Sources and methods

• The Paleobiology DatabaseCC BY 4.0

  Every occurrence, date, classification and formation name. If something here says an animal lived in a place at a time, this is why.

• GPlates Web ServiceOpen, run by EarthByte

  The reconstructed coastlines, computed once per period and stored.

• Wikipedia and Wikimedia CommonsPer image, shown on the card

  The picture on a card, its licence and its author, and the description under the name.

• WikidataCC0

  Matches a taxon name to the right Wikipedia article, and supplies the article count behind the eye-catching ordering.

• OpenAlexCC0

  The papers on an opened card: a name search, most cited first, so a genus that is also an ordinary word can turn up another field's work.

• iDigBioPer record

  Photographs of actual catalogued specimens in museum collections.

• Smithsonian Open AccessPer image, shown on the card

  Specimen photographs, where one exists for the animal on the card.

• GBIFPer image, shown on the card

  Photographs from the collections that publish through it, used when nothing above has one.

• MorphoSource and SketchfabPer model, shown beside it

  The 3D scans, where a museum has published one of a specimen.

• MacrostratCC BY 4.0

  The group and subgroup a formation sits in, and the age stratigraphers give the rock itself rather than its fossils.

None of these are ours. Most run on donations or grants, so the app caches hard and paces its requests.

A fossil occurrence is a record. Almost everything built on one is a reading. Here is each reading, and where it stops being true.

• Found in North America

  Rests on The country a fossil was dug up in, grouped into eleven regions: seven modern continents, three landmasses that spent much of this site's span on their own (the Indian plate, Madagascar, Zealandia), and Open ocean, for drill cores in no country.

  Stops at A modern address, not an ancient one. A Triassic fossil from Morocco is filed under Africa, though Morocco was then joined to eastern North America. The maps show where the land was; the label does not.

• The coastlines on the maps

  Rests on The GOLONKA rotation model, the same one behind the Paleobiology Database's paleocoordinates, so shorelines and fossil dots describe one Earth.

  Stops at One model among several that disagree, drawn at a single date mid-period while continents move throughout. Read a shoreline as roughly where the sea was.

• The continental configurations on Tectonics

  Rests on Written by hand from published reconstructions, not a database field.

  Stops at Six named stretches with dates on them, and a reading rather than a boundary anyone agrees on. They name the period and order the list; they do not decide which region a search continues into, and the maps do not depend on them.

• Which regions were joined, on the schematic

  Rests on Ten landmasses read by hand off labelled reconstructions at thirteen dates, and interpolated between them.

  Stops at The connections are the claim; where any dot sits is illustrative, not a coordinate. It is what decides the region a search continues into when the chosen one runs out.

• Most eye-catching first

  Rests on The only opinion in the app: a coarse weighting over taxonomic class, nudged by how much has been written about the name.

  Stops at It says a ceratopsian is more worth looking at than a foraminiferan, which is a fact about people. It only reorders: nothing is hidden, no count changes.

• Likely alongside

  Rests on Taxa dug out of the same Paleobiology Database collection, which is one site and one horizon.

  Stops at Not proof of coexistence: a collection can span a bed metres thick and thousands of years. It is evidence of the same place and roughly the same time.

• The lighter cap on a diversity bar

  Rests on Taxa the database reports as implied rather than sampled: known before and after a period, so alive during it, but not found in it.

  Stops at Inferred from a range, not dug up, and drawn separately so the difference stays visible.

• Found on the Indian plate

  Rests on The six countries that sit on it: India, Pakistan, Nepal, Sri Lanka, Bangladesh and Bhutan.

  Stops at Whole countries are an approximation of a plate. The suture runs through northern Pakistan, where the Kohistan-Ladakh arc is Eurasian rather than Indian, and through Nepal, so this is still a line drawn on a map of today.

• Found in the open ocean

  Rests on The seven codes the Paleobiology Database uses for sites in no country, almost all of them deep-sea drill cores.

  Stops at A region for searching, not a place on the diagram. Its 2,953 animal occurrences are scattered across every ocean, so there is no one position to draw. The schematic marks it apart from the continents and joined to nothing; the map under the results plots each site where it actually was.

• Recorded as a living group

  Rests on The Paleobiology Database's own extant flag on the taxon.

  Stops at A database field, not a check, and it carries errors: it is set on Raeticodactylus, a Late Triassic pterosaur. Nothing here infers it from the youngest fossil, which is why the range chart's axis always ends in a date and never in “still living”.

• Found in this period only

  Rests on Taxa whose first and last recorded appearance fall in the same period, as the Paleobiology Database bins the record for the search on screen.

  Stops at A fact about the record, not the animal, and scoped to the search rather than the world: narrow the region and almost every group becomes single-period, because its other periods sit outside the search. On the Taxonomy chart it falls inside both arrivals and departures, so the overlap is drawn in its own shade rather than counted twice.

• A formation's rank

  Rests on Distinct taxa the search found at least one fossil of in that rock.

  Stops at It measures the rock and the fieldwork together: a formation ranks highly for being rich, for being reachable, and for a century of collecting. A taxon in two formations counts in both, so the column does not add up to the taxa in the search.

• The rock note under a formation

  Rests on Macrostrat's stratigraphic name authority, matched on an exact formation name whose rock reaches into the period on screen.

  Stops at It appears for roughly a quarter of the formations this site lists, and not evenly: nine in ten for North America in the Late Jurassic, one in ten for Europe, none for Asia, Africa, Antarctica, Madagascar or the Indian plate. Its absence means Macrostrat has not compiled that formation, never that the rock is unstudied. Where two different formations share a name and both reach into the period, nothing is shown rather than a guess.

• A last appearance

  Rests on The youngest occurrence on file.

  Stops at When something stops being found, not when it died out. The gap is why an extinction is normally given as a range.

• The dated events on Tectonics

  Rests on Compiled by hand from the literature, and the app's own selection of what is worth listing.

  Stops at Not exhaustive, and every date is a range because every event is: a rift takes millions of years, and several are argued over. India meeting Asia is dated anywhere from 60 to 35 million years ago.

• The dashed line through a scatter of points

  Rests on A power law fitted by least squares to the points on that chart, on logarithmic axes.

  Stops at A description of that cloud and nothing more. It corrects nothing and claims nothing about how many species existed; it is a line to judge a point against.

• The artwork on a card

  Rests on Whatever illustrates the taxon's Wikipedia article, credited to its author on the card.

  Stops at A life reconstruction is an artist's interpretation of a skeleton, and colour and soft tissue are almost never preserved. A photograph of a specimen is evidence; a painting of a living animal is an argument.

An empty result does not mean nothing lived there. It means nothing has been found, recorded and entered into a database. Fossilisation is rare and collecting is uneven; both favour ground that is dry, exposed, reachable and near a university. Whole periods and continents are thin here for reasons unrelated to what was alive.

It is why the Taxonomy chart draws the number of fossils under the number of kinds. The curves usually have the same shape, and that is the finding.

No analytics and no tracking. Nothing here needs an account. Two things are stored, in your browser and never sent anywhere: your settings, and anything you save. Clearing site data clears both, and neither follows you to another device. The Saved panel offers a download, so you can keep a copy.

One optional sign-in, with Patreon. It is the only thing here that sets a cookie or learns a name, nothing is behind it, and the privacy page says exactly what it stores.

Searches are answered by this server, which caches results so the next person asking the same thing does not cost the databases a second request. What is cached is the answer, keyed by the question, with nothing identifying you.

The rate limiter holds your network address in memory for a few minutes to count requests. It is not written to disk, not logged, and not tied to anything you looked at.

The endpoints the pages call return JSON. A script gets the same answers a browser does, and the same rate limit.

GET /api/explore?region=North America&interval=Late Cretaceous&view=index
GET /api/diversity?group=Ammonoidea&rank=genera
GET /api/taxon-profile?name=Triceratops

A refused request returns HTTP 429 with a Retry-After header; honour it. Every response carries X-RateLimit-Remaining, so a script can pace itself.

For the whole database rather than slices, the Paleobiology Database publishes bulk downloads.

EpochExplorer.Earth@gmail.com

Tell us how you are using Epoch Explorer. Feedback is welcome.

Corrections especially: something wrong, something that reads as more certain than the record supports, or a picture that should not be here. If a photograph is yours and the credit is wrong or you want it removed, say so and it will be. One person reads this, so a reply may take a few days.

Any result set downloads as CSV, and anything with coordinates as GeoJSON, for QGIS and similar tools. Look for Export beside the results, on the formations ranking, and under the diversity table.

Every file carries its credit and licence terms: a comment block in a CSV, fields in a GeoJSON. Attribution is a licence condition, not a courtesy, so leave those lines in and cite the database rather than this site in anything you publish.