Can ground-penetrating radar detect adjacent roots and rock fragments in forest soil?

Ground-penetrating radar (GPR) has been used for estimating root biomass, estimating pull-out resistance force of roots, and reconstructing root system architecture. Although GPR can estimate the diameter of a single root, it has not yet been verified whether it can successfully estimate the diameters of adjacent roots or whether rock fragments occurring in the soil matrix can be misidentified as roots. This study aimed to (1) evaluate the visual properties of GPR images of root bundles and rock fragments, and (2) clarify whether they have the potential to create non-negligible errors in GPR surveys. Root bundle samples buried in a sandy soil bed were scanned using 900 MHz and 1500 MHz antennas. Sole roots of Cryptomeria japonica and rock fragments in the bed were also scanned. A bundle of root formed one hyperbola with an apex in the radar profile, like a sole root. However, the contrast of the hyperbolas was weak and GPR indices relating to diameter were reduced compared to those of the comparably sized sole root. Stones did not form clear images, but boulders were faintly visible with non-negligible GPR indices. For evaluation of root biomass, adjacent roots had the potential to lead to significant errors. Similarly, in reconstructions of root system architecture, adjacent roots might lead to incorrect modelling of root point connections, because of incorrect diameters. For pull-out resistance force, adjacent roots could lead to only underestimation, i.e., safe-side error. Misidentification of rock fragments as roots depends on their sizes.