Regional disparities and social-ecological constraints of soil antibiotic pollution in urban agglomerations of China

• Published in: Progress in physical geography Vol. 48; no. 4; pp. 536 - 550
• Main Authors: Zhao, Fangkai, Yang, Lei, Feng, Qingyu, Ji, Chenxu, Li, Min, Fang, Li, Yu, Xinwei, Chen, Liding
• Format: Journal Article
• Language: English
• Published: London, England SAGE Publications 01.08.2024; Sage Publications Ltd
• Subjects: Agglomeration; Algorithms; Anthropogenic factors; Antibiotics; Carrying capacity; Climate change; Climate models; Climate prediction; Clustering; Economic development; Economic growth; Geographical isolation; Learning algorithms; Livestock; Livestock production; Machine learning; Performance prediction; Pollution; Pollution control; Regional development; Regional planning; Rivers; Soil pollution; Soils; Urban areas; Yangtze River Delta; soil carrying capacity; spatial hotspots; human activities; random forest model; Soil health
• ISSN: 0309-1333; 1477-0296
• DOI: 10.1177/03091333241263796

With intensive socio-economic growth, antibiotics have been heavily discharged into soils, posing a threat to ecosystems and human health. However, limited data are available in broad-scale urban agglomerations which limit the understanding of the driving mechanisms for antibiotic pollution and hinder action plans for pollution control. Here, we examined the underlying mechanisms driving antibiotic pollution and their regional disparities (Central Yunnan vs. Yangtze River Delta), and we predicted antibiotic concentrations in the soils of these two urban agglomerations using machine learning algorithms. Specifically, anthropogenic pressures such as population aggregation and livestock production accounted for the highest contribution to model accuracy in Central Yunnan, suggesting human interference mediated by geographical isolation likely plays a pivotal role in the clustering of antibiotic pollution in soils. However, soil and climate variables were the most important predictors in the more developed Yangtze River Delta, indicating that human-mediated soil carrying capacity was likely the main mechanism controlling antibiotic pollution in this region. Our results showed that machine learning models performed better (area under the receiver operating characteristics curve: 0.91–0.98) in predicting antibiotic hotspots than classic linear models. Our findings highlight the regional disparities in underlying mechanisms for antibiotic pollution in the soils of urban agglomerations, and demonstrate that machine learning algorithms based on spatially available predictors can be extended to other regions.