Root biomass accounts for about 25% of the carbon storage in mid-latitude forests. Estimation of root biomass for carbon cycling studies requires either direct measurement by excavation of root systems, or remote measurement using ground penetrating radar (GPR) or other geophysical methods. This study evaluated the ability of a 2-GHz multi-channel GPR system (IDS Hi-BrigHT) to detect and map white pine roots in managed forest near Turkey Point, southern Ontario. The GPR system employed eight dual-polarized antenna pairs separated at 10 cm intervals. GPR data were acquired as overlapping swaths (2 cm line spacing, 0.4 cm inline) across a 25-m2 test site (TP74-R) containing a juvenile white pine tree. Radargrams were processed to full 3-D radar volumes for time slicing and interpretation of root architecture and comparison with the excavated root network. Radargram signal processing was successful in suppressing airwave and other background noise and improved the detection of root diffractions on radargrams. The majority of roots were found in the rooting zone at a depth of 5-40 cm. Roots as small as 0.5 cm were detected with the 2-GHz frequency, but many roots <1.5 cm diameter could not be detected as continuous root structures. Root detection was strongly dependent on root orientation; large, coarse roots (>3-5 cm) were imaged as continuous root segments when oriented perpendicular to GPR profiles. Roots intersecting GPR profiles at angles <30-45 degrees were either imaged incompletely or not detected on radargrams. The highest rate of root detection was achieved with horizontally polarized (HH) antennas (dipole axis parallel with the root structures). Isosurface root models constructed from the Hilbert-transformed radargrams allowed mapping of the 3-D dimensional root architecture for large (> 3-5 cm diameter) roots. Isosurface models provide a means for estimating the coarse root volume for large roots and could be employed in future work to monitor temporal changes in root biomass by repeat survey at the same measurement site. Radargram signal processing was successful in suppressing airwave and other background noise and improved the detection of root diffractions on radargrams. The majority of roots were found in the rooting zone at a depth of 5-40 cm. Roots as small as 0.5 cm were detected with the 2-GHz frequency, but many roots <1.5 cm diameter could not be detected as continuous root structures. Many roots were not detected due to dependence of root reflection amplitude on root orientation. Roots oriented at >30-45 degrees to the survey swaths were imaged incompletely or not detected. Most large coarse roots (>5 cm diameter) were mapped as continuous structures when the root orientation was either parallel to, or perpendicular to the GPR transects. The highest rate of root detection was achieved with the horizontally polarized (HH) antennas, with the dipole axis perpendicular to the root structures. Isosurface root models constructed from the Hilbert-transformed radargrams allowed mapping of the 3-D dimensional root architecture for large (> 3-5 cm diameter) roots. The isosurface models provide a means for estimating the coarse volume and belowground biomass but further work is required to improve 3-D image resolution to allow detection of the entire root network. The method could be employed to measure the temporal changes in root biomass by conducting repeat surveys at the same measurement site. Radargram signal processing was successful in suppressing airwave and other background noise and improved the detection of root diffractions on radargrams. The majority of roots were found above a depth of 40 cm with the root zone being detected at a depth of10-15 cm. Roots as small as 0.5 cm were detected with the 2-GHz frequency, but many roots <1.5 cm diameter could not be detected as continuous root structures. Many roots were not detected due to dependence of root reflection amplitude on root orientation. Roots oriented at >30-45 degrees to the survey swaths were imaged incompletely or not detected. Most large coarse roots (>5 cm diameter) were mapped as continuous structures when the root orientation was either parallel to, or perpendicular to the GPR transects. The highest rate of root detection was achieved with the horizontally polarized (HH) antennas, with the dipole axis perpendicular to the root structures. Isosurface root models constructed from the Hilbert-transformed radargrams allowed mapping of the 3-D dimensional root architecture for large (> 3-5 cm diameter) roots. The isosurface models provide a means for estimating the coarse volume and belowground biomass but further work is required to improve 3-D image resolution to allow detection of the entire root network. / Thesis / Master of Science (MSc) / Forests play an important role in the global carbon cycle by removing carbon from the Earth’s atmosphere and storing it in tree tissues as biomass. Estimation of the amount of biomass and carbon stored in forests is critical to predictive climate change models, and increasingly employs remote sensing methods to detect both the above ground biomass (e.g. leaves, tree branches) and the belowground carbon in the tree root system. Measurement of the belowground biomass is most difficult, as it cannot be directly observed without destructive excavation of the tree root system. This study investigated the application of new technology, multi-channel ground penetrating radar (GPR), for mapping tree root systems. The GPR system (IDS Hi-BrigHT) employs ‘swath mapping’ using high frequency pulsed radio waves and multiple transmitting and receiving antennas to produce detailed maps of roots structure. The GPR capabilities were evaluated at a test site at the Turkey Point Flux Station (TPFS) in southern Ontario. The root system of a juvenile white pine tree (20-30 cm diameter) was swath mapped over a 25-m2 area with a line spacing of 2 cm. The GPR data were processed to produce a 3-dimensional radar volume, which can be ‘sliced’ in various orientations to reveal the root structure. The time slice maps show that roots as small as 1-cm can be detected and roots larger than 3 cm in diameter can be mapped as continuous root segments.

La végétation arborée sur les digues ou les barrages en remblai est un facteur de fragilité pouvant favoriser l'apparition de mécanismes de détérioration par érosion. Définir la structure géométrique des systèmes racinaires ainsi que la nature des sols dans lesquels ils croissent, à partir de méthodes géophysiques non destructives, est nécessaire afin d'appréhender les conséquences de leur développement sur la sécurité d'un ouvrage hydraulique. Des expériences en laboratoire ont d'abord permis de déterminer les propriétés acoustiques et électriques intrinsèques d'échantillons racinaires menant à l'identification de signatures indispensables à la discrimination de l'anomalie liée à la racine sur le terrain. Le montage de supports expérimentaux adaptés nous a conduits à progressivement étudier des paramètres d'influences en conditions contrôlées. Des essais en conditions semi-contrôlées sur un dispositif d'arbres plantés dans un sol homogène ont permis d'évaluer la pertinence des différentes méthodologies d'acquisition, tels que l'utilisation de la polarisation provoquée en tomographie de résistivité complexe ou encore la géométrie des capteurs pour la tomographie acoustique. Des traitements innovants tels que l'analyse en ondelettes sont associés afin d'exploiter la richesse des enregistrements. Les résultats obtenus ont été validés par le relevé des positions réelles des racines. Enfin, des campagnes d'auscultation sur un ouvrage réel ont été réalisées et ont permis de mettre en évidence une variabilité spatiale du corps de digue associée à la présence d'arbres. Une méthodologie adaptée au diagnostic géophysique de la végétation sur les ouvrages a été mise en place. / Woody vegetation from earth dikes or dams is a fragility factor which can promote mechanisms of degradation such as erosion. An accurate assessment of root system structure, from geophysical non-destructive methods, of root position into the embankment (depth, extension), and a good knowledge of soil conditions are critical in order to anticipate the consequences of vegetation development for the hydraulic structure’s safety. Laboratory experiments allowed determining intrinsic acoustical and electrical root properties leading to identify relevant signatures and discriminate anomalies related to roots in the field. The establishment of adapted experimental devices led us progressively to assess different parameters (roots mass, water content. . . ) under controlled conditions. Experiments in semi-controlled conditions with trees planted into a homogenous soil, were conducted to assess the relevance of different methodologies, such as the use of temporal induced polarization in complex resistivity tomography or the geometry of sensors for acoustical tomography. Innovative data processing such as wavelet analysis were used to valorize the rich database. The results were validated by the determination of actual root position.Finally, field investigations into an embankment have been performed to highlight a spatial variability of dike structures associated with trees presence. A methodology adapted to the geophysical diagnostic of vegetation roots in embankments was developed.