Machine learning for monitoring lobe dynamics of the Yellow River Delta since the implementation of water-sediment regulation scheme

Study region: The active Yellow River Delta (YRD) lobe. Study focus: Over recent decades, the morphology of the active delta lobe has changed frequently and unpredictably under a changing environment. To address this challenge, this study developed a lobe area-tidal level model leveraging satellite images combined with a machine learning (ML)-based approach to monitor the evolution of the lobe area and morphological changes since the implementation of the Water-Sediment Regulation Scheme. This framework enables consistent, large-scale extraction of lobes from long-term satellite images, resolving limitations of subjectivity and low efficiency in conventional methods. New hydrological insights for the region: The results from ML-based monitoring show that the overall area of the lobe has been expanding seaward at a rate of approximately 4.0 km²/yr, and its morphology has exhibited three stages: eastward development (2002–2009), northward development (2009–2017), and northward stabilization (2017–2022). The dynamic spatiotemporal patterns of the lobe reflect the complex interactions between channel dynamics, vegetation feedback, and sediment supply. The migration/bifurcation of the mouth channel have altered the redistribution of sediment, increasing the lobe land-building efficiency by 142–230 %. A critical sediment threshold ranging from 0.48 to 1.5 × 10⁸t is found to sustain the development of the lobe. This study clarifies the importance of multi-factorial interactions in the evolution of the lobe, and emphasizes the need for balanced intervention measures to maintain delta resilience.