Dielectric permittivity effects in the detection of tree roots using ground-penetrating radar

The use of GPR as a method to estimate root biomass has been studied in recent decades as an alternative to avoid traditional methods that are destructive and repetitive. Biomass estimation with GPR is usually done by estimating the volume of roots or other calculated indexes imaged in the radargram and then converting to biomass using correlation methods. The volume estimation varies with the antenna's response, which depends on physical factors of soil moisture, and root properties such as diameter, orientation, depth, and water retention. In this work, we study how the dielectric permittivity variation of roots affects the GPR-derived root diameter estimates. We show that the relationship between dielectric permittivity and the amount of water present in each root affects the imaged diameter. We calculate the dielectric permittivity for roots samples with different moisture content. We acquired sections through a 2 m diameter sandbox, and compared our results with synthetic models. We noticed a correlation between the analyzed indices and the relative dielectric permittivity (er) in the synthetic data that indicates the diameter estimate may vary widely with roots moisture. We observed that the indexes extracted from the radargrams did not correlate with our independent estimates of dielectric permittivity. To estimate the roots' diameter, we used 40% of the image amplitude generated after the Hilbert transform as a parameter, which resulted in an average error equal to 8%. We also propose a correction factor to eliminate the effects of the amount of water on the diameter estimation. This processing step reduced the diameter estimation error on synthetic models.