Large-scale mapping of plastic greenhouses in plateau regions based on Sentinel data

Plastic greenhouses have gained widespread global adoption due to their advantages in yield and quality, yet the film materials used have faced significant environmental concerns. Accurate mapping of plastic greenhouse distribution holds critical importance for both agricultural development and environmental protection. However, the fragmented distribution and spatial heterogeneity of land cover types in plateau regions pose challenges for large-scale mapping of plastic greenhouses. To address this issue, this study compared two research methodologies for plastic greenhouses and proposed a method suitable for large-scale mapping studies in Chenggong District. The experiment demonstrated that: (1) The Plastic Mulch Layer Index (PMLI), Moment Distance Index (MDI), Greenhouse Detection Index (GDI), Plastic Greenhouse Index (PGI), and Advanced Plastic Greenhouse Index (APGI) yielded unsatisfactory classification results for plastic greenhouses in Chenggong District; (2) Shapley Additive Global Importance (SAGE) analysis revealed that the B1, B2 and B12 bands play a crucial role in studying plastic greenhouses in plateau regions; (3) Combining Sentinel-1 and Sentinel-2 data improves model accuracy and plastic greenhouse classification precision compared to using Sentinel-2 data alone; (4) The experimental approach combining SAGE feature selection with the XGBoost model proved to be the optimal classification scheme, achieving an overall accuracy of 96.42%, an F1 score of 96.42% and a Kappa coefficient of 0.9205. This study provides methodological support for plastic greenhouse research in plateau regions.