Data and code for: A narrow and narrowing divide - machine-learning analysis of geopolitical structure in four environmental science literatures

Corpus, derived matrices, analysis code, result tables and figures supporting the manuscript A narrow and narrowing divide: machine-learning analysis of geopolitical structure in four environmental science literatures (Cushman, Zhu, Hashimoto). The study analyses 1,202 Web of Science research articles (1990-2022) retrieved under four umbrella constructs - ecological integrity (217), ecological resilience (263), ecological sustainability (239) and ecological security (483) - and asks whether these literatures are lexically distinct once retrieval vocabulary is removed, whether their geopolitical structure reflects nation-specific writing or the sorting of communities into frameworks, and whether framework separation has widened or narrowed over time. The archive contains the source bibliographic records; the derived document, token and paper-by-word matrices; per-paper framework, year and nationality metadata; R and Python implementations of the tiered lexical ablation, random-forest model selection, bidirectional variance partitioning and multivariate trajectory analysis; all result tables reproduced in the manuscript and supplement; and the published and supplementary figures. A verification script (scripts/verify_archive.py) checks file integrity and re-derives the headline statistics from source. Known data issues are documented in README.md section 4 rather than silently corrected. Note on third-party content: files under data/raw/ include abstract text retrieved from the Web of Science Core Collection. Copyright in that text is retained by the original publishers and is not covered by the licence applied to this record.