A novel dual-polarization SAR vegetation index for crop phenology detection

Abstract

Precise crop phenology information is paramount for smart farming and food security. Nevertheless, traditional optical remote sensing methods are susceptible to cloud cover, leading to data discontinuities, which in turn makes the detection of full crop phenology challenging. Synthetic Aperture Radar (SAR) remote sensing technology enjoys the advantage of all-weather observation. However, SAR data has not been systematically employed for comprehensive crop phenological stage detection to date. Consequently, we introduce a novel dual-polarization SAR vegetation index for crop whole phenological stage detection. This method initially unifies SAR data’s covariance elements and intensity information to establish co-polarized principal component parameters (m_cp). Subsequently, a scale parameter (R_cp) is formulated by combining the polarization degree to represent the co-polarization principal component polarization direction within the SAR signal. By combining m_cp and R_cp, a new dual-polarization SAR vegetation index (DRVIs) is devised. Finally, based on time-series DRVIs data, an enhanced shape model fitting technique is utilized to detect the seven phenological stages of winter wheat growth. For validation purpose, this study focuses on the Eastern Henan Plain, employing 89 Sentinel-1 images captured from 2020 to 2022 for experimentation. The results of the experiments manifest that DRVIs exhibit a close correlation with the crop growth process, ranging from approximately 0.2 to 1.0. Winter wheat phenology is detected using DRVIs and NDVI, achieving a remarkable correlation coefficient of 0.94 between the two. Compared with in-situ data, DRVIs are more effective in minimizing phenological detection errors than NDVI. In general, this method effectively delves deeper into the potential of SAR vegetation indices in crop phenology detection, thus broadening the scope of SAR data applications in smart agriculture.