Advancing Sediment Incipient Motion Modeling: Predictive Neural Network Models for Vegetated Beds

Numerous studies have explored how aquatic vegetation influences sediment transport and the incipient motion velocity (IMV) of particles. Existing IMV prediction criteria primarily focus on either rigid or flexible vegetation independently, yet their accuracy remains limited, and no generalized criterion exists for environments where both vegetation types coexist. Understanding IMV in vegetated channels is not only essential for sediment transport modeling but also plays a crucial role in preserving aquatic ecosystems, mitigating coral reefs stress, and improving water quality. This study aims to bridge this gap by developing a novel predictive framework that integrates experimental data with artificial neural networks (ANN). This research was conducted in three key phases: First, an ANN model was trained on laboratory datasets, demonstrating high predictive accuracy (R² = 0.9974, RMSE = 0.1652) for IMV in mixed vegetation environments. Second, existing IMV criteria were modified to accommodate both rigid and flexible vegetation, with the newly developed relationship achieving a strong predictive performance (R² = 0.89, RMSE = 1.02). Third, a large dataset of 1,000 data points was generated using the calibrated ANN, further validating the new criterion's robustness (R² = 0.81, RMSE = 3.5). The findings highlight the advantages of an integrated approach to IMV prediction, offering a more accurate and adaptable tool for sediment transport modeling while contributing to the sustainable management of aquatic environments.