Research on the Development of an Inland Lake Bathymetry Estimation Model Based on Multispectral Data

Area of Science:

• Earth and Environmental Sciences
• Remote Sensing
• Hydrology

Background:

• Lakes are vital for regional economies and ecosystems.
• Accurate water depth data is crucial for managing lake resources and ecological stability.
• Existing remote sensing models for inland lake depth estimation lack unified evaluation and consistent performance.

Purpose of the Study:

• To systematically compare numerical and machine learning models for inland lake water-depth estimation.
• To propose a novel machine learning-based methodology for constructing accurate water-depth estimation models for inland lakes.
• To integrate machine learning with multispectral remote sensing data for enhanced depth estimation.

Main Methods:

• Systematic comparison of numerical models (MLR) and machine learning models (Random Forest, BP neural networks, AdaBoost).
• Development of a new machine learning-based methodology for inland lake water-depth estimation.
• Integration of machine learning algorithms with multispectral remote sensing data.

Main Results:

• Machine learning models (RF, BP, AdaBoost) outperformed the MLR model in accuracy (R² up to 0.88 vs. 0.59).
• The proposed methodology yielded improved precision in machine learning models (R² up to 0.92).
• The developed model is particularly suitable for medium- and small-sized lakes, providing high-precision underwater topographic maps.

Conclusions:

• Machine learning models significantly enhance the accuracy of remote sensing-based water-depth estimation for inland lakes.
• The proposed methodology offers a more precise and applicable approach for water depth estimation in medium- and small-sized lakes.
• The study provides a valuable tool for rapid, high-precision hydrological data acquisition, supporting effective lake water resource management.