Developing a Phenology- and Pixel-Based Algorithm for Mapping Rapeseed at 10 M Spatial Resolution Using Multi-Source Data

As a major oilseed crop, large-scale and high-resolution maps of rapeseed (Brassica napus L.) are critical for predicting annual production and ensuring global energy security. However, such free maps are still unavailable in large areas. We designed a new pixel- and phenology-based algorithm and produced a new data product for rapeseed planting area (2017–2019) over 33 countries at 10-m spatial resolution based on the multiple data. The product showed a good consistence (R2 = 0.88) with the official statistics (Food and Agricultural Organization of the United Nations, FAO) at national level. Rapeseed maps achieved at least 0.81 F1-scores of spatial consistency when comparing with the Cropland Data Layer (CDL) of America, Annual Crop Inventory (ACI) in Canada and Crop Map of England (CROME) in England. Moreover, their F1-scores ranged 0.84–0.92 based on the independent validation samples, implying a good consistency with ground truth. Furthermore, we found that rapeseed crop rotation is ≥2 years in almost all countries. Our derived maps with high accuracy suggest the robustness of pixel- and phenology-based algorithm in identifying rapeseed over large regions with various climate and landscapes. The derived rapeseed planting areas freely downloaded can be applied to predict rapeseed production and optimize planting structure. The product is available publicly at http://dx.doi.org/10.17632/ydf3m7pd4j.3 (Han et al., 2021).