Deciphering the influence of hydrological variables on water quality via data-driven predictive models

Area of Science:

• Environmental Science
• Water Resource Management
• Machine Learning Applications

Background:

• Miyun Reservoir's water quality is linked to its inflowing rivers, Chao and Bai Rivers.
• Quantitative data on hydrological factors influencing Chao and Bai River water quality is limited.

Purpose of the Study:

• To quantitatively assess the impact of rainfall, runoff, sediment, and dissolved oxygen (DO) on Chao and Bai River water quality.
• To determine optimal ranges for these hydrological parameters to ensure good water quality.
• To predict water quality responses to hydrological changes using machine learning.

Main Methods:

• Utilized machine learning, specifically the XGBoost model, for predicting water quality impacts.
• Employed multi-objective optimization to identify optimal hydrological parameter ranges.
• Conducted scenario predictions to evaluate water quality under extreme hydrological conditions.

Main Results:

• The XGBoost model demonstrated a high fitting effect (R² > 0.9).
• Optimal ranges for hydrological parameters were determined for both Chao and Bai Rivers.
• Scenario analysis indicated that the Chao River is more susceptible to water quality deterioration than the Bai River under extreme hydrological events.

Conclusions:

• The Chao River faces a significantly higher risk of water quality degradation compared to the Bai River.
• Hydrological fluctuations pose a greater threat to water quality in the Chao River.
• Implementing soil and water conservation in the upstream basins of both rivers is recommended to enhance water quality.