Development of a 10-m resolution maize and soybean map over China: Matching satellite-based crop classification with sample-based area estimation

Abstract

Spatially explicit information on crop distribution is essential for market information, food security, and agricultural sustainability. However, high-resolution crop maps are unavailable for most countries of the world. In this study, we developed an operational workflow and produced the first openly-available 10-m resolution maize and soybean map over China. We also derived area estimates for maize and soybean extent for 2019 using a stratified, two-stage, cluster sampling design and ground data collected for the entire country. We developed a multi-scale, multi-temporal procedure for mapping, in which field data were used as training to map maize and soybean over the first-stage sample of 10 km × 10 km equal-area blocks with PlanetScope and Sentinel-2 data. Then, the classified blocks were used as training to map maize and soybean for the country with wall-to-wall Sentinel-2 data using a random forests approach. We used all available Sentinel-2 surface reflectance data acquired between April and October 2019, applied quality assurance, including cloud and shadow masking, and created monthly image composites as inputs for the random forests analysis. We derived maize and soybean area estimates using the field sample data and a regression estimator. Utilizing the probability output layer of the random forests models, we found and applied empirical probability thresholds that matched map-based crop area estimates with sample-based area estimates. Maize area in China in 2019 was estimated to be 330,609 ± 34,109 km2 (± value is the standard error), and soybean area was estimated to be 78,107 ± 12,969 km2. Validated using the field sample data as reference, our crop map had an overall accuracy of 91.8 ± 1.2%. The user's and producer's accuracies for the maize class were 93.9 ± 2.5% and 79.2 ± 3.6%, and for the soybean class were 63.6 ± 12.1% and 61.9 ± 11.8%. Our map-based maize and soybean area estimates had close agreement with government reports at the provincial and prefectural levels, with r2 of 0.90 and 0.92 for maize, and 0.93 and 0.94 for soybean, respectively. Our workflow can generate internally consistent results for crop area estimation and crop mapping simultaneously. As Sentinel-2 data are being acquired consistently and very-high-resolution commercial satellite data are increasingly available, our established workflow may be applied in an operational setting for annual crop mapping in China and other countries.