Tracking the Rise and Fall of Riparian Vegetation: A Temporal Network Approach

Abstract

Fluvial landscapes are the product of complex interactions between physical and biological processes, including hydrological fluctuations, sediments transport dynamics, and vegetation. In recent years, vegetation dynamics have been increasingly recognized as a key factor shaping riparian landscapes. Traditionally, vegetation dynamics have been assessed by comparing land cover classes between timesteps, which captures broad trends but fails to track individual vegetation patches over time. This study addresses that gap by introducing a novel object-based temporal network framework that links classified land cover objects from 15 aerial image years (1946–2019) to reconstruct the lifecycles of large, high riparian vegetation (HV) patches on a stretch of the Allier River. Here we demonstrate the value of the novel approach, revealing spatial and temporal dynamics that are obscured in traditional class-based change analyses. The results show that most large HV patches follow a rise–and–fall trajectory, typically lasting around 10 years. Growth most often begins through emergence from bare soil or low vegetation, while further development occurs through merging and eating of nearby patches. Degradation usually occurs through direct disappearance to non-high-vegetation patches, like bare soil, low vegetation or water, while an intermediate step of patches falling apart to smaller HV patches is also observed. These patterns reveal key insights into patch-scale dynamics that are not accessible through traditional class-based methods. By shifting focus from aggregated land cover change to individual patch trajectories, this method reveals the temporal and spatial processes underlying vegetation development. The approach offers a tool for ecological monitoring and provides a basis for future predictive models that integrate vegetation dynamics with hydrological and geomorphological change.