cubebuild

cubebuild is the open-source construction pipeline of the Solid Earth Data Cube (SEDC) v1.0. It builds the cube; it is not the cube. The cube data are released separately — see Data.

The pipeline is driven by a single declarative layer contract (cubebuild/layer-mapping-v0.yaml) and produces a tiered Zarr v3 store (1°, 30′, 6′, 3′ and 30″ tiers), the vector sidecars, a machine-readable manifest (cube_manifest.json) and structural validation reports.

This repository contains the construction pipeline only:

• the cubebuild Python package, with the cubebuild command-line entry point;
• the layer contract that drives every build (cubebuild/layer-mapping-v0.yaml);
• the test suite;
• the input-dataset inventory (input-datasets.csv).

It does not contain the data cube itself or the raw input data.

Requires Python 3.11 or newer. Install from source:

git clone https://github.com/w-r-b/cubebuild.git
cd cubebuild
pip install -e .

To run the tests as well, install the test extra:

pip install -e ".[test]"

The full test suite requires the raw input data on a local disk and is therefore not covered by CI; CI runs the data-independent subset (tests/test_contract.py and tests/test_pixel_area.py).

The full cube cannot be rebuilt from this repository alone. Construction requires about 22 GB of raw input data that are not included here, and part of that raw data cannot be redistributed by the authors. Readers who want to use the cube should use the public version (see Data) instead of re-running the pipeline from scratch.

The public version of SEDC v1.0 is archived on Zenodo (DOI: 10.5281/zenodo.23098600) as 25 data parts (21 tar archives of the Zarr store split by tier and latitude / time bands, plus 4 binary parts of the consolidated sediment parquet) together with a catalog.json and a sha256 checksums.txt file. Reassembly commands are provided in the record. It contains 25 layers in the public Zarr v3 store and 4 vector sidecars. The full cube — 47 layers and 10 vector sidecars — is not publicly released: the redistribution licences of the remaining source datasets are unconfirmed or restrictive.

input-datasets.csv lists the input datasets with their sources, DOIs and licences.

Citation metadata for the software is provided in CITATION.cff. If you use the cube data, please cite the dataset release — see Data — as well as the original input datasets listed in input-datasets.csv.

The code and the data are licensed separately:

• Code (this repository): MIT — see LICENSE.
• Data (the public SEDC v1.0 release): CC BY 4.0.

Licences of the original input datasets differ per dataset; see input-datasets.csv.