[Prediction Model of Groundwater Sulphate Based on Combined Multi-source Spatio-temporal Data]

Area of Science:

• Hydrogeology and Environmental Science: Focuses on groundwater chemistry and spatial variability analysis.

Context:

• Groundwater sulfate (SO42-) concentration prediction is vital for water resource management.
• The Yarkant River Basin's plain area exhibits complex spatial variations in groundwater SO42-.

Purpose:

• To develop an accurate model for predicting spatial variation trends in groundwater SO42- concentration.
• To identify key environmental factors influencing groundwater SO42- levels using multi-source spatio-temporal data.

Summary:

• A Bayesian optimization algorithm-optimized random forest regression (BOA-RFR) model was employed, integrating land cover, soil, elevation, and pH data.
• Groundwater pH, ground elevation (GE), and percentage of bare land (BAR) were identified as significant negative predictors of SO42- concentration.
• The BOA-RFR model, enhanced with geostatistical interpolation, achieved R² > 0.96, improving prediction accuracy and reducing errors.

Impact:

• Provides a robust method for predicting spatial SO42- distribution in groundwater.
• Identifies critical factors influencing groundwater hydrochemistry, aiding targeted management strategies.
• Highlights expanding areas of high SO42- groundwater in the Yarkand River Basin, informing regional water quality planning.