Screening and optimization of interpolation methods for mapping soil-borne polychlorinated biphenyls

Area of Science:

• Environmental Science
• Geospatial Analysis
• Soil Science

Background:

• Optimal interpolation methods for mapping soil-borne organic pollutants, like polychlorinated biphenyls (PCBs), lack scientific consensus.
• Accurate spatial prediction of soil contaminants is crucial for environmental risk assessment and management.

Purpose of the Study:

• To compare classic interpolation methods for mapping soil-borne PCBs using a high-resolution soil monitoring database.
• To identify the most accurate interpolation technique and assess the impact of data transformations and parameters on prediction accuracy.
• To propose a standardized workflow for improved interpolation accuracy in soil pollutant mapping.

Main Methods:

• Comparison of interpolation methods including empirical Bayesian kriging (EBK), ordinary kriging (OK), and inverse distance weighting (IDW).
• Utilized a high-resolution soil monitoring database for polychlorinated biphenyls (PCBs).
• Investigated the effects of logarithmic data transformation, search neighborhood size, and outlier handling on prediction accuracy.

Main Results:

• Empirical Bayesian kriging (EBK) demonstrated the highest prediction accuracy for total PCB concentration.
• Inverse distance weighting (IDW) exhibited among the highest root mean squared errors (RMSE).
• Logarithmic transformation improved semivariogram modeling in kriging; outliers and increased search neighborhood impacted accuracy differently across methods, often leading to map oversmoothing.

Conclusions:

• EBK is a superior method for accurate spatial prediction of soil-borne PCBs compared to IDW and OK.
• Data preprocessing, including logarithmic transformation and careful outlier management, is essential for accurate geostatistical modeling.
• A standardized interpolation workflow can significantly enhance accuracy and reduce errors in soil pollutant mapping.