Although China is the world's largest soybean consumer and fourth largest soybean producer, there is a lack of high-resolution annual maps depicting the area of soybean cultivation in China. To address this gap, we have developed a new method called Phenology and Pixel Based Soybean Area Mapping (PPS), which is based on Sentinel-2 remote sensing images from the Google Earth Engine (GEE) platform. We use various auxiliary data, such as cultivated land layers and detailed phenological observations, to select the unique features that most effectively distinguish soybeans from other crops in different regions. Then these features are input into an unsupervised classifier (K-means) and the most likely type is determined through a post classification method based on dynamic time warping (DTW). We have generated the first dataset of soybean planting area in China from 2017 to 2021, with a spatial resolution of up to 10 meters. From 2017 to 2020, the value between the mapping results at the county and prefecture levels and the census data remained around 0.85. In addition, the overall accuracy of on-site mapping results in 2017, 2018, and 2019 was 77%, 84%, and 88%, respectively. Compared with the existing 10m crop type map in Northeast China based on field samples and supervised classification methods, the mapping accuracy has significantly improved by 31%, especially at the county level, based on its consistency with census data. ChinaSoyArea10m spatially matches well with the existing two datasets. ChinaSoyArea10m provides important information for sustainable soybean production and management, as well as agricultural system modeling and optimization.
Sentinel-2 remote sensing images on the Google Earth Engine (GEE) platform.
Unsupervised classifier (K-means) and determines the most likely type through a post classification method based on dynamic time warping (DTW).
In order to compensate for the lack of long-term soybean planting area maps in China, we first generated a dataset of soybean planting areas in major production areas from 2017 to 2021, with a spatial resolution of 10 meters (ChinaSoyArea10m). Compared with existing datasets, ChinaSoyArea10m has higher consistency with census data and further improves spatial details. This dataset can provide reliable support for subsequent yield monitoring and food security research.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB6433.2024
10.5281/zenodo.10071427