20. Settlement Layouts
Everyone had a favorite picture. Domes like pearls. Hexagons to the horizon. Greenhouse cathedrals in tasteful rows. Solar farms glinting to the horizon.
I watched them parade through design reviews, each accompanied by evangelism and a napkin sketch. The problem with pictures is they do not leak, jam, drift off spec, or ask for maintenance bays. Mars did all four. So we asked a blunter question: not “what looks right?” but “what works?” Then we let the math mark our homework.
♦
First, the tropes, briefly and mercifully.
“The Giant Dome”: beautiful, brittle. One great pressure zone turns a pinhole into a city-wide problem. You can partition with décor; unless you can slam bulkheads, you are role-playing resilience. Sadly we hadn’t (still haven’t) cracked Banksian materials science that make them viable.
“The Camp of Misc Cans”: good for sorties, not for living. EVA to fetch a wrench is charming until it is daily. If you hard-connect everything later, you inherit a hedge of doors and joints that never sleeps.
“The Cruise Ship Megastructure”: a single vast building with clever interiors. Better than a dome; still hostage to shared failure modes, hard to expand without cutting into your own lungs.
CLS — the Complex Layout Score — dispatches these quickly with cold math. It grades the pressurized complex as a network of buildings (nodes) and streets (links), not the furniture inside. If you fail the gates — two independent routes to refuge, sane compartment sizes, dirty↔clean interlocks — you do not get a number; you get revision notes, penned with “Do better next time.”
Once the render-hazing cleared, three layout families kept turning up solid numbers when done well: spines, rings-and-radials, and meshes. Think of them as ways to arrange barrels (links) and barns (nodes) so shirtsleeve life scales without drama. None is religion. Each has its era on Mars.
♦
Spines (e.g. a main corridor with branches).
Strengths: the straightest path between “we landed yesterday” and “we can breathe inside.” Heavy-access is excellent — rovers, cranes, pallets all love a trunk. Junction variety stays low; SKU discipline is easy; phasing is almost insultingly simple: add another branch, repeat.
Struggles: chokepoints on the trunk, brittle egress until you close a loop, single-feed utilities unless you plan a counter-run. You will want two short cross-links just to stop the forklift queue from learning philosophy.
Bottom line: quick to stand up, easy to service. With a couple of cross-ties and a counter-utility run, a well-kept spine lands roughly in the low-to-mid-60s on CLS 100 index. But it wants to become a ring the way water wants downhill.
♦
Ring + Radials (e.g. one or two loops with spokes).
Strengths: redundant routes “baked in” — two ways to a haven without thinking; balanced flow; clean zoning (hot uses on spokes, quiet uses around the loop); expansion edges everywhere. Utilities love loops; so do sector bulkheads. Especially if you design them as double-two-level-loops. Sweet.
Struggles: inner areas need planned heavy-access (boulevard or bridged cuts), hubs can load up if you over-spoke, and your signage budget doubles if you let every spoke become a main street.
Bottom line: the natural mid-founding-era topology; tidy, resilient, expandable. Done with discipline, the ring family tends to sit mid-70s on CLS 100 Index and upgrades gracefully toward a mesh by adding regular cross-ties. (An Orbital would approve; Trantor might grumble about the lack of vaulted halls and wasted under-developed space.)
♦
Meshes (e.g. a rectilinear or hex lattice with sector lines and a service-alley lattice).
Strengths: shortest detours, graceful degradation when something fails, many small loops so a single cut rarely traps anyone. Expandability is trivial; recombining bays later is civilized work, not archaeology.
Struggles: temptation to “over-mesh” into spaghetti; junction count rises if you are not ruthless with T-junctions; heavy-access must be platted as alleys, not painted on a poster later.
Bottom line: the premium layout for large, long-life districts. It rewards planning and punishes flourish. Kept disciplined, planned and executed well, meshes score low-80s on CLS, and walk calmly into the 90s as utilities and sectoring go from good to smug. But to get these working at that level your design must be excellent, and your execution even better.
♦
If you prefer your conclusions condensed, the short TLDR version of our ‘crawl-walk-run’ layout approach was ‘spine-ring-mesh’:
Spines are fastest to build, easiest to service, and inherently brittle until looped.
Rings balance safety, flow, and growth; they are honest mid-scale founding layout design citizens.
Meshes are the end-game for resilience and operations — if you plan the service alleys as seriously as the streets.
CLS did not pick winners by taste. It priced egress time, path redundancy, link length density, envelope compactness, joint counts, service access, and expansion edges. It rewarded compact, looped, sectorable, serviceable, and expandable; it punished decorative complexity and anything that forced a human to fight a door twice an hour.
♦
A note on scale, because many early arguments ignored it. Below about 10,000 m² of connected, pressurized floor, you are still in “expedition with amenities” mode. CLS can be computed, but the noise dominates: two barns and three barrels do not a city make. Around 50k m², rings begin to make sense; by 500k m², meshes pay for their extra thought every day. The dials do not change; the bands do.
Another note, on physics in red dust. These layout families are not abstract exercises; they are what pressure, shielding, and maintenance ask of a civilization. Low and wide, buried and bulkheaded, with straight chords and gentle knees — that is not a style; it is what happens when the roof wants compression, the air wants to leave, and you want to swap a liner at 03:00 without waking a district.
♦
Engineers enjoy heresy, so here is one: everyone was right, and everyone was wrong. Well at least when it came to early twenty-first century Mars settlement designs.
The hexagon people were right about redundancy and wrong about joints. The dome people were right about compact envelopes and wrong about practical consequences. The grid people were right about operations and wrong about “more links always better.” CLS made those conversations short. It turned “I feel” into “it fails,” or, better, “it passes — if you add two cross-ties and stop insisting on four-way junctions.”
What does any of this have to do with barrels and barns? Everything. Spines, rings, and meshes are only the choreography; the steps are still those two mega-SKUs. A spine is a run of barrels with barns on the frontage; a ring is a closed run with spokes and barns appended; a mesh is straight chords joined by knees with barns where it matters. If you can keep the link kit small and the node kit honest, any of the three can be made to sing — within its proper season.
♦
I still remember the first time we closed a settlement loop on Mars. The forklift traffic went quiet like a lake in a pause of wind. Egress drills stopped being theatre. Utilities stopped caring which end the power came from. Someone taped a paper ring above the mess hatch and wrote “NO MORE DETOURS” on it. A childish act, but accurate.
So: why did we not skip the awkward adolescence and build meshes at dawn? Because the city you can operate beats the one you can imagine. Early yards could not move enough dirt to bury what a mature mesh needed. We had one battery swap depot, not three. We had six link SKUs, not sixteen. And minimal shims. A spine with two cross-ties was not just “good enough” — it was the only thing that did not fall over when the weather turned and the night ran long.
♦
But it is overdue we stopped speaking in shapes and started speaking in eras. Port Harmonia began as a line with ambitions, became a loop with opinions, and grew into a lattice with manners. The kit stayed small — barrels for streets, barns for rooms — while the choreography matured. The math nudged; the metrics barked; the dust taught fast. Here’s how the ground actually moved under our feet.
♦
The Set-Up Settlement Era (Pop. ~10s to low 100s; IL-0…2)
Spine-layout - stand it up, keep it reachable
Speed and heavy-access beat elegance. Things started when we laid a short spine from donor-shell barrels — four × 30 m staves per 120 m pressure cell — tipped from uncrewed ships, clamped with standard knees, liner pre-installed to infuriate purists and delight schedulers. We set the spine shallow or no-cut, held it down with ballast cassettes, and promised the berm later.
Two cross-ties were roughed-in on day one and sealed — future doors, not drawings. We distributed two havens on the trunk so every occupied node could reach one within a minute by two independent routes once the cross-ties opened. Bulkheads were placed where a tired human would actually close them; compartments stayed in the 800–1,200 m² band so a leak was local, not a legend.
Dirty stayed dirty: garages and fab shops sat behind an interlock so the shortest path to sleep or med crossed a filter, not a hope. Utilities ran one direction along the spine, but we reserved a counter-run and installed sectionalizer stubs every ~25–30 m so the second loop could snap in without re-plumbing the city.
Flow was honest, not heroic. Average trips were short; chokepoints existed and were tolerated; pressure-door crossings stayed low enough that people didn’t start propping them open with crates. Heavy-access was nearly perfect: cranes and rovers could nose up to almost any barn from the perimeter. We lived with more steel and anchors than elegance, because earthmoving was a rumor and power was a budget. It worked. It was survivable. It was not permanent.
What we counted: minutes to haven, joints per node, meters of link per 1,000 m², median drive distance to a door. What we changed: we opened one cross-tie as soon as it was safe; we sized three doors so they could convert to bulkheads later. What we refused to change: four-way junctions (entropy with signage) and systems hidden in walls (maintenance with swearing).
♦
The Start-up Settlement Era (Pop. ~100s to low 1,000s; IL-3…5)
Ring & Radials layout - close the loop, plan the streets.
The first ring was a civic emotion. We connected the two cross-ties and stopped pretending detours were a character-building exercise. Egress improved by arithmetic: two disjoint routes to haven baked in; sector bulkheads partitioned consequences into >1,500m² bites; a small inner loop around the busiest cluster (galley/med/admin) kept hub loads sane. We added a second utility loop running the other way and cross-tied the two; most nodes became dual-fed in a week, and the complaints about brownouts aged out.
Rings are honest because they balance things we used to argue about: they shorten detours without multiplying joints; they distribute commons without losing buffers; they let you grow in any direction without asking permission from geometry. We learned to put bulkheads on spokes, not every doorway; to keep pressure-door crossings per trip in a band that didn’t train sabotage by fatigue.
Heavy-access dipped when we encircled ourselves — predictable, fixable. We legalized a service boulevard: a driveable route that touched the inner side of the ring every 80–120m where a junction had an alcove. Bridges, not underpasses, whenever possible; the civil-works penalty for digging under pressure is paid in ulcers.
Meanwhile the hardware matured. Barrels grew up from cut-and-shut staves into haunched horseshoes: ~12m clear, ~9m crown, straight chords at a 6m frame pitch, knees passing hoop forces without inventing curves. Outside, regolith panels did the compression and radiation work; inside, the same liner family ran continuous past the joints. Cover thickened from 1–1.5m to 2–3m as the dirt-moving fleet got teeth. Barns learned long habits: 60m spans got ordinary, >120m runs got common, the mid-deck carried tie action so pressure ran with us, not against.
We kept reserves real inside the settlement core site: ~50% platted, with width, utility stubs, and rights-of-way touching each lot. Later, a clinic arrived on one; somewhere else a quiet agri-garden that saved a winter. People remember these more than they remember a new pump.
What we counted: OD times between sleep and mess; the worst link’s utilization at peak; “two-feed” coverage; distance to the nearest valve; percent of nodes a crane could still reach. What we changed: we added cross-ties at a regular cadence (about every 100–120 m), we set a haven at each significant loop intersection, and we codified the link/node SKU set (≤ six of each). What we refused to change: bespoke curves for beauty, and X-junctions except at a handful of high-value places. The ring earned its manners by saying no often.
♦
The Scale-up Settlement Era (Pop. ~1,000s to ~10,000; IL-5…7)
Mesh layout - lattice the city, degrade gracefully.
The mesh arrived the way spring arrives: by accumulation that suddenly looks like a season. Those cross-ties you sprinkled? Keep going — every two to three cells. The service boulevard you argued about? Duplicate it as an interior alley lattice so heavy kit can reach anything without apologies. Sector lines become civic lines; havens are as common as good exits in an old theatre. You do not eliminate chokepoints; you reduce their importance.
Operations changed character. Two workable routes for most important trips became normal. We kept path stretch (the penalty for not using the absolute shortest path) near 1.2–1.3 so the “second way” was a real option, not a sightseeing tour. Bulkhead crossings per trip stabilized around a civil number; most people could go about their day without fighting safety, and when the night asked for drills, everything clicked like well-worn tools.
Utilities matured into tri-loop where it mattered; critical nodes took three feeds on different chases; sectionalizers and by-passes landed at the same steady spacing the valve crews asked for a century earlier. We did not chase 100% perfection; we chased graceful degradation. Remove any single link and egress still worked; isolate any two adjacent utility segments and lights stayed boring in the med bay.
The kit stayed small on purpose. Straight 30 m chords with 7.5° knees, T-junctions bowed to sparing X’s, bulkhead door sets identical across links and nodes. Liners rolled from reels; gland cassettes swapped alive. Barns still re-fit on a decade cadence without touching their shells. Link-length density stayed in its sweet band so maintenance didn’t drown; the curved fraction stayed low so spares fit on shelves; bridges stayed more common than underpasses so no one had to cut the ground to deliver a compressor.
And we kept our manners: dirty stayed interlocked; noisy buffered quiet; commons sat near sleepers without mixing fumes with food. Meshes are not an excuse to forget zoning; they are a way to survive forgetting until you fix it.
What we counted: haven coverage and time under single-edge failure; link density and junction discipline (so the lattice didn’t become ivy); heavy-access coverage to interior nodes; outage footprint required to add the next district (kept small). What we changed: we legalized rolling renewal — one district offline for upgrade while the others carried the day—and we set aside replacement corridors so a future pipe run didn’t require repentance. What we refused to change: the temptation to over-mesh (straight lines are a virtue; excess links are a hobby), and the urge to co-route “independent” utilities in the same trench (malice by convenience).
♦
Peace with the past.
Spine logic never died; it moved to the edges. Starport ops still run on trunks with branches because heavy-access is king there and distance is short. Power yards and solar workshops do well on stout spines: straight runs, low SKU, fast service. In satellite towns, small rings anchor identity and resilience with minimal overhead; they grow by adding a spoke or closing a second loop, not by reinventing the grammar. And science outputs still look like landing zone scout-out EVA heavy camping.
Materials and ISRU did what economics always do: they bent the curves. When regolith panels gained strength and production, burial thickened and horseshoes dominated; when in-situ metals got cheap enough, we leaned more on ribs and less on imported plates. None of this changed the choreography much. Barrels became horseshoes; barns grew calmer with mass; the layout families stayed what they were because pressure, maintenance, and human legs stayed what they are.
And yes, domes did arrive… eventually. They were not crystal cathedrals. They were careful, ribbed, and buried to the shoulders; they sheltered parks the size of a small village green, not Earth’s national parks. People loved them anyway. Poetry is permitted when it does not break drills.
♦
The moral sits where the sawdust fell. We grew Port Harmonia by building the city we could operate, then teaching it better moves. We counted what mattered — minutes, meters, joints, doors, drives — and let the counts change the plan. We respected the kit: two mega-SKUs and a handful of helpers, iterated without apology. We refused to jump to “finished” because finished would have finished us.
You can still walk the old spine in the core and feel the loop close under your feet. Keep going and you find the ring and radials, and then how this lattices, quiet and indifferent to failures that would have ruined our sleep in the first decade. That indifference is not arrogance. It is civilization.
Next, we’ll stop talking about the inside of the settlement, and start to look at the wonderful world of working wedges that surrounded it.