Integrating image processing and machine learning for phase specific estimation of Manning roughness coefficient in...

Area of Science:

• Agricultural Engineering
• Hydrology
• Computer Science

Background:

• Accurate Manning's roughness coefficient (n) estimation is crucial for hydrological modeling and furrow irrigation.
• Traditional methods are limited by spatial-temporal variability and labor-intensive measurements.

Purpose of the Study:

• To develop a novel algorithm for dynamic prediction of Manning's n using image processing and machine learning.
• To evaluate the algorithm's performance across different data input scenarios.

Main Methods:

• Integration of high-resolution image processing with machine learning (Random Forest).
• Evaluation of three scenarios: full-field data, images only, and images plus selected field data.
• Manning's n computation using SIPAR_ID model and Manning's equation.

Main Results:

• Random Forest achieved 99% accuracy with full-field data.
• Scenario with images plus selected data maintained 95-96% accuracy with reduced inputs.
• Excluding hydraulic variables reduced performance by approximately 50%.

Conclusions:

• The developed algorithm offers a robust, cost-effective, and practical solution for real-time Manning's n estimation.
• This approach supports sustainable water management and precision agriculture.
• Image processing and machine learning significantly enhance n estimation accuracy and efficiency.