The dependence of synthetic aperture radar backscatter on forest structure and biomass: Potential application for global carbon models

In this thesis, the NASA airborne P-band (0.438 GHz), L-band (1.25 GHz), and C-band (5.3 GHz) quadpol. SAR system was used to collect data for a series of tropical broadleaf evergreen forests on the Island of Hawaii. The SAR data were regressed against biomass measurements made in the field and the response curves for the tropical forests were compared to those made for coniferous forests in North America and Europe using the same SAR instrument and imaging angles (40{dollar}spcirc{dollar}-50{dollar}spcirc{dollar}). Results indicated that the response curves for the tropical forests and the coniferous forests were similar and that the radar signals were saturating at relatively low biomass levels. Biomass saturation points were determined at {dollar}approx{dollar}100 tons/ha for P-band, {dollar}approx{dollar}40 tons/ha for L-band, and {dollar}approx{dollar}20 tons/ha for C-band (HH, VV, and HV polarization). The possible existence of a universal saturation point for the SAR response to forest biomass distinctly limits the usefulness of P-, L-, and C-band SAR for global biomass mapping. A relatively small percentage of the world's vegetated systems fall below the highest estimated saturation level. Approximately 46% of the world's vegetated surface area containing 82% of the estimated total store of biomass lies above the saturation limit of the current radar systems ({dollar}>{dollar}100 tons/ha for P-band). While 54% of the Earth's vegetated surface area is below the saturation level for P-band, this class contains only an estimated 18% of the total biomass represented by terrestrial vegetation. Theoretical modeling indicated that the primary forest canopy structural factor influencing SAR backscatter was the surface area to volume ratio (SA/V) of the branches. This proved true for both broadleaf evergreens and conifers. As a forest stand matures, the form of the phytomass coalesces into fewer larger components and the calculated SA/V declines as biomass increases. As the SA/V declines the backscatter tends to increase since the radar has larger components from which reflections can occur.