Detection of Soybean Pod Formation Stage Using Sentinel-1 SAR Data

Soybean pod formation and its variability during the crop cycle has direct impact on the final crop yield. Observation using optical data could not be a better alternative for a crop like soybean, which is grown in summer monsoon in India in the Kharif season (crop grown mid June to end September). This study provides an effective method for soybean phenology monitoring and pod formation detection using Sentinel-1 synthetic aperture radar (SAR) based Dual-pol Radar Vegetation Index (DpRVI) and cross-pol ratio (VH/VV). We observed a trend followed by DpRVI and VH/VV. The mean DpRVI observed for 330 test sites soon after the sowing (DOY 181) was $0.28\pm 0.07$, increased to $0.38\pm 0.06$ (DOY 229) then followed a downward trend (dip formation) with mean DpRVI $0.35\pm 0.08$ (DOY 241) and finally reaches to its maximum $0.48\pm 0.09$ just before the harvest (DOY 265). Previous five year data over the same fields showed same pattern of growth and dip, with more accuracy for those years having enough dataset. DpRVI and cross-pol ratio (VH/VV) observed as a better indicator of soybean growth compared to VH and VV polarization backscatter. This study also observed impact of early and delayed rainfall over previous five years on the pod formation. This study finds a pattern of first derivative of phenological profile and identifies that the second minima of first derivative of phenology profile represent the date of pod formation stage for soybean growth.