A new adaptive approach for detecting crop sowing and harvesting dates from Sentinel-1 time series

Spatially-explicit crop sowing and harvesting dates are crucial for crop modeling, field management, and yield estimation. Synthetic Aperture Radar (SAR) technology, especially the globally available Sentinel-1 time-series data, offers significant potential to monitor these cropping activity dates (CADs) due to all-weather observation capabilities. Recent studies have highlighted the effectiveness of two Sentinel-1 indicators, the VH and Cross-polarization Ratio (CR), in CAD detection. However, their performance varies significantly depending on crop types and growing environments, making the SAR-based methods unreliable for large-scale applications. Hence, we developed an Adaptive Sentinel-1 Indicator Selection (ASIS) framework for CAD detection by comprehensively investigating the mechanisms influencing the capabilities of Sentinel-1 indicators. Our findings suggested that CR is more effective for CADs detection in rain-fed crops with rough soil, while VH performs better in irrigated or aqua fields. Based on this new knowledge, ASIS first employed an iterative time-series clustering strategy to group Sentinel-1 pixels into clusters which represent different crop types and land surface conditions. Then, a new index, RCGI, integrated VH and CR to reliably delineate the crop growing seasons. Last, VH intensity out of the growing seasons was used to automatically identify land surface conditions for selecting the optimal Sentinel-1 indicator (VH or CR) in sowing and harvesting detection. Results show that ASIS had superior accuracy for sowing (MAE = 7.62 days) and harvesting dates estimation (MAE = 8.48 days), making the estimation error of sowing dates (or harvesting dates) significantly decrease by 32.4% (or 7.5%) compared to using VH only and by 52.7% (or 9.5%) compared to CR only. This study demonstrated the effectiveness of the optimal Sentinel-1 indicator in CAD detection, especially for regions without prior information on crop labels and field boundaries.