Sustainable utilization of agricultural residues is essential for reducing environmental pollution while enhancing long-term soil carbon sequestration. Converting crop straw into biochar has emerged as a promising environmental technology for agricultural waste utilization, however, whether this conversion fundamentally alters soil organic carbon (SOC) stabilization pathways remains poorly understood. A three-year field experiment in an intensive rice–wheat cropping system was conducted to compare the effects of direct straw return and straw-derived biochar application on SOC fractions, microbial responses, and microbial necromass formation. Results showed that straw-derived biochar combined with N fertilization (NBC treatment) significantly increased total SOC (TSOC) content and stock by 17–22% and 12–15% compared to the CK, CS, and NPK treatments (P < 0.05), respectively, while it exhibited a 48% increase (P < 0.05) in mineral-associated organic C (MAOC) relative to the CK treatment. In contrast, direct straw return (NS treatment) promoted the accumulation of plant-derived lignin phenols. Biochar amendment also increased fungal α-diversity, strengthened bacterial–fungal network connectivity, and enhanced microbial necromass C accumulation. Microbial necromass C, especially bacterial necromass, contributed up to 16.7% of TSOC and was identified as the strongest predictor of SOC accumulation in partial least squares-path modeling. Our results suggest that converting crop residues into biochar prior to soil return, combined with N fertilization, promotes a transition in SOC formation from plant residue accumulation toward microbial-mediated stabilization pathways. Our results highlight that integrating straw pyrolysis with nutrient management represents an effective strategy to optimize residue management and support climate change mitigation efforts in intensive rice-wheat system, though evaluating its long-term persistence requires continued multi-site field validation.
Nitrate pollution poses a significant threat to aquatic ecosystems and human health globally. The nitrate-nitrogen isotopic composition (δ15N-NO3−) serves as a powerful tracer for identifying nitrate sources and transformations. However, direct measurement of δ15N-NO3− is cost-prohibitive and technically demanding, limiting its widespread application. This study utilized a comprehensive five-year (2020–2024) dataset from the Wen-Rui Tang River watershed (Southeast China) to develop three machine learning models including Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) for predicting δ15N-NO3−. The results indicated that the XGBoost model yielded the best performance, with r = 0.80, NS = 0.63, RMSE = 1.98‰, and MAE = 1.43‰ during the testing stage. The scatter plots and Taylor diagram further visually confirmed this superiority, with tighter clustering along the 1:1 line and the closest position to the reference point. Furthermore, SHapley Additive exPlanations (SHAP) analysis was employed to capture the complex, non-linear relationships between water quality parameters and δ15N-NO3−. The SHAP results identified electrical conductivity (EC), ammonium-nitrogen, nitrate-nitrogen, total nitrogen, and nitrite-nitrogen as the most influential predictors. The SHAP-identified associations were consistent with the nitrification process. High EC conditions correlated with lower predicted δ15N-NO3− values, suggesting dilution by low-δ15N sources such as synthetic fertilizers or industrial effluents. Conversely, higher predicted values corresponded to manure and sewage inputs. This study confirmed that ensemble models, particularly XGBoost, when coupled with SHAP analysis, offer a robust and interpretable supplementary approach for estimating environmental isotope signatures.
High-strength acetonitrile (ACN) wastewater poses a dual challenge: rapid nitrile hydrolysis must be sustained under substrate stress, while the released ammonium imposes a downstream nitrogen-oxidation burden. Here, an immobilized aerobic biofilm was challenged with ACN loads equivalent to 500–10,000 mg COD/L to resolve how carbon conversion and nitrogen oxidation diverge under increasing load, and its engineering applicability was further evaluated using real wastewater. ACN was nearly completely removed within 16 h across all loads, despite a 41.1% lower apparent pseudo-first-order rate constant at H5 (10,000 mg COD/L) than at H3 (2500 mg COD/L). High loading depressed ATP and NADH and increased oxidative and membrane stress, yet electron transport capacity remained high and nitrilase, nitrile hydratase, and amidase were enriched. Soluble EPS became polysaccharide-enriched, and the microbial community showed load-dependent selection and restructuring. Metagenomics indicated enhanced potential for ACN hydrolysis, acetate activation, and central carbon metabolism, whereas nitrogen-oxidation genes decreased to 55–63% of low-load levels, consistent with reduced nitrification-enzyme activities. In a 30-d continuous-flow test with real ACN-dominated wastewater (influent COD 102,429 ± 2663 mg COD/L), COD removal remained 93.21 ± 2.25%. These results demonstrate a load-dependent, system-level functional differentiation characterized by preferentially maintained ACN/carbon conversion and increasingly constrained nitrogen oxidation, supporting immobilized biofilms as high-load biological pretreatment before targeted nitrogen polishing.
