10-m Soybean Cropland Maps of China Produced
Jun 07, 2024
Researchers from the Aerospace Information Research Institute (AIR) at the Chinese Academy of Sciences (CAS) have developed a spatiotemporal migration method to accurately map China's soybean cropland. Leveraging Sentinel-1/2, the innovative technique has yielded 10-m maps of China's soybean cropland spanning from 2019 to 2022. The study was published in Scientific Data.
China, the world's largest consumer and fourth-largest producer of soybeans, faces challenges in achieving domestic soybean self-sufficiency, which stood at only 15% in 2021, far below the internationally recommended level for food security.
Accurate and high-resolution maps of soybean cropland are essential for effective cultivation management and ensuring global food security. However, there are many challenges to draw national soybean cropland maps in China. These include time-consuming sample collection, fragmented small-scale farming, and cloud interference affecting optical remote sensing imagery.
Remote sensing technology can provide large-scale, long-term land observation data and obtain crop growth information. By analyzing the physical and chemical properties of the soybean canopy and its growth characteristics, researchers extracted various temporal features using time-series Sentinel-1/2 images. They then employed the random forest algorithm to accurately map soybean cropland.
Researchers collected soybean samples in multiple provinces from 2019 to 2022, and used a spatiotemporal migration method to generate samples of non-sample areas and years for model training and results verification. The sample migration process started by creating random points of cropland and other areas using the ESA·WorldCover dataset. Then, they used the soybean canopy's high chlorophyll content to propose the Concavity-Convexity Index (CCI) for initial sample filtering. Finally, they built feature spaces using six bands and vegetation indices to calculate Mahalanobis distances, selecting the most reliable soybean samples for accurate mapping.
The accuracy of the maps was validated and found to be about 93%. This high level of accuracy, in alignment with statistical yearbook data, confirms the reliability of the mapping results.
The new data could be used to analyze spatial patterns of soybean cultivation, estimate soybean crop yields, and enhance soybean production and management practices in China, thus contributing to global food security.
Researchers will continuously refine and expand their mapping methodologies with the primary goal of enabling near-real-time extraction of soybean cropland, not only in China but also in other major soybean-producing regions worldwide.
Contact: luyq@aircas.ac.cn
News & Events