Threshold‐based classification of rapeseed using multi‐spectral vegetation indices from Sentinel‐2

Abstract Accurate mapping of rapeseed ( Brassica napus ) fields is critical for effective crop management, pollinator support, and yield forecasting. This study systematically evaluated the performance of 23 vegetation indices (VIs) for rapeseed detection during the peak flowering stage using high‐resolution Sentinel‐2A imagery across two contrasting agricultural regions: North Dakota and Spain. Each VI was assessed within a threshold‐based binary classification framework. The normalized difference yellowness index (NDYI) consistently emerged as the most effective index in both study areas, achieving the highest classification performance in North Dakota (overall accuracy = 0.97, F1 score = 0.95, Kappa = 0.93, and producer's accuracy = 0.94) and in Spain (overall accuracy = 0.96, F1 score = 0.64, Kappa = 0.62, and producer's accuracy = 0.51). Other indices, notably green leaf index and CI, also showed relatively strong performance, while traditional greenness‐based indices such as normalized difference vegetation index, soil‐adjusted vegetation index, and modified soil‐adjusted vegetation index achieved only moderate accuracy. Several color‐sensitive indices clearly outperformed conventional metrics in distinguishing rapeseed from spectrally similar summer crops, whereas indices such as modified yellow index and high‐resolution flowering index exhibited poor classification results in both regions. The optimal NDYI thresholds—0.69 in North Dakota and 0.60 in Spain—were closely aligned, indicating minimal regional variability and underscoring the robustness and transferability of NDYI for rapeseed mapping. Overall, these findings provide practical guidance for selecting suitable VIs and highlight the importance of spectral yellowness for accurate crop classification during the flowering stage.