Machine-Learning Surrogate Modeling of SOC and N2O Responses to Diversified Crop Rotations in the Texas High Plains

Diversified crop rotations and cover crops are increasingly promoted as cli-mate-smart management strategies for improving soil organic carbon (SOC) while minimizing nitrous oxide (N₂O) emissions in semiarid agroecosystems. However, regional-scale assessment of SOC–N₂O trade-offs remains computationally chal-lenging because process-based simulations across large spatial and climatic do-mains are highly demanding. This study combined long-term DSSAT simulations with machine-learning (ML) surrogate models to quantify and spatially predict SOC and N₂O responses to diversified rotations and cover-crop systems across the Texas High Plains (THP). Random Forest models were trained using DSSAT outputs representing multiple crop rotations, climate scenarios, and soil condi-tions. Two complementary modeling frameworks were developed: (i) BAU-relative responses, which quantified SOC and N₂O changes relative to base-line management, and (ii) added cover-crop effects, which isolated the additional benefits of diversified cover-crop systems relative to a simplified no-cover crop improved rotation system. Model interpretation was conducted using permuta-tion importance and SHAP analysis to identify the dominant environmental and management controls. The surrogate models accurately reproduced DSSAT-derived responses, particularly for BAU-relative SOC and N₂O changes reaching R² ≈ 0.90, while added cover crop SOC was less predictable than added N2O. Spatial predictions revealed strong geographic variability in climate-smart response zones, with larger SOC benefits generally observed under future climate conditions, particularly during the 2070s. However, some regions also exhibited stronger N₂O trade-offs, highlighting the need for spatially targeted management strategies. The results demonstrate the potential of combining process-based simulations with ML surrogates to generate computationally efficient deci-sion-support tools for climate-smart agriculture in semiarid cropping systems.