Abstract
Natural archives record environmental change over timescales that instrumental series cannot reach, and global environmental assessments rest on the evidence they supply. To our knowledge, however, where that evidence is produced, and whether it reflects the processes it describes, has not been established. Here we reconstruct the sampling and publication geography of 33,145 studies of past environments, tracing each to the Sustainable Development Goals it addresses. Sampling and publication are not decoupled but nearly isomorphic. Twenty-six countries account for 89% of publication credit and 81% of sampling, and all eleven goals with evidence flows draw 82–90% of their evidence from them. Those countries hold 57.5% of world population and produce that evidence roughly 28 times more intensively per head. National income acts on research output as a threshold rather than a gradient. Below about 41,000 PPP dollars per head additional income buys no additional output, while above it output rises steeply with income. Because credit and sampling are read from affiliations and abstracts, the analysis captures where research is led rather than the provenance of every sample. The evidence base for global goals therefore reflects the research geography of a minority rather than the distribution of the processes being measured. Our findings provide the first baseline against which commitments to widen participation can be measured, and identify analytical capacity, rather than access to samples, as the binding constraint on whose evidence the planet is studied.