Phenology-pigment based automated peanut mapping using sentinel-2 images

Reliable spatiotemporal crop data are vital for sustainable agricultural management. However, efficient algorithms that can be automatically applied to large regions scarce, especially cash crops, since it is hard distinguish their uniqueness merely from temporal profiles of traditional vegetation indices. The efficiency knowledge-based features and red-edge pigment indices in characterizing growth has been reported the literature, but potential combined applications identifying crops not validated yet. This study fills this gap by developing a automated Peanut mapping Algorithm with consideration Phenology Pigment content variations (PAPP). earlier longer flowering stages compared other such as paddy rice maize. fields distinguished less anthocyanin chlorophyll well higher carotenoid concentrations. Herein, three phenology pigment-based indicators were proposed peanut exploring concentration chlorophyll, indices, respectively. PAPP algorithm was over (around 250 thousand km2 cropland) covering provinces Northeast China using Sentinel-2 time-series images. results there 8,371 area 2018, concentrated western Jilin Liaoning provinces. Validation 1,102 field survey sites revealed overall accuracies 94%, kappa index 0.87 F1 score 0.91. sensitive thresholding, high classification accuracy could obtained once threshold one indicator roughly defined. thresholds determined based on proportions staple (i.e. maize) historical statistical either show least or largest values these indicators. demonstrates capabilities automatic no requirements further training modifications. makes contributions management society given significant role legume co-delivering food security adapting climate change.