Winter wheat classification within heterogeneous agricultural landscapes

Winter wheat plays a crucial role in ensuring food security and promoting sustainable agriculture. In the Liangshan Yi Autonomous Prefecture of Sichuan Province, which serves as a significant source region for wheat stripe rust in China, accurately mapping the distribution of winter wheat is essential for monitoring wheat diseases. However, the region’s challenging terrain and diverse crop systems complicate the precise mapping of winter wheat cultivation. To address this issue, this study utilizes high-resolution Gaofen-2 satellite (GF2) data in Xichang City, Liangshan Yi Autonomous Prefecture, to establish a novel Winter Wheat Extraction Index (WWEI). An object-oriented approach is employed to quantify the likelihood of winter wheat cultivation across various land patches. The WWEI capitalizes on the distinctive characteristics of the Kernel Normalized Difference Vegetation Index (kNDVI) and the Normalized Difference Water Index (NDWI), both derived from GF-2 data during the germination-to-headinggrain-filling stages of wheat growth. Furthermore, adaptive parameters and binary classification thresholds are utilized to convert WWEI values into a comprehensive winter wheat map. The results indicate that the WWEI generates an accurate classification map, achieving an overall accuracy of $89 \%$ and F1 scores exceeding 0.89. The wheat cultivation area derived from the map, when compared with official statistics, shows an $R^{2}$ value above 0.8. Thus, the WWEI demonstrates considerable flexibility and applicability in mapping the distribution of winter wheat in Xichang City due to its independence from reference samples and predefined parameters. This delineation of the spatial distribution of wheat stripe rust host crops enhances the accuracy of disease monitoring and forecasting, thereby enabling agricultural managers to implement swift and targeted control measures to ensure stable and sustainable agricultural production.