From GPR Amplitudes to Roots: A Numerical Modelling Framework for Mapping Root System Architecture

Abstract Observations of root architecture of in-situ living trees remain limited despite the potential to inform arboriculture, forestry, and ecological investigations. We used dense 500‑MHz ground‑penetrating radar (GPR) to populate a steady‑state groundwater flow model and characterise three‑dimensional root architecture. Collapsed, depth‑normalised diffraction hyperbolae were used to populate a numerical flow model that provided radial continuity between amplitude anomalies. We present the root systems of a Swamp Oak (Quercus palustris) and a River Red Gum (Eucalyptus camaldulensis). The root architecture is dendritic with pronounced asymmetry and different depth distributions. Approximately 80 % of roots were within 5.25 m of the trunk edge, and ~80 % at depths < 60 cm. Azimuthally, the oak had roots growing into a hillslope exposed to wind, while the gum had three dominant roots at 0 degrees, 100 degrees and 200 degrees. Our results represent a robust, repeatable, non‑destructive approach to compare root architecture across species and environments. Scripts.zip — This directory contains the Python scripts I used to create the main figures in the paper and the interactive HTML files for the supporting information. I have not attempted to create a .yaml file describing the library dependencies; the code is provided for reproducibility. I have tried to modify all paths so they direct to files or other directories contained within the other two .zip files. The scripts in the main directory are labelled according to the figure numbers in the manuscript. Many of the final labels and cosmetic adjustments were completed in Illustrator, but these scripts reproduce the underlying data. OakTree.zip — This directory contains the files used for processing and creating the models described in the text. It also includes the raw data and a CSV of their locations. These data were also used in a manuscript submitted in October 2025 and resubmitted with minor revisions in February 2026. The preprint is available here: essopenarchive.org/users/618679/articles/1364950-mapping-coarse-tree-roots-using-ground-penetrating-radar-with-an-angular-grid-acquisition. The data were collected with a Sensors & Software antenna, so every profile is individual. The folders contained within the directory represent each “angular” grid section. GumTree.zip — This directory contains the files used for processing and creating the models described in the text. It also includes the raw swath data. It does not include the drone acquisition data—only the final DEM and hillshades used for mapping. Here we used a MIRA GPR system, so the files are in the form of swaths, not individual lines.