Remote Sensing Inversion of Water Quality Grades Using a Stacked Generalization Approach

Area of Science:

• Environmental science
• Remote sensing
• Water quality assessment

Background:

• Effective water quality monitoring is essential for environmental management and policy.
• Current methods struggle to integrate multi-source remote sensing data.
• The Songhua River Basin (SHRB) requires advanced water quality assessment tools.

Purpose of the Study:

• To develop and validate a stacking algorithm for water quality grade classification in the SHRB.
• To integrate multi-source remote sensing data for improved water quality assessment.
• To explore correlations between environmental factors and water quality.

Main Methods:

• Employed a Stacked Generalization (SG) model within Google Earth Engine (GEE).
• Utilized a combination of multiple machine learning models for enhanced classification.
• Analyzed correlations between the normalized difference vegetation index (NDVI), precipitation, and water quality grades.

Main Results:

• The SG model achieved a high accuracy of 91.67% in classifying water quality grades.
• Demonstrated significantly improved classification performance over traditional methods.
• Identified substantial correlations between NDVI, precipitation, and water quality.

Conclusions:

• The developed stacking algorithm is effective for water quality monitoring in the Songhua River Basin.
• The method successfully integrates remote sensing data for comprehensive assessment.
• Findings provide insights into the influence of natural factors on water quality and pollution.