Mapping Soil Heavy Metal Levels in England and Wales for Application in Public Health

摘要

Background/Aim: Heavy metals, especially Lead (Pb), Cadmium (Cd) and metalloid Arsenic (As), have known health impacts. Using extensive soil sampling data from the British Geological Survey (BGS) we aimed to generate reliable soil metal concentration surfaces, with accompanying measures of uncertainty, to identify areas exceeding health-based guideline values. Methods: The BGS soil metals data for Pb, Cd and As comprise an amalgamation of six different surveys collected over a 34-year period and have up to four samples per km2. Ordinary Kriging (OK) and Empirical Bayesian Kriging (EBK) were applied to an 80% stratified sample (i.e. 80% random sample of each survey, comprising 36,802 'training samples'). The remaining 20% 'test sample' (9,202 samples) was used for cross-validation. The predicted standard error generated via the EBK approach was mapped to quantify the uncertainty/error in the interpolated surface. Results: The OK and EBK interpolation methods generated similar soil concentration surfaces, and cross validation with the held-back test data indicated moderate to good model performance: Spearman's rank correlation coefficients 0.83 (Pb), 0.72 (As) and 0.57 (Cd). Overall, the models tended to underestimate metal concentrations, although at lower levels (where the majority of the samples clustered), the models both over and under-estimated metal concentrations. For each metal, the areas of high uncertainty (i.e. high standard errors) were located in the areas with high soil metal concentrations. In terms of small areas exceeding health based guideline values, almost 4% of Lower Layer Super Output Areas (LSOAs) exceeded the UK soil guideline value for As, and 3.6% exceeded the USA EPA hazard standard for Pb for residential children's play areas. No LSOAs exceeded the UK soil guideline value for Cd. Conclusions: These soil metal concentration surfaces can be used to identify areas exceeding health based guidelines.