Crop Biophysical Parameters Estimation with a Multi-Target Inversion Scheme using the Sentinel-1 SAR Data

In this paper, a multi-target inversion scheme is adopted for joint estimation of crop biophysical parameters from dual-pol SAR data. The single-output support vector regression (SVR) method is extended to a multi-output support vector regression (MSVR) method to estimate biophysical parameters. The MSVR is implemented for simultaneous retrieval of plant area index (PAI) and crop biomass from the Sentinel-l C-band dual-pol (VV + VH) data. In this particular study, the inversion algorithm is trained and validated for the canola crop using in-situ measurements collected during the Soil Moisture Active Passive Validation Experiment 2016 (SMAPVEX16-MB) Manitoba campaign. The validation results indicate a good correlation coefficient (r) of 0.72 and 0.85, with a RMSE of 0.35 m
²
m
^-2
and 0.48 kgm
^-2
for PAI and wet biomass respectively. In addition, the mapped PAI and wet biomass values at flowering stage of canola capture the variability in crop growth from Sentinel-l data.