labor costs. We use Landsat images to monitor the channel of the Lower Yellow River and calculate the overall and segmental channel activity separately. Our results show that the overall channel activity of the Lower Yellow River has gradually decreased between 1990 and 2020, with decreases of 33.04% and 41.06%, respectively. Analysis of channel activity reveals that the water sediment pattern of the Lower Yellow River changed from siltation to scouring after the completion of Xiaolangdi Reservoir, and the Lower Yellow River is gradually becoming stable.