Spatial resolution can reverse scientific conclusions in Earth observation: data and code

This repository accompanies the manuscript “Spatial resolution can reverse scientific conclusions in Earth observation”. It contains the processed analytical data, Source Data, evidence-group mapping, source-location and bibliographic records, analysis and figure-generation code, software/environment information, fixed random seeds, publication figures, reproducibility documentation, and file checksums.
The primary synthesis comprises 496 directly comparable ranking or sign contrasts from 20 papers and 19 independent analytical groups, including 35 conclusion reversals and 461 non-reversals. Nine of the 19 groups contained at least one reversal. Giving each independent group equal weight, the mean within-group reversal proportion was 10.16% (95% percentile interval from 10,000 whole-group bootstrap replicates, 3.68–18.20%). Six classification comparisons from three additional groups retain separate eligible-unit denominators and are not pooled with the ranking/sign evidence. A group-weighted resolution-ratio analysis of 484 contrasts from 18 groups did not support a monotonic association (odds ratio per doubling of linear support = 0.987; 95% interval, 0.621–2.002).
The archive supports the conclusion that substantive scientific interpretations can differ across spatial supports within matched Earth-observation evidence. It does not estimate a universal Earth-observation reversal prevalence, establish a resolution-only causal effect, identify a preferred resolution route, or define a universal resolution threshold.
Copyrighted source articles and publisher supplementary files are not redistributed; bibliographic identifiers and source locations are provided instead.