Spatially Explicit Crop Planning for Water–GHG–Profit Trade-Offs in Northeast China’s Black Soil Region: An End-to-End Land Use Optimization Framework

Land use planning in the Black Soil Region of Northeast China must be sustainable, taking into account food security, water use, GHG emissions, and economic returns. Current crop suitability mapping and single-objective optimization studies tend to analyze crop occurrence, crop structure, and spatial allocation independently, which is of little value in spatial planning. In this study, a three-stage integrated approach is proposed, involving deep learning crop occurrence mapping, multi-objective crop structure optimization, and suitability-guided spatial allocation. During Stage I, a lightweight U-Net semantic segmentation model, BlackSoilCropNet, is developed to provide per-pixel occurrence probabilities of rice, maize, soybean, and other types of crops based on Sentinel-2 time series and auxiliary environmental predictors. In stage II, NSGA II will optimize the area structure of the crops and reduce water consumption and GHG emissions with the maximum profit under the constraints of the cropland, water, and production. Selected Pareto optimal solutions are transformed to crop allocation maps and transition hotspot outputs in Stage III. The framework resulted in three viable planning options. The economic priority scenario resulted in the highest profit (USD 27.9 billion), with higher water consumption and emissions. The environmental-priority scenario resulted in a reduction in water use to 118.2 × 109 m3 and emissions to 50.9 MtCO2e, but at the cost of lower production and profits. There was a balance between economic stability and an improved environment in the balanced scenario. The framework provides a reproducible, geospatial decision support approach for sustainable farming planning and black soil conservation overall.