Unveiling the Role of Wetland Strategies in Antibiotic Risk Reduction across China by Machine Learning

Pervasive antibiotic pollution in water environments
has emerged
as a serious threat to global ecosystem functions and public health.
While wetland expansionincluding protection, restoration,
and construction, is widely promoted for sustainable water quality
improvement, its effectiveness in mitigating antibiotic pollution
remains poorly understood. Here, we develop a machine learning model
based on a compiled data set of 337 experimental observations to quantify
antibiotic removal and map risk distribution in wetlands across 2,833
counties/districts in mainland China. Between 2010 and 2020, the wetland
area across China expanded by 34.7%, yet antibiotic removal improved
by only 0.1%, failing to meaningfully reduce the risk. We find that
antibiotic removal in wetlands is primarily constrained by input magnitudes
rather than the wetland area. To address this, we proposed a multistage
wetland management strategy to enhance antibiotic removal by 27.6%
in 2020 and high-risk area reduction by 90.6% under optimal policies
by 2035. Furthermore, we further identified the importance of wetland
management strategies through an interpretable model. Our findings
provide novel wetland strategy insights for policymakers and highlight
the fact that wetland expansion without targeted management is insufficient
for controlling antibiotic pollution, although it is an important
cornerstone characteristic for water quality improvement.