CCD — evolution and drivers of the global carbonate compensation depth over the past 170 million years

Code, derived results and figures for:

Dutkiewicz, A., and Müller, R.D., Evolution and drivers of the global carbonate compensation depth over the past 170 million years (submitted). EarthByte Group, School of Geosciences, The University of Sydney.

The repository reconstructs an area-weighted global carbonate compensation depth (CCD) from Atlantic, Pacific and Indian Ocean records, calibrates it against long-term sea level, extends it to 170 Ma, converts it into global carbonate sediment thickness, and drives a solid-Earth carbon-degassing model with the result. It contains everything needed to reproduce the analysis except the large input and output grids, which are archived on Zenodo (see Data availability).

---

Quick start

git clone https://github.com/EarthByte/CCD.git
cd CCD
pip install -r requirements.txt

python run_ccd_core.py          # steps 1–7: timescale normalisation → global CCD → 170 Ma hybrid

run_ccd_core.py is self-contained and deterministic: every input it needs is in data/, and it regenerates every step-1–7 output and diagnostic figure. It takes a few minutes on a laptop.

The downstream stages need a local geoscience stack and the Zenodo grids:

./run_carbon_cycle.sh           # steps 8–10: thickness grids → volumes → CO2 degassing

run_carbon_cycle.sh re-runs run_ccd_core.py first, so the carbon stages always start from the current CCD and nothing stale propagates. See pipeline_carbon/README.md for its options (--skip-ccd, --only-step9, CO2_START_FROM=… to resume a notebook sequence).

---

Workflow

Step	Folder	What it does
1	steps/step1_timescale_conversion	Normalise every CCD and sea-level curve to GTS2020, then resample onto a common grid
2	steps/step2_ocean_basin_areas	Reconstructed ocean-basin area fractions through time (notebook; result stored in data/ccds_and_oceanbasin_fractions.xlsx)
3	steps/step3_global_ccd_synthesis	Area-weighted global CCD synthesis and inter-basin dispersion, 0–52 Ma
4	steps/step4_sealevel_envelope	Hybrid Miller et al. (2024) / Haq et al. (1987) sea-level peak-following envelope
5	steps/step5_ccd_lowpass_filter	Butterworth low-pass of the global CCD to match the sea-level bandwidth
6	steps/step6_sealevel_ccd_regression	Reduced-major-axis sea-level–CCD calibration, prediction and diagnostics
7	steps/step7_planktogenic_hybrid_ccd	Planktogenic correction and splice → hybrid CCD, 0–170 Ma
8	steps/step8_carbonate_sediment_thickness	Compacted carbonate sediment thickness grids from the new CCD (min / mean / max sedimentation-rate scenarios)
9	steps/step9_carbonate_volume_analysis	Global carbonate volume and area statistics; DSDP/ODP ground-truth at 38 basement sites
10	steps/step10_carbon_cycle_degassing	Deep-Earth carbon-cycle notebooks: subducted carbon → degassing → atmospheric flux

Steps 1 and 3–7 are pure Python (numpy / scipy / pandas / matplotlib). Step 2 is a notebook. Steps 8–10 additionally need pygplates, gplately, GMT/pygmt and jupyter, plus the Zenodo grids.

Timescales

The published source curves sit on three different geomagnetic polarity timescales: GTS2020 (Ogg, 2020) for the Miller et al. (2024) sea level and the Dalvand et al. (2025) Pacific and Indian CCDs, GTS2012 (Ogg, 2012) for the Atlantic CCD and the Haq sea-level compilations, and Cande & Kent (1995) for the Pälike et al. (2012) record that extends the Pacific CCD from 35.5 to 52 Ma. Step 1 is the single place where these age models are reconciled. Everything downstream reads GTS2020 curves from steps/step1_timescale_conversion/outputs/. Only curves that enter the analysis are converted; published curves shown purely for comparison, including the Haq (1987) pair used to validate the sea-level envelope, keep their own age models. Segments already on GTS2020 pass through untouched, splices are re-imposed after conversion, and the converted series are resampled onto a 1 Myr grid for the CCDs and 0.1 Myr for sea level. See steps/step1_timescale_conversion/README.md.

Each step writes to its own outputs/ (or output/) folder. Those outputs are committed here so the results can be inspected without rerunning the workflow.

Notebooks are committed with their outputs cleared to keep the repository small. Their executed results are included as CSV under steps/step10_carbon_cycle_degassing/Alfonso_etal_2024_DM26/Outputs/Notebook*/csv/.

---

Figures

figure_scripts/ holds the scripts that build the published figures and the two supplementary animations; figures/ holds the rendered output (PDF + PNG), the hybrid CCD curve and the carbonate-budget table that panel 3D plots.

Figures 1, 2 and 4 are drawn with pyGMT at the journal's column widths, so the type sizes on the page are the sizes the reader sees; Figure 3 and the supplementary figures are matplotlib. All of them are set in Helvetica.

Figure	Script	Output
Fig. 1 — regional and area-weighted global CCD, 0–52 Ma	figure_scripts/make_fig1_gmt.py	figures/Fig1_regional_global_ccd_gmt.*
Fig. 2 — long-term sea level with the CCD calibration inset, and the hybrid CCD to 170 Ma	figure_scripts/make_fig2_gmt.py	figures/Fig2_combined_gmt.*
Fig. 3 — reconstructed carbonate thickness at 0, 34 and 115 Ma, with the carbonate budget	figure_scripts/make_fig3_maps.py (panel D via carbonate_budget.py)	figures/Fig3_carbonate_thickness_maps.*, figures/Fig3_carbonate_budget.csv
Fig. 4 — the CCD against the solid-Earth carbon fluxes, their correlations, and the variance each explains	figure_scripts/make_fig4_gmt.py	figures/Fig4_combined_gmt.*
Fig. S1 — regression-method sensitivity	steps/step6_sealevel_ccd_regression/regression_diagnostics.py	figures/FigS1_regression_method_sensitivity.*
Fig. S2 — modelled vs observed present-day thickness	steps/step9_carbonate_volume_analysis/08_ground_truth_carbonate_thickness.py	figures/FigS2_modelled_vs_observed_thickness.*
Fig. S3 — residual distribution at 38 sites	steps/step9_carbonate_volume_analysis/08_ground_truth_carbonate_thickness.py	figures/FigS3_thickness_residual_histogram.*
Fig. S4 — residuals by ocean basin	steps/step9_carbonate_volume_analysis/08_ground_truth_carbonate_thickness.py	figures/FigS4_thickness_residuals_by_basin.*
Fig. S5 — net-outflux end-member scenarios	steps/step10_carbon_cycle_degassing/standalone_plots	figures/FigS5_net_outflux_endmembers.*

figures/Fig1_regional_global_ccd.* is the same panel rendered by steps/step3_global_ccd_synthesis/global_ccd_synthesis.py as that step's own diagnostic.

figure_scripts/attribution_closure_analysis.py performs the attribution (correlations, AR1-adjusted significance, incremental R², first differences) and writes figure_scripts/attribution_matched_series.csv and figure_scripts/component_correlations.csv. Figure 4 reads both, so the panels and the numbers quoted in the text cannot drift apart.

The figure scripts carry their own layout checks — text-collision detection, minimum type size against the journal's floor, line weights and the typefaces embedded in the PDF — and report them when they run.

Animations

Movie	Script
Movie S1 — global paleobathymetry, 170–0 Ma	figure_scripts/make_paleobathymetry_video.py
Movie S2 — compacted carbonate sediment thickness, 170–0 Ma	figure_scripts/make_carbonate_thickness_video.py

Both share figure_scripts/_ccd_video_common.py and render 171 frames plus MP4s into figures/videos/ (git-ignored). The rendered movies are on Zenodo. Options:

python figure_scripts/make_carbonate_thickness_video.py --time-step 5   # quick preview
python figure_scripts/make_paleobathymetry_video.py --force             # re-render all frames

The figure and video scripts locate the workflow root and the figure folder automatically, so they run unchanged from this repository or from the authors' working tree.

---

Supplementary datasets

The three datasets cited by the supplement are in supplementary/:

Dataset	File	Contents
S1	supplementary/Supplementary_Dataset_S1_CCD.xlsx	the global and regional CCD curves and the reconstructed ocean-basin area fractions
S2	supplementary/Supplementary_Dataset_S2_carbon_model.xlsx	the deep-Earth carbon-cycle model results: plate influx, atmospheric influx by source, and net outflux
S3	supplementary/Supplementary_Dataset_S3_carbonate_budget.csv	the carbonate budget plotted in Figure 3D — net volume gain and seafloor area above the CCD

S1 and S2 are rebuilt from the current workflow outputs by figure_scripts/make_supplementary_datasets.py; every sheet records the file its numbers came from, and each workbook carries a provenance sheet that marks any source predating the current CCD. S3 is written by carbonate_budget.py at the same time as the figure panel, from the same arrays.

---

Data availability

Small inputs — the regional CCD curves, the sea-level records, the reference CCDs of Boss & Wilkinson (1991) and Delaney & Boyle (1988), the ocean-basin area fractions and the present-day carbonate-thickness validation grids — are in data/ and are all that run_ccd_core.py needs.

The large grids are archived on Zenodo:

Zenodo archive: https://doi.org/10.5281/zenodo.22785465

It contains:

• Carbonate sediment thickness grids, 170–0 Ma at 1 Myr, as three separate scenarios — minimum, central and maximum carbonate sedimentation rate — each carrying the compacted thickness, the decompacted thickness and the deposition mask at every time step
• Paleobathymetry grids, 170–0 Ma at 1 Myr, which drive the thickness model, computed on the Alfonso et al. (2024) plate model following the pyBacktrack 1.5 method of Müller et al. (2026)
• Reconstructed continental masks, matching those times
• the plate model the reconstructions run on
• the present-day carbonate thickness grids used for validation (also in data/)
• the two supplementary animations, each rendered forward and backward in time

All grids are global NetCDF at 0.25°, in reconstructed (paleo) coordinates. zenodo/README_zenodo.md is the archive's own README and describes every file, what the three scenarios mean and how to use them; zenodo/make_zenodo_archive.sh builds the archive from a working tree and writes a checksummed manifest.

Download the archive and place its contents at the paths given in pipeline_carbon/config.sh before running steps 8–10. The step-10 notebooks read the same grids through environment variables, so nothing in them is tied to one machine:

Variable	Points at	Default
CCD_SOURCE_DATA	the unpacked grids, holding CarbonateThickness/ and Paleobathymetry/	source_data
CCD_PLATE_MODEL_ROOT	the folder containing Alfonso_etal_2024_modClennettMuller/	.
CCD_MASK_ROOT	the continental-mask collections	continent_masks
CCD_SUBDUCTED_WATER	the Cao et al. (2024) subducted-water grids, a separate dataset	subducted_water
seafloor-age, spreading-rate, sediment-thickness, crustal-carbon and reservoir
grids that step 10 also reads are not redistributed; their sources are given in
steps/step10_carbon_cycle_degassing/.

---

Citation

If you use this code or the derived CCD record, please cite the paper, this repository and the grid archive:

What	DOI
This repository, all versions	10.5281/zenodo.22792794
Grids and animations	10.5281/zenodo.22785465

Cite the all-versions DOI unless you need to pin a particular release; it always resolves to the current one. CITATION.cff carries the machine-readable form.

Two pieces of Zenodo metadata live here and describe different records. .zenodo.json at the repository root is what Zenodo reads when it archives a GitHub release of this repository, giving that release its own DOI. zenodo/zenodo_metadata.json describes the separate archive of grids and animations, which is too large to live in a repository.

Licence

Code is released under the MIT Licence (LICENSE). Figures, derived data tables and the reconstructed CCD record are released under CC BY 4.0 (LICENSE-FIGURES). Third-party inputs retain their original licences; the carbonate-sediment-thickness workflow in steps/step8_carbonate_sediment_thickness/ carries its own upstream LICENSE.