Data and code for "Has algorithmic innovation improved reported satellite rice-mapping accuracy? Evidence from 189 rice studies"

This dataset and code archive supports the revised manuscript “Has algorithmic innovation improved reported satellite rice-mapping accuracy? Evidence from 189 rice studies”. The release contains 437 assessment records: 251 retained studies (189 primary rice-mapping studies and 62 secondary crop-classification studies) and 186 excluded records. It provides study-level extraction data, eligibility and coding documentation, source-linked decision records, a separate screening sample, reproducible Python scripts, analytical results, sensitivity analyses and figures. The analysis distinguishes between-study associations from within-paper comparisons evaluated on shared test data. The primary rice shared-test analysis includes 15 papers and 68 dependent method pairs. Original figure layouts and colour palettes are retained. The authors report manual Google Scholar, IEEE Xplore and arXiv searching during April–May 2026 and completion of subsequent Scopus/Crossref and Web of Science searching on 30 June 2026. Individual query histories and complete upstream screening counts remain unavailable. Stream B’s accounting is reconciled as 831 − 1 text-extraction failure − 545 without extracted OA − 148 duplicate-filter removals = 137; the earlier figure label omitted the separately logged extraction failure. The separate 500-record September screening sample includes 37 records requiring full-text assessment; it is not evidence of complete historical search coverage. Licensing: the authors’ original data compilation, documentation, results and generated figures are licensed under CC BY 4.0; their original scripts are licensed under MIT. These licences apply to distinct components, as specified in LICENSING.txt. Third-party excerpts, source publications, external data and software dependencies retain their applicable rights and notices; see NOTICE.txt. Please cite this specific Zenodo version and the underlying sources where relevant.