Rock moisture as a globally important yet overlooked driver of terrestrial water-carbon coupling

Rock moisture as a globally important yet overlooked driver of terrestrial water-carbon coupling This archive contains the analysis/figure code and the project-generated data needed to reproduce the figures and tables of the study. It is a self-contained, pruned mirror of the working repository: unpack it ('tar -xzf bedrock-reproduction.tar.gz') and the relative paths resolve as-is. Contents scripts/ - all code, organized in four layers plus shared tools: preprocessing/ - public source datasets to standardized products on the 0.05-degree grid calculations/ - standardized products to derived diagnostics (water balance, vegetation activity, carbon, random forest, trend, CoLM) stats/ - diagnostics to figure tables, supplementary tables, and in-text numbers figures/ - derived data and stats tables to the manuscript figures tools/ - shared utilities, plotting styles, and grid definitions data/ - project-generated data the figures read: derived/ - diagnostics organized per method stats/ - figure CSV/NetCDF tables and in-text numbers external/, processed/ - public inputs re-gridded to the project grid that figures read directly README.md - per-figure reproduction recipe (which scripts to run, in order) MANIFEST.md - path, size, and SHA-256 of every file LICENSE-CODE.txt, LICENSE-DATA.txt Reproducing the figures Requires Python 3.11 with numpy, xarray, netCDF4, pandas, scikit-learn, shap, matplotlib, geopandas, and the cdo command-line tool. From the unpacked root: export PYTHONPATH=$PWD python scripts/figures/fig_sbedrock_overview.py Each script locates the archive root from its own path, so the tree runs wherever it is unpacked. The README maps every figure and table to the scripts and data needed to regenerate it. All manuscript figures regenerate from the data included here. (One figure, fig_dtb_multiproduct, also composites a hand-drawn concept panel that is not part of the data pipeline and is not distributed; its data panels reproduce normally.) What is not included, and why - Public source datasets (precipitation, evapotranspiration, GPP, depth to bedrock, CoLM forcing, etc.) are not redistributed; obtain them from their original providers as listed in the manuscript Data Availability statement. The preprocessing scripts document the re-gridding recipe but cannot be re-run without these sources. - Large regenerable intermediates (full water-balance state stacks, root-zone vegetation-activity stacks, the multi-member precipitation-by-evapotranspiration combinations) and the trained random-forest model files are omitted to keep the archive compact; the products and statistics derived from them are included. Key project-generated datasets - Weathered-bedrock plant-available water capacity (S_bedrock), 0.05-degree, global - CoLM control/experiment dry-season diagnostic, 0.25-degree - Compiled field-site database License Code is released under the MIT License ('LICENSE-CODE.txt'); data are released under CC BY-NC 4.0 ('LICENSE-DATA.txt').