Independent reproduction implementation for COMMSENV-26-2919: Environmental Governance of Artificial Intelligence Infrastructure across Territories, Sectors and Time

Abstract

Reproduction code for the statistics reported in the associated article on the environmental governance of artificial-intelligence and data-centre infrastructure. reproduce_gate3.py recomputes 247 prespecified output quantities from the final event-level data (Supplementary Tables S1, S3 and S5 and the 32 subnational search-frame jurisdictions of Supplementary Table S2) and compares each with the locked value in expected/EXPECTED_OUTPUTS.csv; a run reproduces all 247. The code is an independently written reproduction implementation; it is not the original analysis execution script, which was not preserved in the archived analysis package. Environment: Python 3.11.15, NumPy 2.4.4, SciPy 1.17.1 and pandas 3.0.2 (requirements.txt); root seed 20260919; 10,000 bootstrap and 50,000 permutation replicates; exact enumeration of 81 and 729 unresolved-case allocations. Run: python -m pip install -r requirements.txt; python reproduce_gate3.py. The run checks that the reported quantities follow computationally from the published data; it does not validate event selection, source interpretation or substantive classification, and the associations are descriptive, not causal. The input files in data/ reproduce content of the Supplementary Tables, which are published under CC BY 4.0 in the dataset record https://doi.org/10.5281/zenodo.22876740. This version supersedes the previous version of this record.