In-season crop mapping over China’s three major grain-producing regions using progressive Sentinel-1/2 time-series

In-season crop mapping is critical for real-time agricultural monitoring, yet remains challenging under incomplete time series and complex cropping systems. This study develops anin-season crop mapping framework across China’s three major grain-producing regions, namely the Northeast China Plain, the North China Plain, and the Middle and Lower Yangtze River Plain, by integrating Sentinel-2 and Sentinel-1 time-series data at 10 m resolution with a Random Forest classifier under progressive time-series accumulation. Classification performance and earliest identifiable time (EIT) are evaluated across different training strategies, and temporal separability analysis is employed to reveal the phenological mechanisms underlying early crop identification. Results demonstrate that classification accuracy increases with accumulating temporal observations and stabilizes at key phenological stages. Crop- and region-specific differences in EIT are identified: in the Northeast China Plain, rice becomes identifiable as early as DOY 140, while maize and soybean require mid-to-late season observations (DOY 190–200); in the North China Plain, winter wheat can be distinguished before overwintering (~DOY 340), whereas summer maize and soybean converge later (~DOY 200–210); in the Middle and Lower Yangtze River Plain, early, middle-, and late-season rice stabilize around DOY 140–150, 160–190, and 190–200, respectively. Cross-year transfer reduces performance compared with within-year training, while feature selection based on the top-50 features improves cross-year transferability, particularly for winter crops. The integration of Sentinel-1 SAR data provides limited benefits for upland crops but yields noticeable improvements for complex rice systems and certain winter crops. Temporal separability analysis confirms that separability peaks align with key phenological stages and remain broadly consistent across years, confirming that early identification capability is primarily driven by phenology-induced spectral contrasts. These findings provide a scientific basis for advancing in-season crop monitoring under complex cropping systems, with implications for timely agricultural management, market forecasting, and policy decision-making.