Version log

Work in progress

• Sign in with ORCID. Sign-in and a prompted, optional import of a researcher's ORCID works are built; activation waits on client registration with ORCID.
• Study sites of member papers. A paper added by a member is read for the place the work was done, and that place is resolved through the same gazetteer and the same checks as the atlas study sites. Until a place resolves, the coordinate shown is labelled an estimate from the abstract rather than a study site.
• Tags on member papers. A paper added by a member is tagged by the same expert panel, the same codebook and the same checks as an atlas paper: every tag must be a codebook value, must quote the words in the title or abstract that support it, and must be confirmed by an independent judge. How this works, and how its accuracy is measured, is described in How member papers are tagged. No accuracy figure for member papers is published yet; it will carry its sample size and date when it is. Papers tagged before this keep their existing tags, marked as unmeasured rather than removed.
• Global Biodiversity Framework targets. A validated crosswalk from codebook tags to GBF targets 2, 4, 6 and 9 exists; a searchable target facet on the explorers is planned.

Released

20 October 2026

• Rubiaceae research atlas v1.0 published, the largest corpus in the series: 14,400 accepted species on the Kew WCVP version 16 census, 41,317 indexed studies, 39,980 of them tagged with 640,292 tags. 10,803 species (75.0 percent) have no indexed paper. The genus Coffea holds 14,298 studies, 34.6 percent of the record, and 74 of its 134 accepted species have no indexed paper; the family’s largest genus, Psychotria, holds 2.0 percent of the research across 1,645 species, 1,385 of them with none. 1,477 species assessed as threatened have no indexed paper, 290 of them Critically Endangered. 10,228 studies (24.8 percent) carry a geolocated study site across 170 countries.
• Blind precision on the finished corpus 89.0 percent of 200 sampled tags (95 percent interval 83.9 to 92.6), the highest a finished corpus has measured on this platform; read separately, coffee 87.9 percent and the rest of the family 89.6 percent. Link precision 85 percent of 179 sampled links. The accuracy figure covers the abstract-bearing corpus only, leaving 7.9 percent of tags outside the instrument, and 0.93 percent of tags are refusal pass-throughs; both are stated on the page.
• Release post, home page carousel entry, case-studies card, interactive map entry and atlas picker group for the new atlas. Figures audit extended to the Rubiaceae page: 2,380 checks, zero drift.

16 October 2026

• Report 1 of the series Research coverage of the world’s vascular plants, 2026 to 2030, titled Research coverage of three plant families under the Global Biodiversity Framework, published at version 1.0 as a page and a PDF. It covers 39,052 accepted species of Orchidaceae, Apocynaceae and Lecythidaceae, 10.7 percent of the Kew WCVP version 16 checklist, and the 63,782 indexed papers linked to them: 29,018 species (74.3 percent) have no indexed paper, 1,082 of 1,574 threatened species (68.7 percent) have none, 3,375 species (8.6 percent) carry any Red List assessment, and 14 papers name an instrument of the Convention on Biological Diversity. Country coverage and author institutions are tabulated for 231 countries. Prepared for CBD COP17, Yerevan, 19 to 30 October 2026.

14 October 2026

• Felidae atlas v1.1: species range countries now from the Mammal Diversity Database version 2.5 country distribution, confirmed countries only (688 country records for all 45 species), in place of GBIF occurrence countries of every basis (997 records for 37 species), which counted zoo, museum and escaped-animal records as presence. Species with the United States as range country 26 to 5; range countries 161 to 151. The range-shortfall table is now led by Chad, Brunei, Guinea and Guinea-Bissau, where Germany, Denmark and Japan had led it; all 45 species have a study site inside their native range. Study counts, tags and study sites are unchanged. French Guiana and other overseas territories are now named as such in the generated country tables, not under their sovereign state.
• Funding and opportunities: filter by career stage (undergraduates, master's students, PhD candidates, postdocs, early, mid and senior career). Every listing was re-read against its own text and the funder's page; a stage is recorded only where eligibility is limited to it, each supported by a sentence from that text. Choosing a stage shows the listings limited to it, listed first, together with those open at any stage or stating no limit.
• Antarctic atlas: the species range countries (20,743 range records for 1,191 species) and the study-site corrections recorded on 28 September 2026 are now carried in the engine database, so the spatial layer republishes exactly as published: 3,967 of 5,550 studies with a located study site, 6,978 placements. Range coverage recomputed on the restored records: 900 of 1,191 species with range data have a study site inside their range (895 before), 86 have sites only outside it (89 before).
• The 28 September corrections (sub-national phrases resolved to a country point, points outside their own country polygon, concatenated legacy sites, northern-hemisphere sites on Antarctic-only studies) had been applied to the published atlases but not to the engine database, and the October family republishes had re-trusted 730 of those placements. They are withdrawn again on every atlas. Studies with a located study site: Felidae 9,084 to 8,898; Orchidaceae 6,770 to 6,677; Apocynaceae 4,679 to 4,625 across 148 countries (150 before); Brazil-nut family 936 to 928; fungi 604 to 599; desertification 15,838 to 15,766. Report 1 table 3: Brazil located studies 1,806 to 1,700, Madagascar 113 to 109, United States 532 to 527.

13 October 2026

• Study-site places re-keyed to GeoNames identifiers across the Felidae, Orchidaceae, Apocynaceae, Brazil-nut family and fungi atlases. 9,538 name-keyed places became 7,455 identifier-keyed places; 26 names that had merged sites in different countries under one point (Sahel, South Africa, Ontario, Rocky Mountains, South Florida, Black Hills, Sonoran Desert, Kalahari, Amur, Kimberley among them) are now separate places, and 930 of 36,089 site placements moved to the correct point. Species-range checks downgraded 805 placements that fell outside the known range of every species in the study.
• Studies with a located study site after the correction: Felidae 9,220 to 9,084 across 149 countries (170 before); Orchidaceae 7,081 to 6,770 across 150 countries (163 before); Apocynaceae 4,681 to 4,679; Brazil-nut family 952 to 936; fungi 592 to 604. Twenty-one Felidae studies whose sites had been lost in an earlier run recovered 19 placements.
• Desertification atlas: all 20,635 study-site placements on 15,239 studies re-judged against their abstracts; 2,845 placements withdrawn as range statements, passing mentions or affiliations, 2,458 kept after a second review, and 5,649 re-resolved to a more precise point. Studies with a located study site 16,055 to 15,838 (57.1 to 56.3 percent of the 28,111-study core); distinct precise places 7,400 to 7,746. The Sahel province of Burkina Faso now carries 109 located studies (334 before, when the Sahel region as a whole shared its point).
• Spatial-unit assignments completed for every located place; the expected-research sections now count 8,164 Felidae and 5,707 Orchidaceae unit-studies over ecoregions (4,567 and 2,461 before, when places added in the October republishes had no unit assignment). Spatial, change and expected sections, case panels, map data, Report 1 country table and home-page totals regenerated; figures audit 2,159 checks, zero drift on the regenerated pages.

12 October 2026

• Felidae research atlas v1.0 published: 45 accepted species on the Mammal Diversity Database version 2.5 census, 18,019 indexed studies, 17,027 of them tagged with 280,437 tags. No accepted species is without indexed research; five species hold 8,853 studies, 49.1 percent of the record, and the genus Panthera holds 41.4 percent. Median studies per species 398 for Least Concern against 59.5 for Endangered. 9,220 studies (51.2 percent) carry a geolocated study site across 170 countries. Whole-corpus blind precision 79.4 percent of 199 sampled tags; link precision 81 percent of 233 sampled links.
• A judged common-name pass over 41 English names added 1,047 links on 1,039 studies, 145 of them for the Iberian lynx, 29.5 percent of its record. Three names and eight homonym-dominated names remain unsearched.
• Release post, home page carousel entry, case-studies card and interactive map entry for the new atlas. Figures audit extended to the Felidae page, post and cards: 2,167 checks, zero drift.

10 October 2026

• Desertification atlas v1.1. The 28,111-paper core corpus was re-tagged under the current codebook with the gated pipeline. Whole-corpus blind precision 69.0 percent (138 of 200 sampled tags, 95 percent interval 62.3 to 75.0); 620,794 tags on 28,105 papers, against 680,874 before. The measurement standard tightened on 5 October 2026 when the domain rubric was clarified: the previous tags measure 70.0 percent on the same standard, and the September figure of 72.5 percent came from 40 tags. Corpus and study sites unchanged.
• Orchidaceae atlas v1.1. Seven large genera were re-harvested after a result cap in the earlier harvest: Dendrobium, Vanilla, Gastrodia, Phalaenopsis, Cypripedium, Anoectochilus and Bletilla, 2,077 species. The corpus is 26,501 indexed studies, 578 of them new records, with 24,888 tagged and 374,719 tags. Species with no indexed paper 24,503 of 31,892 accepted (76.8 percent, from 24,512 and 76.9 percent). Blind precision on the new papers’ tags 79.8 percent of 119 judged tags; link precision re-measured at 84 percent of 180 sampled links.
• Report 1 figures revised on the same corpus: three-family papers 63,782, species with no indexed paper 29,018 of 39,052.

4 October 2026

• Antarctic atlas v1.1 and Cordycipitaceae atlas v1.1. Both corpora were re-tagged under the current codebook with the gated pipeline. Whole-corpus blind precision: Antarctic 79.0 percent (158 of 200 sampled tags; before 72.5), 100,265 tags on 5,541 papers; fungi 89.4 percent (169 of 189; before 78.9), 127,363 tags on 7,782 papers. Species and paper counts unchanged.
• Home page totals recounted over every classified corpus; this version log published.

3 October 2026

• Apocynaceae atlas v1.2 on the Kew WCVP version 16 census: 6,759 accepted species, 4,292 with no indexed paper. Range countries now from WCVP distributions instead of GBIF occurrences. Species link precision measured and stated on the page.
• Place Explorer: collapsible side panels, every study site of a unit plotted, a taxonomic group picker, readable popups, faster unit queries.

2 October 2026

• Orchidaceae research atlas v1.0 published: 31,892 accepted species, 76.9 percent with no indexed paper, 25,919 papers, blind precision 85.5 percent. Release post, home page feature and case-study card.
• Brazil-nut family atlas v1.2 on the WCVP version 16 census: 401 accepted species, corpus 2,938 to 3,588 papers.
• Figures audit extended to every number on the new pages (1,005 checks at release, zero drift).

30 September 2026

• Spatial sections on every atlas: geographic distribution of study sites, research against environmental change, and an expected-research model (beta) with prediction intervals.
• Census provenance became a publish-gate check; atlas census counts default to accepted species.
• Tagging pipeline: reproducible sampling for accuracy checks, and each paper is attempted once per codebook version.

28 and 29 September 2026

• Classification accuracy is now measured by a blind re-judge of a random sample of published tags, and each atlas states the result with its sample size and interval. Earlier pages had cited figures from small reference sets; they were replaced with whole-corpus samples (79.8 percent across published corpora at the time).

August 2026

• Desertification atlas published from the 28,111-paper core of the land-degradation corpus.
• Researcher profiles, dashboard and follow system; four-domain geocoding hardening sweep.

Each atlas carries its own dated revision note with the before and after figures for every change to its data. Measured accuracy figures on this site are blind re-judge results, with the sample size and 95 percent interval stated.