Crop Phenology Mapping using Polarimetric Parameters extracted from Sentinel-1 Images

Abstract

Accurate and high-resolution Spatio-temporal information about crop condition and phenology is a vital component for crop management and yields estimation at the local scale to regional scale. In this research, the crop phenology estimation is carried out using time-series Sentinel-1 dual-pol data. Sentinel- 1 data was acquired from January 2019 to December 2020 for Chinoat city of Pakistan. Backscattering coefficients (σ_o) for VH and VV channels were computed for each acquired image. Crop calendar for the local crop of Chinoat was acquired and (σ_o) were stacked according to the cropping season of rice, maize, and wheat. The unsupervised classification was performed using the ISO-Data clustering technique. The mean of each cluster was extracted corresponding to each data of acquisition and polarimetric parameter-based phenological profiles were plotted. Hermite polynomial fitting was performed to acquire smooth phenological profiles. Extracted phenological profiles were compared with the local crop calendar and the following crops were identified: rice, maize, and wheat-based on sowing, growth, and harvesting time information. The (σ_o) in VV channel does not provide consistent results that is why it was discarded from the analysis. However, (σ_o) in VH channel provides very precise crop profiles that coincide with the cropping pattern in the crop calendar. Finally, crop phenology mapping was carried and final crop maps are prepared.