Spatial distribution prediction of soil heavy metals based on sparse sampling and multi-source environmental data.

Area of Science:

• Environmental Science
• Geospatial Analysis
• Soil Science

Background:

• Accurate spatial prediction of soil heavy metals is vital for environmental management and remediation.
• Limited soil sampling points pose a significant challenge to achieving precise heavy metal distribution predictions.

Purpose of the Study:

• To introduce and evaluate a novel hybrid approach, Light Gradient Boosting Machine plus Ordinary Kriging (LGBK), for predicting soil heavy metal spatial distribution.
• To assess the predictive performance of the LGBK model compared to traditional methods using limited soil sampling data.

Main Methods:

• Collected 137 soil samples from the Shengli Coal-mine Base, Inner Mongolia, China.
• Measured heavy metal concentrations (Cu, Zn, Cr, As) and utilized environmental covariates.
• Developed and validated the LGBK model for spatial prediction of heavy metals.

Main Results:

• The LGBK model demonstrated superior predictive performance over traditional models.
• Cross-validation yielded high coefficients of determination (R²): 0.65 for Cu, 0.52 for Zn, 0.57 for Cr, and 0.63 for As.
• The LGBK model effectively captured intricate spatial features and accurately forecasted heavy metal distribution trends.

Conclusions:

• The LGBK approach provides higher-precision soil heavy metal predictions, particularly valuable when sampling points are limited.
• The study analyzed the spatial distribution of Cu, Zn, Cr, and As and identified key influencing environmental factors.
• This research offers significant implications for the environmental regulation and remediation of heavy metal pollution.