National-scale mapping of time-dependent apparent degradation rates of tetracycline and sulfonamide antibiotics in Chinese cropland soils under a spring fertilization scenario.

Quantifying antibiotic degradation at large spatial scales is essential for assessing environmental persistence and exposure risks in agricultural soils. Here, we developed a machine learning-based framework to map time-dependent apparent degradation rates (μapp) of tetracycline- and sulfonamide-class antibiotics in Chinese croplands under a spring fertilization scenario. A literature-derived database comprising 952 degradation datasets from 66 studies and 120 soils was compiled for tetracyclines and sulfonamides. Models were trained using soil-grouped cross-validation and linked μapp to soil physicochemical properties, environmental conditions, and experimental factors. The tetracycline model showed higher predictive accuracy (R2 = 0.724) than the sulfonamide model (R2 = 0.576), reflecting stronger soil-driven controls on tetracycline degradation. Spatial predictions reveal pronounced east-west gradients for tetracyclines, with slower degradation in western and northwestern China, whereas sulfonamides exhibit generally faster and more spatially uniform degradation with a reversed northwest-southeast contrast. Uncertainty analysis indicates more robust predictions for tetracyclines and higher uncertainty for sulfonamides. These results provide the first parallel national maps of time-dependent apparent degradation rates for major veterinary antibiotics in Chinese croplands, offering a quantitative basis for large-scale fate modeling and region-specific risk management.