Retrieving Rice FPAR Using Freeman-Durden Decomposition of RADARSAT-2 Data

This paper aims to explore the feasibility of the Freeman-Durden decomposition parameters from RADARSAT-2 data for the fraction of photosynthetically active radiation (FPAR) retrieval in paddy rice. The study area was located in Meishan City, Sichuan Province, China. Four fine quad-polarized single-look complex (SLC) RADARSAT-2 SAR images were acquired, and field observations were conducted about every 12 days from 2 May to 8 August in 2016 at 30 sample sites. The powers of the three scattering mechanisms (surface, double-bounce and volume scatterings) of Freeman-Durden decomposition were extracted from the multi-temporal RADARSAT-2 data. It was found that in a complete rice growth cycle, as the rice plants grow, the surface scattering component power decreases, the volume scattering component power increases, and the double-bounce scattering component power peaks in the tillering phase of rice, reflecting the changes of the interaction between electromagnetic waves and rice plants at different growth stages. Regression analysis of the three-component powers and their mathematical combination with the measured FPAR data of rice showed that the ratio between the volume scattering and the surface scattering had the highest correlation with FPAR, and the coefficient of determination reached 0.7. At the same time, the backscattering coefficient extracted from the RADARSAT-2 data was also used for regression analysis with FPAR, and the best-performing polarization combination was the ratio of VV and VH polarization, and the coefficient of determination was 0.64. The results show that the Freeman-Durden decomposition parameters, which retain information on intensity and phase, can better respond to changes in rice physiological structure compared to the backscatter coefficient that retains only the intensity information.