Crop Mapping Based on Temporal and Spatial Sample Migrations: A Case Study Over Three Counties in Heilongjiang Province, Northeast China

Abstract

Crop mapping is crucial for agricultural management and yield prediction. Currently, remote sensing-based crop mapping over a large region is still challenging due to the requirement of sufficient in-season crop samples, which is commonly costly and time-consuming. To address this challenge, a spatial and temporal sample migration method was proposed and evaluated in three typical agricultural counties in Heilongjiang province, Northeast China. On one hand, ground crop samples collected from the previous two years (2020 and 2021) in Nenjiang County were temporally migrated to the target year (2022). On the other hand, ground crop samples collected from the year 2022 in Fujin County were spatially migrated to the adjacent Tongjiang County for mapping crops for the same year. This enabled crop mapping in the absence of current crop samples. In the proposed method, the Sentinel-2 data were primarily used to obtain target curves and reference curves for crop samples. In addition, by balancing the quantity and quality of migrated samples, an optimal migration rule was designed to obtain migrated samples using the dynamic time-warping algorithm over the study area. Finally, the migrated crop samples were used to mapping crop distribution for the target year and region. The results indicated that the overall accuracy can reach 95.7% with the temporal migration in Nenjiang, whereas the spatial migration from Fujin to Tongjiang was approximately 75.6%. The proposed approach reveals significant potential for crop mapping without the knowledge of in-season crop samples, especially for the use of historical crop samples for temporal migration.

Keywords

• Crop mapping
• historical crop samples
• Sentinel-2
• spatiotemporal migration