A risk-based approach for accurately delineating the extent of soil contamination: The role of additional sampling in transition zones

Abstract

Accurate soil contamination delineation is crucial for deciding where remediation efforts are required. However, misjudgments, either in underestimating or overestimating contamination extents could incur different risks: underestimation may result in environmental risks, while overestimation may lead to financial risks. This study proposed an approach based on environmental and financial risks (loss risk) to improve the performance of contamination delineation. Additionally, the impact of additional sampling in the transition zones on the contamination delineation was evaluated. This approach was demonstrated in Hechi, southwest China, where the soil was polluted by arsenic and cadmium. Initially, geostatistical simulation and 512 initial soil sampling were utilized to generate two maps: the conditional coefficient of variation (CCV) and the conditional probability of exceeding a critical threshold (CPT). These two maps were integrated to quantify the uncertainty in identifying the transition zones, guiding additional sampling. Out of 189 candidate sampling sites, we selected 100 additional sites to address high uncertainty. Subsequently, the minimization risk principle was employed to delineate contamination boundaries. The results showed that contaminated areas in the initial phase were significantly underestimated. Additional sampling in the transition zones improved the performance of soil contamination delineation. The performance metrics of Recall and F1 score for arsenic exhibited a notable enhancement of 6 % and 7 %, respectively. As for cadmium, there was an enhancement with Recall and F1 scores increasing by 4 % and 7 %, respectively. Adding 100 extra samples reduced the financial risks of arsenic and cadmium by 13 % and 11 %, respectively. In comparison, the 100 additional samples reduced the environmental risks of arsenic and cadmium by 55 % and 72 %, respectively. The study demonstrates that combining CCV and CPT for additional sampling efficiently mitigates the risks of delineating contaminated areas, which could help better understand the boundaries and gradient of contamination.