Results from both the model calibration and validation confirmed
that when using C-band RADARSAT-2 data, the strongest correlations and lowest errors of estimationwere foundwhen the linear cross polarization (HV)was used in the model. Thiswas true for both corn and soybeans.Considering only independent validation data, correlations (Rvalues) between estimated and measured LAI were reported as 0.83 (HH–HV) and 0.81 (VV–VH) for corn; 0.80 (HH–HV) and 0.80 (VV–HV) for soybeans. For corn, RMSE (MAE) errors ranged from 0.84 m2 m-2 (0.65 m2 m-2) to 0.75 m2 m-2 (0.62 m2 m-2) for HH–HV and VV–VH, respectively. When considering soybeans, the like-cross polarization combinations resulted in RMSE errors of 0.64 m2 m-2 (HH–HV) and 0.63 m2 m-2 (VV–HV) between estimated and observed LAI.MAE values were reported as 0.44 m2 m-2 for both like-cross polarization
retrievals. Longer L-Band microwaves penetrate further into
canopies. Although this may be advantageous for larger canopies,
when biomass is low greater direct soil and soil-canopy or canopy-soil interactions are expected. The best results for corn were found when HH–HV backscatter was used to estimate LAI (RMSE of 0.91 m2 m-2;MAE of 0.64 m2 m-2). Although L-band was able to estimate LAI of corn with relatively lowerrors, this lower frequency did not adequately estimate LAI for soybeans. With this lower biomass crop, the lowest errors were achieved with VV–HV (RMSE of 1.26 m2 m-2).