Paleontology Analytics

Uncovering the hidden patterns of life's history

How “separated” were ecosystems long ago?

Quick summary:

  • We split the map into squares, then check which squares share the same genera.
  • High score = regions look different from each other (more “separate”).
  • Low score = regions look similar (more “mixed together”).

What this score really is: a network “clumpiness” number (modularity) built from locality ↔ genus links. It uses paleocoordinates when available (otherwise modern coordinates), with 5°×5° bins. It’s a helpful proxy, but it can also be influenced by sampling and climate. Modularity also has known limitations (including a “resolution limit”), so treat it as a trend/indicator, not a ground‑truth measure of biogeographic provinces.

Is this just random?

Quick summary:

  • We keep how many (unique) genera each map square has…
  • …and how many squares each genus appears in…
  • …then shuffle genus labels across square ↔ genus links.
  • If the real score is much higher than the shuffled scores, the pattern is likely real.

Where are fossils on the ancient Earth?

Quick summary:

  • Pick a time. We show a sample of fossil locations on a world map.
  • Most points use paleocoordinates (where the rocks were back then).
  • Use Play to watch the points move through time.

When did lots of life disappear?

Quick summary:

  • A spike in the purple line means many genera are present in one bin but missing in the next (younger) bin.
  • We mark a few famous events (like the dinosaur-ending one at ~66 Ma).
  • This is “what the database shows”, not a perfect measure of true global extinction.

Before vs After (examples)

Quick summary: For each event, we compare a “before window” to an “after window” and show example genera.

Can a model guess who disappears next?

Quick summary:

  • We look at one time bin, then ask: “does this genus show up in the next bin?”
  • We feed the model simple clues: how widespread it is, how often it’s found, and its latitude spread.
  • The scores below are from a held‑out split with no genus overlap (more honest than random row splits).
  • This is still a rough, “for learning” model — not a causal claim or a conservation tool.

Do warmer times have more genera?

Quick summary:

  • Blue line: how many genera show up in each time bin.
  • Orange line: a simple “how warm Earth was” curve.
  • We measure how much they move together: r = ….

Big warning: this is not proof that temperature causes diversity changes. Extinctions, rock availability, and sampling can also change what we see.

What kinds of life are in the dataset?

Quick summary:

  • This chart is a “category wheel”: phylum → class → order.
  • Bigger slices mean more records in this dataset (not “most important in nature”).

🏆 Survivor Champions

Quick summary: These genera show up across the biggest time span in this dataset.

Important context: some long‑lived “genera” here are trace‑fossil ichnogenera (behavioral labels), which can persist across many unrelated lineages — so don’t read this as “the same organism survived for 400+ Myr.”

🦖 The Age of Dinosaurs

Quick summary: A dino-focused filter, then “how many dino genera per time bin”.

Note: counts reflect this dataset (and may include synonyms, uncertain assignments, and some trace fossils), not a curated “official” list of dinosaur genera.

🔍 Explore the Data

Quick summary:

  • One row per genus (not one row per fossil).
  • Age range = oldest to youngest finds in this dataset.
  • Occurrences = how many records mention that genus.
Genus Age Range (Ma) Occurrences Sample Reference
Page 1

🚀 Future Analysis Ideas