AI-driven predictions of geophysical river flows with vegetation

Area of Science:

• Hydrology and Water Resources Engineering
• Environmental Fluid Mechanics
• Computational Science

Background:

• Accurate forecasting of river flow velocity in vegetated channels presents a significant challenge in river research.
• Existing empirical equations often struggle to capture the complex dynamics influenced by vegetation.
• Machine learning (ML) offers a promising alternative for improving flow velocity predictions.

Purpose of the Study:

• To quantify the forecasting performance of various independent and hybrid machine learning (ML) models for flow velocity in vegetated channels.
• To compare the efficacy of ML models against traditional empirical equations.
• To identify the most influential parameters affecting flow velocity prediction.

Main Methods:

• Utilized flow velocity measurements from natural and laboratory flume experiments.
• Assessed four standalone ML models: Kstar, M5P, Reduced Error Pruning Tree (REPT), and Random Forest (RF).
• Evaluated eight hybrid ML algorithms: AR-Kstar, AR-M5P, AR-REPT, AR-RF, BA-Kstar, BA-M5P, BA-REPT, and BA-RF, combined with Additive Regression (AR) and Bagging (BA).

Main Results:

• Vegetation height was identified as the most sensitive parameter influencing flow velocity.
• All evaluated ML models outperformed traditional empirical equations.
• Optimal performance for most ML algorithms was achieved when all input parameters were utilized.
• The hybrid AR-M5P model achieved the highest accuracy (R²=0.954, R=0.977, NSE=0.954).

Conclusions:

• Hybrid ML algorithms provide superior flow velocity forecasting in vegetated rivers compared to standalone ML models and empirical equations.
• The AR-M5P model is recommended as the optimal choice for accurate flow velocity prediction in vegetated riverine environments.
• Incorporating all relevant input parameters enhances the predictive power of ML models for river flow dynamics.