Towards automation of national scale cropping pattern mapping by coupling Sentinel-1/2 data: A 10-m map of crop rotation systems for wheat in China

Abstract

Context

Wheat, as the world's largest cereal crop, contributes significantly to agricultural intensification through crop rotation systems. Updated knowledge of cropping patterns (CP) describing crop rotations is crucial for the development of sustainable agricultural systems. However, there is a gap in data availability and finer resolution CP maps are not available for most countries, which hampers our knowledge of geographically targeted crop rotation for sustainable management. It is challenging to automatically map CP at large scales due to the lack of ground-truth datasets, the complexity of crop rotation systems, and the limited applicability of existing algorithms.

Objective

This paper has three objectives: 1) propose approaches for automatic mapping of wheat cropping patterns; 2) assess its capability through its applications over conterminous China; 3) explore the distribution patterns for wheat of crop rotation systems in China.

Methods

This study introduced a novel framework for automatic agricultural mapping by proposing CP indices based on coupled patterns of multi-source imagery and inter-seasonal variations. This study developed the first 10-m wheat Cropping Patterns (ChinaCP-Wheat10m) distribution map over conterminous China by proposing a robust algorithm for mapping Wheat cropping Patterns by fusing Sentinel-1 SAR and Sentinel-2 MSI data (WPSS).

Results and conclusion

The ChinaCP-Wheat10m map showed that wheat dominated the north of the Yangtze River and east of the Taihang Mountain, with a distinctive spatial pattern of winter wheat-rice or upland crops divided by the Huaihe River. There was 206,919 km² of wheat sown area in China in 2020, and over 90 % of national wheat cultivation was implemented by double cropping. More than half of national wheat farming was intensified through rotation by maize (51.39 %), followed by paddy rice (21.12 %) and other upland crops (18.90 %). There was a small proportion of single cropping by spring wheat (6.86 %) and winter wheat (1.73 %). The reliability of the WPSS was validated by 17,627 widely distributed reference sites with an overall accuracy of 92.57 % and good agreement with the agricultural census data (R² = 0.96).

Significance

This study opens a new direction to move from crop type identification to the automatic generation of crop rotation maps at the national scale, which would facilitate the progress of the Sustainable Development Goals (SDGs) to reduce poverty and hunger. The processing codes and wheat CP records produced in China can be downloaded from the following link: https://doi.org/10.6084/m9.figshare.28668173.v1