Data and code for "Geophysical imaging reveals subsurface control on vegetation patterns on granitic outcrops"

This repository contains all data, processing scripts, and figure-generation code associated with a study of subsurface weathering structure and its control on vegetation patterns at a granitic outcrop in the Piedmont of South Carolina, United States. The study combined seismic refraction tomography, airborne LiDAR, multispectral (NAIP and Planet Labs) imagery, and soil geochemistry to test whether observed vegetation patterns are controlled by pre-existing subsurface structure rather than being generated by the vegetation itself. Study siteThe field site is Boggs Rock, a granitic outcrop located in the South Carolina Piedmont, southeastern United States. Two seismic refraction profiles (Line 1 and Line 2), each 192 m long, cross sharp gradients in vegetation from exposed bedrock to shrub and closed-canopy forest. Thirty-seven soil samples were collected along Line 1 and analysed for pH, cation exchange capacity (CEC), elemental concentrations (P, K, Ca, Mg, Na, Zn, Mn, Cu, B, NO₃), organic matter, total carbon, total nitrogen, and C:N ratio. Data contentsThe repository is organised into the following directories: data/Seismic/ — Raw and processed seismic data for both profiles. Includes stacked SEG-Y shot gathers, LiDAR-derived topography files used for mesh construction, manually picked first-arrival travel-time files, the full ensemble of 120 individual VTK inversion results per line (located in runs_multiThread/), and ensemble-mean model outputs including velocity, standard deviation, ray coverage, and vertical velocity gradient. Diagnostic figures (shot gathers, model fits, misfit by shot location) are also included. data/SoilSamples/ — Soil sample data in CSV format (Boggs_SoilSamples.csv, soil_sample_results.csv) and the original laboratory analysis reports in PDF format (Soil_Reports/). data/Spatial/ — KML files defining the GPS coordinates of both seismic line locations (Boggs_L1.kml, Boggs_L2.kml). Note: large raster datasets (LiDAR point clouds, NAIP imagery, Planet Labs imagery) are not redistributed here as they are publicly available from their original sources (NOAA Digital Coast for LiDAR; USDA for NAIP). src/ — Shared Python modules. boggs_data.py is the central data-loading class that reads all datasets and provides extraction and interpolation methods used by the figure scripts. plot_utils.py defines shared plotting constants and helper functions. figures/ — One Python script per manuscript figure. Each script is self-contained and reproducible from the data in this repository. outputs/ — All manuscript and supplementary figures generated by the figure scripts, provided as PNG files at 300 dpi. Processing workflowSeismic data were processed in three stages. First, repeated shot records were stacked using stackData.py. First-arrival travel times were then picked manually. Finally, travel-time tomography was performed using a bootstrap ensemble approach implemented in StreamLineInversion_Pwave_bootstrap_mulitThread.py (Line 1) and StreamLineInversion_Pwave_bootstrap_mulitThread_L2.py (Line 2), using the PyGIMLi open-source geophysical inversion framework. Ensemble results were averaged and ray coverage computed using averageVTKFiles_traceRays.py. Full details of the acquisition parameters, inversion workflow, and uncertainty quantification are provided in the Supplementary Material of the associated publication. Software and dependenciesAll scripts are written in Python 3. Required packages are listed in requirements.txt and include numpy, scipy, matplotlib, pandas, segyio, and pygimli. A conda environment file (environment.yml) is also provided. Scripts were developed and tested on macOS.