The forward modeling exercise compared backscatter estimated fromEq. (7)to observed backscatter coefficients (Fig. 2;Table 4). For C-band RADARSAT-2 and corn, the correlation coefficients (R values) between the observed and estimated backscatter coefficients were 0.40 (HH), 0.46 (VV) and 0.82 (HV) (Table 4). Comparable results were found for soybeans where R values of 0.45 (HH), 0.47 (VV) and 0.80 (HV) are reported. These results confirm that the WCM model
was better calibrated for the linear cross-polarization (HV) than for the HH or VV polarizations. The sensitivity to HV backscatter is explained by the presence of significant crop canopy and the known response of cross-polarization to volume scattering within these canopies.At longer L-Band wavelengths, the model was well calibrated for corn using any of the three linear polarizations (R values of 0.91 (HH),0.88 (VV) and 0.93 (HV)). Correlations were most likely higher at LBand relative to C-Band given the ability of this longer wavelength to penetrate further into the canopy. Deeper penetration means that Lband microwaves interact with a larger volume of the corn canopy.
The reverse was observed for soybeans. For this lower biomass crop, a longer wavelength may not be advantageous as greater direct soil and canopy–soil (or soil–canopy) interactions are present. Given these interactions, correlations at L-Band were lower (R values of 0.24 (HH),0.31 (VV) and 0.37 (HV))