Long-term water quality simulation and driving factors identification within the watershed scale using machine...

Area of Science:

• Environmental Science
• Water Quality Management
• Data Science

Background:

• Surface water quality improvement in China is ongoing, but regional challenges persist.
• The Liao River Basin, a vital industrial and agricultural region, experiences unstable water quality despite management efforts.

Purpose of the Study:

• To compare data-driven models for filling historical water quality data gaps (1980-2022) in the Liao River Basin.
• To identify and quantify the driving factors influencing water quality trends.

Main Methods:

• Evaluated Random Forest (RF), Support Vector Machine Regression (SVR), K-Nearest Neighbors (KNN), stacking, Long Short-Term Memory (LSTM), and Convolutional-LSTM models for data imputation.
• Utilized the SHapley Additive exPlanations (SHAP) model to assess the impact of various factors on water quality parameters.
• Analyzed long-term water quality data for total nitrogen (TN), ammonia nitrogen (NH3-N), total phosphorus (TP), chemical oxygen demand (COD_Cr, COD_Mn), and electroconductibility (E).

Main Results:

• The Random Forest model demonstrated superior predictive performance for water quality data.
• Total Nitrogen (TN) increased by approximately 20% from 1980 to 2022; other parameters were controlled.
• Anthropogenic activities (agriculture, urban areas), climatic factors (extreme rainfall, precipitation, temperature), and geographical factors (soil, slope) significantly impacted water quality.

Conclusions:

• Data-driven models, particularly RF, are effective for reconstructing historical water quality data.
• Targeted management strategies are needed to address the rising TN levels driven by human activities.
• Integrated approaches considering anthropogenic, climatic, and geographic factors are crucial for effective watershed water quality management.