Growing Continents, Not Fracturing Them
Two ways to build tectonic plates for a procedural world — and why the one we bet against won.
When you generate a world from tectonic plates, the very first decision — how do you make the plates? — quietly determines how natural everything downstream looks. Get it wrong and your continents read as Voronoi blobs with mountains painted on the seams. We spent a while on this, talked ourselves into the “obvious” approach, then tried the opposite out of curiosity. The opposite won. Here’s the story, and why.
Two philosophies
There are basically two ways to carve a planet into plates:
Top-down (fracture). Place a handful of big plates. Find their active margins — where plates converge, or where ocean meets continent — and shatter those margins into smaller child plates over a few passes. Big stable interiors, fragmented edges.
Bottom-up (grow). Place many small plates. Then agglomerate them: group neighbouring small plates into a few large “macro” plates that move as a unit. The macros are your tectonic plates; the small plates become provinces inside them.
Same end product — a hierarchy of large plates made of smaller pieces — built from opposite directions.
Why we bet on top-down first
Top-down felt obviously correct, for three reasons:
- It guarantees a backbone. Placing N big plates with spacing rejection gives you N comparably-massive continents from the start. No risk of ending up with a planet of gravel.
- The fracturing is tectonically motivated. Margins shatter; interiors stay whole — exactly the structure real continents have: ancient stable cratonic cores ringed by younger mobile belts. You get it for free.
- It’s a shortcut. Real plates assemble over billions of years; we’re generating a snapshot. Fracturing a few majors is a cheap way to reach the end-state structure without simulating the history.
All three are true. We shipped it. It looked fine.
The geology had other ideas
Then we read the microplate-tectonics literature — Li et al.’s “microplate paradigm” — and it’s blunt about one thing: real plates don’t fracture down from a few big ones. They amalgamate up from many small ones. Earth’s surface is a mosaic of ~1,000 microplates that accrete and weld into megaplates and cratons over geologic time. Cratons aren’t primordial blocks; they’re frozen collages of old microplates, welded along sutures that stay mechanically weak forever (which is why supercontinents later rift along those old seams).
So our “obvious” top-down approach was running the movie backwards. It reached a plausible end-state, but by the opposite mechanism from the real one. That nagged enough to try the real direction.
The experiment
Bottom-up turned out to be a small change. We already grow plate regions with a noise-warped flood-fill (organic lobes instead of straight Voronoi walls), so:
- Place ~50 small plates instead of 6 big ones.
- Pick a handful as macro nuclei.
- Every other small plate joins its nearest nucleus and inherits that macro’s rigid motion.
- The macro is now the moving plate; the small plates are provinces — and the boundaries between provinces of the same macro classify as sutures/faults: fossil seams.
Everything downstream — boundary classification, orogeny, erosion, the “old cores are hard cratons” pass — reads the exact same data it always did. We just built the hierarchy from the bottom.
It looked much better. Here’s why.
The difference was not subtle. Three things drove it.
1. Emergent size variety. Top-down spacing-rejection makes your big plates come out uniform — six continents, all roughly the same size. Real continents range from Eurasia to Madagascar. Bottom-up gets this for free: some macros greedily agglomerate many provinces and become huge, others stay small. The size distribution emerges instead of being imposed.
2. Internal structure everywhere. A top-down major has a smooth, featureless interior — fracturing only ever touched its edges. A bottom-up macro is a mosaic of provinces all the way through, and every province boundary is a fossil suture. Run erosion and those sutures carve into lineament valleys threading the whole continent — exactly the look real shields have (think of the fault-controlled valley grids in the Canadian or Australian interiors). Top-down simply can’t produce that; the interior never had any seams to exploit.
3. Organic outlines. A macro’s coastline is the union of its provinces’ organic boundaries — a naturally irregular, assembled shape — rather than one big blob with a wiggled edge.
And the kicker: it’s also the physically correct mechanism. We stopped fighting the geology and the terrain got better. That’s usually a sign you’re doing it right.
The honest trade-offs
Bottom-up isn’t free:
- It’s a touch more expensive — you assign more small regions before grouping them.
- You give up the guaranteed backbone. In principle a macro could agglomerate too little and come out a runt; in practice, with enough small plates it’s fine, but it’s a knob to watch.
- Size variety is emergent, which means less direct control. If you want a specific, art-directed continent layout, top-down’s explicit placement is easier to steer.
For a procedural generator chasing natural-looking variety, those are good trades. For a hand-authored map, maybe not.
The takeaway
The lesson generalises past tectonics: when you’re choosing between imposing structure top-down and growing it bottom-up, and the real-world process is bottom-up, bet on bottom-up. Emergent structure tends to beat imposed structure, and matching the physical mechanism — even loosely — buys you realism you’d otherwise have to fake by hand.
We spent real effort justifying the top-down shortcut, and it was a reasonable bet. But the planet looked more like a planet the moment we grew the continents instead of breaking them.
Both modes still ship behind a flag in our generator, so you can A/B them on the same seed. If you try this in your own project: place many small plates, group them by nearest nucleus, let provinces ride their macro’s motion, and treat the intra-macro boundaries as old sutures. Then turn on erosion and watch the interiors come alive.