Living and working elsewhere: from landing site to society
Living elsewhere connects site, energy, water, transport, construction, maintenance, industry, health, and skills. Every choice opens and closes options.
This module assembles earlier tools and compares architectures at 4, 20, 100, and 1,000 inhabitants without prescribing one Mars city.
1. Choose a site: science, resources, operations
NASA used the Exploration Zone concept: regions of interest roughly within 100 km of a central landing site so science, resources, and operations could be considered together. The radius is a historical study framework, not a universal settlement rule.
Site choice combines altitude for EDL, slope, rocks, accessible ice or minerals, energy, dust, communications, geologic hazards, and science value. An excellent resource in a cargo-hostile location can be less useful than a modest resource inside a robust architecture.
Site study — turn a radius into an operations problem
NASA’s historical Exploration Zone concept grouped regions of interest within roughly 100 km of a central landing site. A 100 km-radius circle has geometric area πr² ≈ 31,400 km², but that area is obviously not all usable: slopes, rocks, terrain hazards, and scientific targets fragment it. The calculation only exposes the spatial scale of the mobility problem.
A round trip to a point 80 km from the centre is at least 160 km of ideal travel before detours. At a teaching average operational speed of 10 km/h, that is 16 hours of driving alone. Add sampling, stops, charging, weather, and rescue margin and “a resource 80 km away” becomes a mission rather than a dot on a map. Site selection must therefore connect resource value to actual reachability.
System intuition
Choosing a site solves several problems at once: landing safety, water access, energy, science, traversable terrain, communications, and room to expand. NASA’s Exploration Zone concept places several regions of interest within roughly 100 km, but that geometry does not guarantee operational access.
Criterion weights change the decision; a score is useful only when its assumptions are explicit.
Site case — multiple criteria
Compare two sites: A has closer water but harder terrain; B requires 20 km more travel to water but offers easier landing and routes. A trade matrix should translate each advantage into measurable consequences—transport mass, EVA time, energy, failure frequency, and rescue access—rather than adjectives such as “better” or “promising.”
2. Predeploy before crew
Sending power, communications, inventory, and perhaps local production before people converts time into evidence. Equipment can accumulate operating hours and reveal faults before crew commitment.
Predeployed cargo can fail, land far away, or age in place. A robust campaign identifies what must work before departure, what requires duplication, and which failure combinations cancel a crew mission.
Quantified balance
Predeployment reduces human risk only when systems can be remotely tested in representative states. Cargo that has landed but cannot be connected, calibrated, or repaired is not an available capability. Cargo campaigns therefore need activation sequences, telemetry, spares, and acceptance criteria.
Adding flights does not automatically multiply reliability when they share a common design defect.
Cargo case — real availability
A water plant is predeployed in two modules. Both land, but a water-quality sensor fails and no backup standard is available. The mass is on Mars, yet the capability “qualified water” has not been demonstrated. Remote acceptance must test the end function, not merely electrical presence of hardware.
Decision check. Predeployment should include a formal readiness matrix: landed, powered, communicating, calibrated, functionally tested, stocked, and recoverable. A check mark under “landed” is only the first state. The crew should leave Earth only after the campaign has demonstrated the set of functions that must already exist on Mars.
3. Energy: infrastructure beneath all other infrastructure
Pumping, heating, compressing, lighting, recycling, manufacturing, and communication consume power. Budgets must separate average load, peaks, storage, availability, and critical loads. Installed power is not the same as available power during a fault.
NASA still describes the Mars Architecture Trade Space in March 2026 as relatively open. Some choices narrow the options, while transport and surface architecture remain evolvable. A Mars study should preserve technology status rather than treat a concept as accomplished capability.
Degraded mode
Energy connects every function to a time constraint. An industrial plant may draw much power yet tolerate interruption; ECLSS may draw less but require near-continuous service. Loads must therefore be ranked by criticality with shedding, restart, and black-start plans.
Storage must support peak power and startup current, not merely total energy.
Energy case — black start
After a total outage, critical loads total 35 kW but pumps and compressors need a 60 kW startup peak for 20 s. A battery with 100 kWh energy but only 45 kW power cannot restart the system without sequencing or another source. Stored energy and black-start power are separate constraints.
4. Water, air, food: shift from inventory to flow
Early missions can be dominated by imported stock. As duration and population grow, flows dominate. Ten tonnes sounds large until a 400 kg/day net loss consumes it in 25 days.
Recycling reduces imports while adding filters, sensors, power, and maintenance. Crops can reduce some food imports but demand water, nutrients, light, and sanitation. Autonomy is a network of coupled flows.
Safety reading
Water, air, and food move from inventory to flow as population and duration increase. The useful metric becomes daily deficit after recycling and local production. Large stocks remain valuable for failures but cannot indefinitely hide a structurally deficient loop.
A loop stable on average can still face demand peaks; buffers size timing mismatches, not just annual balance.
Flow case — buffer
An oxygen plant averages 45 kg/day while demand averages 40 kg/day. During five maintenance days production falls to zero. The nominal 5 kg/day surplus does not describe safety: at least 200 kg is needed to cover maintenance, plus margin, before the daily surplus can rebuild inventory.
Decision check. Flow thinking also changes how margins are reported. Instead of saying “we have 100 tonnes of water,” operations should know days of reserve at nominal demand, days at emergency demand, local production rate, repair time, and the rate at which the reserve can be rebuilt after an outage.
5. Local construction: think in tonnes
Teaching case: 2 m regolith cover, assumed bulk density 1,500 kg/m³, over 100 m². Moved mass = 2 × 1,500 × 100 = 300,000 kg = 300 t.
At average 250 kg/h, that is 1,200 h or 50 ideal continuous days. Charging, maintenance, relocation, weather, and quality control extend it. “Use regolith” becomes measurable industrial work.
Construction study — 300 tonnes of regolith is work, not a sketch
Take a teaching shielding layer 2 m thick, assumed bulk density 1,500 kg/m³, over 100 m². Moved mass is 2×1,500×100 = 300,000 kg or 300 t. A chain actually processing 250 kg/h requires 1,200 productive hours, equal to 50 ideal continuous days.
Calendar time is longer. At 70% availability, 1,200 productive hours require about 1,714 calendar hours or 71.4 days. If the machine averages 15 kW while working, productive energy is 18,000 kWh before charging and conversion losses. Local construction, power, maintenance, and logistics are therefore one architecture. Regolith is local; the work required to move it is not free.
System intuition
Local construction should be expressed in tonnes, machine hours, and verified quality. Shielding thickness, area, and volume quickly become hundreds of tonnes, turning a material choice into excavation, transport, power, and maintenance.
Raw material is not always directly usable; moisture, grain size, perchlorates, porosity, and process choice can change qualified output mass.
Construction case — throughput
Shielding requires 1,600 t material. Two machines move 12 t/h each but are available only 60% of 20 h/day. Daily capacity=2×12×20×0.60=288 t/day, about 5.6 ideal days. Soil preparation, travel, failures, thickness verification, and safety must be added before this minimum becomes a schedule.
6. Mobility: distance costs more than propellant
A resource 30 km away implies routes, vehicles, energy, navigation, spares, rescue, and two-way movement. Catalogue rover range is insufficient.
Ask how far the vehicle can travel, lose a function, and still return or wait safely. Logistics evaluates complete missions with failure margin.
Quantified balance
Mobility must be sized with degraded return. Nominal rover range is not safe range after loss of a battery, wheel, sensor, or communication link. Architecture can use paired vehicles, caches, rescue robots, or distance limits depending on scenario.
The one-half factor is only a round-trip intuition; terrain, detour, and rescue needs can require much larger margin.
Mobility case — rescue
A rover has 200 km nominal energy range but only 120 km with degraded battery and high heating load. A target 50 km away needs 100 km round trip before detour. With 20% detour the trip reaches 120 km and all margin disappears. Excursion limits must use degraded range, not nominal headline range.
Decision check. Mobility planning needs a rescue architecture. One option is paired crewed vehicles; another is an uncrewed recovery rover, route caches, or strict turn-back points. Each choice moves mass and complexity between the vehicle fleet and stationary infrastructure, so “range” is an architecture property rather than one battery number.
7. ISRU: demonstrate process, then design plant
MOXIE demonstrated oxygen production from Martian CO₂, with NASA reporting up to 12 g/h and 122 g total. A hypothetical 1 t/day demand is 41.7 kg/h, more than 3,400 times the demonstrated peak rate.
That ratio is not a plan for 3,400 copies. Industrial design changes compressors, cells, thermal rejection, power, storage, controls, maintenance, and availability. The correct status is small-scale principle demonstrated on Mars; industrialization remains to be engineered and qualified.
ISRU study — from demonstrated flow to guaranteed service
MOXIE demonstrated up to 12 g/h oxygen production on Mars and 122 g in total. Imagine a future nominal unit rated at 20 kg/h, a purely teaching value. At 90% availability across an 8,760-hour Earth year, output is 20×8,760×0.90 = 157,680 kg or 157.7 t/year. At 70%, it falls to 122.6 t/year. Twenty availability points are therefore worth about 35.0 t/year in this scenario.
This sensitivity is why industrial sizing includes maintenance, spares, contamination, cell life, compressors, storage, and power. Nameplate flow is not enough. Robust architecture separates instantaneous flow, availability, storage capacity, and cumulative demand. MOXIE supplies Mars proof of principle; availability of a future industrial plant remains a design variable.
Degraded mode
ISRU means replacing an import with a complete local chain. Extraction alone is insufficient: capture, purification, conversion, storage, measurement, and maintenance are required. MOXIE demonstrated one oxygen-production step from Martian CO₂; MARS-C is a 2026 active technology project, not an operational plant.
Scaling must be demonstrated in power, lifetime, maintenance, and purity, not merely nominal flow rate.
ISRU case — availability
A nominal 2 kg O₂/h plant operating 70% of the year produces 2×8,760×0.70=12,264 kg/year. At 90% it produces 15,768 kg/year, a 3.504 t difference. Maintainability can therefore be worth several tonnes per year without changing nominal flow by one gram.
8. Workshop, metrology, spare parts
A manufactured part is not automatically qualified for critical service. Material, dimensions, tolerances, surface state, treatments, and defects must be measured. NIST’s 2026 in-situ metrology work explicitly connects measurement with qualification and certification for metal additive manufacturing.
A Mars workshop therefore needs machines, instruments, standards, material data, and acceptance criteria. A simple bracket may become local while a bearing, sensor, or electronic component stays imported.
Workshop study — machine capacity and qualification
Suppose a workshop has one critical machine available 20 h/day after maintenance, tool changes, and checks. Three part families require 6 h, 8 h, and 10 h of machine time per batch. One day cannot produce all three batches because demand totals 24 h. The bottleneck is not “owning a printer” but scheduling a scarce qualified resource.
If a life-critical repair needing 7 h arrives, it must displace another job or wait. A second machine doubles capacity only if operators, feedstock, metrology, power, and tooling also scale. NIST’s 2026 workshop on in-situ metrology for metal additive manufacturing explicitly links measurement with qualification and certification. A part leaving a machine is not automatically an acceptable life-critical part.
Safety reading
A workshop is not an isolated printer: it needs feedstock, preparation, machine tools, metrology, qualification, design data, operators, and configuration control. NIST’s 2026 in-situ metrology workshop highlights measurement, qualification, and certification as central metal-additive challenges.
A produced part is not yet released for service until critical characteristics are verified.
Workshop case — qualification
A printer makes 10 parts/h but 15% fail inspection and 10% of time is lost to changeover. Over 16 scheduled hours, accepted output≈10×16×0.90×0.85=122 parts. The raw “160 parts” figure overstates useful capacity by ignoring support time and acceptance.
9. Population and skill depth
Four people are a mission team; twenty allow more specialties; one hundred begin to support permanent services; one thousand can sustain training pipelines but require much more complex infrastructure.
Headcount alone is insufficient. Track critical skills, available labor, training of replacements, and maintenance burden. A durable society must reproduce knowledge as well as parts.
System intuition
Population growth requires skill depth, not simply more people. A critical function needs backups, hands-on training, documentation, and a training time compatible with crew turnover.
A theoretical skill unused for a long time may not equal active operator capability; recurrent drills are part of human maintenance.
Skills case — crew rotation
A critical function has three specialists. One is on a long EVA, one is ill, and the third must supervise another emergency. Nominal “three qualified people” becomes zero practical availability. Planning must combine skill matrix, scheduling, simultaneous unavailability, and cross-training.
Decision check. Skill depth can be visualized as a matrix of people versus functions, with levels such as awareness, supervised operation, independent operation, diagnosis, and instructor. The goal is not to make everyone expert in everything; it is to prevent one absence from erasing the only local capability to restore a vital system.
10. Operational governance and safety
Life-critical systems need clear authority: who can shut down a plant, reject a water batch, draw emergency reserve, or isolate a zone? These are safety mechanisms before they are a full political constitution.
Broader institutional choices come later: resource allocation, investment, ownership, emergency duties, transparency, conflict. Separate constraints imposed by physics from choices created by institutions.
Quantified balance
Operational governance turns principles into concrete authorities: stop a line, isolate a habitat, use a reserve, change configuration, accept a part, or postpone an EVA. Rules need traceability while retaining a fast emergency path.
A procedure requiring immediate Earth communication is not robust to Mars delay and outages.
Governance case — plant shutdown
An ISRU line exceeds a temperature threshold but shutdown would sacrifice needed oxygen production. Rules must specify who decides, which threshold forces automatic shutdown, what reserve survives the outage, and how any override is recorded. Without those elements, “safety first” remains an intention rather than an executable rule.
Decision check. Operational governance should also define evidence thresholds. A technician may be allowed to stop a machine from one alarming sensor but require two independent checks before permanently rejecting a batch. Different actions justify different standards of evidence because delay and false action have different consequences.
11. Physical economy: mass, energy, machine time
Before currency, a settlement has a physical economy. A part consumes material, kWh, machine hours, human time, and risk. Saving one imported kilogram may be less valuable than tying up the rarest machine for twenty hours.
Indicators such as imported kg, kWh, availability, days of inventory, maintenance hours, and repair time compare architectures without pretending to forecast an entire future economy.
Degraded mode
Physical economy reveals bottlenecks before money is discussed: available energy, imported mass, machine hours, crew time, pressurized area, and inventory. A currency or price can allocate scarcity but cannot create additional kilowatt-hours.
The chain minimum dominates: improving a non-bottleneck station may not increase output.
Physical-economy case — bottleneck
The workshop has feedstock for 500 parts and energy for 700, but metrology can inspect only 80 parts/day. Production at 150 parts/day does not raise qualified output above 80 while inspection is the bottleneck. The useful investment may therefore be a measurement station rather than a second production machine.
Decision check. A bottleneck analysis should be repeated after every major expansion. Adding population changes demand, but adding one new machine can shift the limiting factor to power, operator time, metrology, storage, or feedstock. The useful question is always “what constrains qualified output now?” rather than “which subsystem looks busiest?”
12. Failure scenarios as design tools
A useful scenario is not “everything fails.” Combine plausible events: one cargo lost, water pump unavailable, post-EVA dust, a metrology instrument offline, and communication delay. The goal is to expose dependencies.
A good scenario produces action: increase a reserve, isolate an interface, simplify repair, train a second specialist, or add independent sensing. Failure analysis becomes design work.
Safety reading
Failure scenarios should expose couplings: power plus water, mobility plus radiation, workshop plus spare stock, sensor plus common software. Examine detection, isolation, degraded mode, repair time, and return to a qualified state.
Positive margin in a nominal scenario is insufficient when repair-time uncertainty is of the same order as the margin.
Combined failure — power plus water
A generator failure sheds industrial loads while the water processor needs heater power to restart. If emergency mode did not reserve that restart power, water remains unavailable even after nominal generation returns. Combined scenarios must therefore follow the time sequence of functions, not merely a list of failed components.
Decision check. Failure scenarios become more realistic when they include recovery order. Restoring power does not instantly restore water if heaters, pumps, and quality checks must restart in sequence. The timeline should therefore include detection, isolation, minimum safe state, repair, restart, verification, and replenishment of depleted reserves.
13. Four scales, four architectures
| Scale | Character | Question |
|---|---|---|
| 4 | integrated mission | Does one fault remain survivable? |
| 20 | specialized outpost | Which skills/spares need duplication? |
| 100 | permanent base | Which services become infrastructure? |
| 1,000 | industrial society | Which local chains replace imports? |
Progression is not automatic. One hundred people can still depend on one imported component. Scale reveals architectural transitions; it does not provide a colonization date.
14. Final project: a 100-person base
Build a ten-function matrix: power, water, atmosphere, food, habitat, health, mobility, communications, workshop, spares. For each, record nominal capability, capability after one fault, emergency stock, repair time, and common dependencies.
The project succeeds when each function can be explained as ready, acceptable in degraded mode, or blocking, with missing evidence clearly identified. A credible architecture is a network of explicit functions and limits, not a collection of impressive technologies.
Final project — physical architecture of a 100-person base
Start with ten functions: energy, water, atmosphere, food, habitat, health, mobility, communications, workshop, and spares. For each, record nominal capacity and, more importantly, capacity after one fault. A 100-person base with 500 kW generation but only 120 kWh storage has a different resilience profile from one with the same generation and several days of reserve. Quantities are not interchangeable: power, energy, flow, inventory mass, and repair time answer different questions.
For water, use a teaching net loss of 40 kg/day after recovery. Thirty days of makeup is 1,200 kg. A ninety-day strategic reserve is 3.6 t. If loss doubles during degraded operation, a nominal ninety-day reserve becomes forty-five days. Inventory has not changed; endurance changed because flow changed. The base must therefore monitor loss trends as well as tank levels.
For construction, a 300 t regolith job at 250 kg/h is 1,200 productive hours. If two identical machines are genuinely independent, nominal throughput may approach 500 kg/h. Yet one maintenance workshop, one charging system, or one shared wear part can limit the gain. For every capacity, record the dependency that can make failures common. That column often reveals more than the machine count.
For industry, assume one critical machine tool is available 18 h/day and planned demand is 15 h/day. Nominal utilization is 83%. An unexpected eight-hour repair cannot be absorbed the same day without delay. Sufficient average capacity can still lack time margin. Add a backlog indicator measured in waiting machine-hours. If backlog rises for several days, the workshop is structurally under-capacity even though no machine is completely failed.
For skills, count more than people. List functions in which only one specialist can authorize or perform an operation: surgery, quality control, high voltage, critical software, chemistry, compressor maintenance. A hundred-person settlement can be demographically large and technically fragile. The plan needs apprenticeship, documentation, simulators, drills, and protected training time. Reproducing skills is infrastructure just as surely as inventorying spare parts.
Finish with three combined scenarios: loss of an annual cargo, a 72-hour ECLSS train outage, and simultaneous loss of a metrology machine. For each, identify functions maintained, degraded services, inventories consumed, recovery time, and the decision that becomes irreversible. The goal is not to predict the first Mars city but to show that settlement architecture can be discussed through balances, dependencies, degraded modes, and evidence.
| Function | Useful measure | Resilience question |
|---|---|---|
| Energy | kW + kWh + availability | how long after generation loss? |
| Water | kg/day loss + reserve tonnes | what endurance at degraded flow? |
| Workshop | machine h/day + backlog | which repair delays others? |
| Skills | qualified people per function | which task has one human point of failure? |
Corrected drill set — independent check
1. 100 km radius: approximate geometric area?
π×100²≈31,400 km². The number does not mean 31,400 km² is traversable; it only gives a geometric envelope for the problem.
2. 300 t at 250 kg/h: productive hours?
300,000/250 = 1,200 h. At 50% availability, ideal calendar time doubles to 2,400 h before other constraints.
3. Machine available 18 h/day, demand 15 h/day: utilization?
15/18≈83%. High utilization leaves little margin for urgent repairs and can create backlog without a long failure.
4. Why can locally making a part be more expensive than importing its mass?
It may consume scarce machine time, energy, operator effort, metrology, qualified feedstock, and schedule margin. Compare the complete resource system, not transport mass alone.
5. What indicator reveals a single human point of failure?
Count independently qualified people for each critical function. A skill held by one person remains a dependency even in a large population.
Deep practice workshop
In these settlement exercises, do not treat a Mars base as a simple list of equipment. Start from the required function and trace its dependencies through power, water, mobility, workshop capability, skills, and spares. Each solution looks for the bottleneck and shows why an architecture that works for four people may fail at one hundred or one thousand.
1. Exploration Zone
A target is 80 km straight-line with 25% route detour. One-way real distance?
Reasoned solution : 100 km. Geographic radius does not replace a traversable route.
The detour turns 80 geometric kilometres into 100 operational kilometres. That difference becomes energy, crew time, wear, and return margin; an Exploration Zone is therefore more than a circle on a map.
2. Energy
150 kW average for 24 h.
Reasoned solution : 3.6 MWh/day.
3. Local shielding
150 m² covered by 1.5 m material at 1,500 kg/m³.
Reasoned solution : Volume=225 m³; mass=337,500 kg=337.5 t.
The 337.5 tonnes comes from 225 m³ multiplied by the assumed density. Before promising that shielding, the mass has to be converted into excavation, hauling, compaction time, and equipment availability.
4. Workshop availability
Machine 90% available, 2 h/day support, 4 parts/h.
Reasoned solution : 24×0.9=21.6 h; minus 2=19.6 h; 78.4 parts/day before yield/scrap.
Availability first reduces twenty-four theoretical hours to 21.6; subtracting support time leaves 19.6 productive hours. Applying rate last shows why people, maintenance, and uptime matter as much as nominal machine speed.
5. Cargo campaign
4 independent flights at 96% each: all succeed?
Reasoned solution : 0.96^4≈84.9%.
The 0.96⁴ product assumes independent flights. The 84.9% result shows how high single-flight reliability becomes less comfortable when a campaign requires every element to succeed.
6. Water
100 people, 0.4 kg/day makeup each.
Reasoned solution : 40 kg/day, 14.6 t/year.
7. Mobility
60 km map distance, 30% detour, 12 km/h.
Reasoned solution : 78 km; 6.5 h ideal one-way drive time.
8. Growth capstone
A base grows from 20 to 100 people. Water makeup scales linearly at 0.4 kg/person/day, but maintenance has two machines rated 60 parts/day each at 80% availability. Compare the changes.
Reasoned solution : Water rises from 8 to 40 kg/day, a factor of five. Corrected workshop capacity is 2×60×0.8=96 parts/day; it does not automatically multiply with population. If part demand also grows fivefold, machines, staffing, hours, or productivity must increase. Some flows scale with population while infrastructure grows in steps.
Water scales linearly with population in this exercise, but the workshop does not: machines and skilled shifts are discrete and have availability limits. Moving from 20 to 100 people can therefore require an architectural step change rather than a simple factor of five.
Additional advanced problems
1. Local material
500 m² × 2 m × 1,600 kg/m³.
Reasoned solution : 1,600,000 kg=1,600 t.
2. Workshop
2 machines, 6 parts/h, 18 h/day, 85% availability, 90% yield.
Reasoned solution : 2×6×18×0.85×0.90≈165 parts/day.
The roughly 165 parts per day multiply machine count, rate, time, availability, and yield. Each factor has a different loss mechanism; the product is useful because it exposes where an improvement or failure acts.
3. Campaign
5 independent cargo flights at 97%. All succeed?
Reasoned solution : 0.97^5≈85.9%.
Five cargoes at 97% each give only about 85.9% probability that all succeed under independence. A robust campaign can size inventories and sequence work to survive one missing delivery instead of assuming perfection.
4. Growth mini-project
100 people need 40 kg/day water makeup and a plant produces 50 kg/day. Ideal annual surplus?
Reasoned solution : 10 kg/day×365=3.65 t/year. It becomes reserve only if storage, quality, and plant availability remain adequate.
The 3.65 t/year surplus exists only if plant output is sustained and the product remains qualified and storable. Production margin is different from reserve water that is already available after a failure.
Sources and references
These references frame site selection, the still-open Mars architecture, the MOXIE demonstration, and industrial qualification. Plant, availability, and construction values remain clearly identified Delta-Sierra teaching scenarios.
- NASA — Mars Architecture Trade Space — 2026 status of the still-open Mars architecture trade space.
- NASA NTRS — Human Mars Landing Site and Surface Operations — Exploration Zone concept and roughly 100 km regions-of-interest framework.
- NASA/JPL — MOXIE completes Mars mission — 122 g total O2, up to 12 g/h, at least 98% purity.
- NIST — In-Situ Metrology for Metal Additive Manufacturing — 2026 measurement-science roadmap for qualification-ready metal AM.
- NASA TechPort — MARS-C — Primary or institutional source used for definitions, technology status, or module data.