Correlating radar backscatter with components of biomass in loblolly pine forests

A multifrequency, multipolarization airborne SAR data set was utilized to examine the relationship between radar backscatter and the aboveground biomass. This data set was also used to examine the potential of SAR to estimate aboveground biomass in these forests. The total aboveground biomass in the test stands used in this study ranged from <1-50 kg m/sup -2/. Not only was total aboveground biomass considered, but the biomass of the tree boles, branches, and needles/leaves. Significant correlations were found in all three frequencies of radar imagery used in this study (C-, L- and P-band), At P- and L-bands, the greatest sensitivity to change in biomass occurred in the HH and VH polarized channels, while at C-band, the greatest sensitivity was in the VH polarized channel. The results of the correlation analyses support modeling studies which show the dominant scattering mechanisms from these pines should be double-bounce, ground-trunk scattering and canopy volume scattering. To produce equations to estimate biomass, a stepwise, multiple-linear regression approach was used, using all the radar channels as independent variables, and the log of the biomass components as the dependent variables. The results of this regression analysis produced equations with high coefficients of linear correlation and low standard errors of the regression equation for estimating total stand, bole and total stem biomass. Statistically-significant regression equations were also generated for large stem, small stem and needle/leaf biomass. Even though the biomass estimation algorithms had high correlation coefficients and low standard errors, when the predicted biomass estimates were expressed in arithmetic terms and compared to actual values, low levels of accuracy were found. A second method was developed using total stem biomass to estimate the other components, with total stem biomass being estimated from the radar image intensity values. This two-step method reduced the coefficient of variation to between 16 and 27% for all biomass components.<
>