Author(s): Ziyi Wang; Chao Qin; Haifei Liu; Yuan Xue; Ximeng Xu And Fenli Zheng

Linked Author(s): Chao Qin

Keywords: Gully erosion; Photogrammetry; Deep learning; Delayed response model

Abstract: Channel sidewall expansion dominants channel evolution process after bed incises to a less- or non-erodible layer, significantly contributing to sediment yield in river systems. Current knowledge on channel evolution lacks the quantification of full-cycle process of channel widening. This study investigates the full-cycle dynamics of channel sidewall expansion under varying hydrological and topographic conditions using artificial runoff simulation experiments and a delayed response model. Photogrammetry and deep learning-based image processing were employed to capture the temporal evolution of channel width. An empirical model was developed to capture the active expansion phase within the experimental duration and a delayed response model to simulate the full-cycle of channel sidewall expansion. Results indicate that channel sidewall expansion exhibits a two-phase pattern: an initial rapid expansion phase followed by an exponential decay phase as the system approaches equilibrium. The delayed response model effectively predicts the temporal evolution of channel width, showing high agreement with observed data (R2>0.8, ENS > 0.7). Equilibrium channel width and the time required to reach equilibrium were found to be related to flow rate, slope gradient, and soil properties. Notably, the increase in channel sidewall expansion rate due to higher inflow rates diminished as slope steepness increased. These findings enhance our understanding of gully evolution and provide a theoretical framework for predicting and managing soil erosion in dynamic landscapes.

DOI: https://doi.org/10.64697/978-90-835589-7-4_41WC-P2042-cd

Year: 2025