Mars Site Selection

What is being decided

Site Selection Is a Decision Framework, Not a Measurement

Nobody has the data to pick a Mars site. What can be done is to make the trade explicit and find out what the answer is sensitive to.

A landing site has to satisfy four different customers at once. Entry, descent and landing wants low elevation for atmospheric column and flat, rock-free ground. The propellant plant wants water in whatever form its architecture needs. The power system wants insolation, which means low latitude. The crew wants a place they can leave from, which loops back to propellant. These pull in different directions, and the weighting between them is where the argument actually lives.

The screen used here scores eight named candidates against a criteria set, then sweeps the weights to find where the ranking changes. That second step is the useful one. A ranking is an output of a weighting; a break-even weight is a fact about the trade — and the break-even is what survives when somebody disagrees with the scoring.

One caveat, stated up front rather than buried: the qualitative scores behind this page are judgement read off published site characterizations, not values pulled from SWIM ice maps, MOLA topography, THEMIS thermal inertia, Odyssey gamma-ray data or HiRISE rock-abundance products. This is a decision framework. It is not a site selection, and no site on Mars has ever been evaluated with the subsurface data that an actual resource decision requires.

The candidates

Eight Candidate Regions, 4°N to 40°S

All are low-lying plains or basins, because entry, descent and landing rules out everything else.

The legend is the finding in miniature. The set splits into two families by water strategy, and that split — not distance, not geology — is what makes the ranking unstable.

The criteria

What the Screen Measures, and Which Architecture Each Criterion Favors

Only one criterion is architecture-sensitive. It happens to be the heaviest.

Seven of the eight are properties of the place and stay put. The first one is not: how heavily ice should count depends entirely on how much water the program needs, and that is set by the vehicle. One architecture-sensitive criterion, carrying enough weight to move the answer — which is the whole mechanism of the flip below.

The flip

Arcadia Under Direct-to-Surface, Isidis Under Orbit Staging

The break-even is an ice weighting of about 12 out of 100. Below it the equator wins; above it the north does.

A vehicle that lands and refuels for Earth return needs roughly 570 t of water per window. At that demand only high-grade, shallow, well-mapped ice will do, and that means going north to about 39°N — Arcadia Planitia, with its widespread shallow ice, and the winter insolation penalty that comes with the latitude.

A vehicle that stays in orbit while a ferry shuttles crew needs roughly 75 t. At that demand ordinary equatorial hydrated regolith at 3–8 wt% suffices, no special deposit required, and ice stops being a discriminator. The weighting shifts onto insolation, terrain and orbital access — all of which favor the equator — and Isidis Planitia at 13°N moves from the back of the field to the front while Arcadia slides to the middle.

Sweeping the ice weight locates the crossing at about 12.4 out of 100. Orbit staging justifies a weight near 8; direct-to-surface justifies something near 22. So the equatorial answer holds as long as ice matters less than roughly an eighth of the total — meaning the water requirement would have to be wrong by about 50% before the ranking reverts north. That is the number worth defending, rather than the ranking it produces.

Amazonis Planitia is the low-regret answer. It sits near the front under both weightings, moves less than a point between them, and is the most rank-stable site across the entire sweep. A program that must choose a site before it has chosen an architecture — before MMX resolves Phobos, before anyone funds a descent and ascent vehicle — should take the site that does not care which way those decisions go.

Isidis carries one soft spot worth naming. Its strong showing on hydrated minerals leans on Nili Fossae adjacency, and Nili Fossae sits hundreds of kilometers from any plausible landing ellipse. Adjacency on a global map is not access on the ground.

Hellas

Hellas Planitia: The Option That Looks Compelling and Is Not

It wins the single criterion everyone cares about most, and finishes last under every weighting.

Hellas is the deepest basin on Mars, roughly 7 km below datum, which gives it about double the atmospheric column of a typical northern plain. That is a genuine and large entry advantage: it is the only candidate where a Starship-class vehicle’s terminal velocity drops below about 350 m/s, which buys real propellant margin on landing and real forgiveness in the descent profile. On elevation alone it scores a perfect mark.

It finishes last anyway, under both weightings. Hellas is a dust storm nursery — regional storms initiate in and around the basin with high frequency — and its floor terrain is poorly characterized and probably rough. Its water is the least understood of the eight. The southern latitude puts it on the wrong side of the hemispheric insolation asymmetry for part of the year even though perihelion favors southern summer.

It earns a section because the pattern generalizes. A site that dominates one criterion and fails several others will keep resurfacing in proposals, because the dominant criterion is easy to explain and the failures are diffuse. The honest statement is that the entry advantage is real, it is not enough, and anyone reviving Hellas needs to answer the storm question first.

The triple-point coupling

Every Candidate Sits Above the Triple Point, and That Breaks the Suits

The preference that makes landing easy is the same one that disables sublimation cooling.

Entry, descent and landing wants atmospheric column, so every serious candidate sits below datum. Below datum means higher pressure: Arcadia Planitia runs about 867 Pa against the 611.657 Pa triple point of water, and Hellas runs higher still. Every site on the list above is above the triple point, and none of them got there by accident — the selection criterion put them there.

Spacesuits reject heat by sublimating water, a method that has worked on every spacewalk since Apollo and requires ambient pressure below the triple point. At 867 Pa it does not work. A Mars suit therefore needs a thermal rejection scheme that 60 years of EVA experience does not cover, and the requirement was created by a landing-site preference chosen by a different community for unrelated reasons. Neither side of that trade has ever been priced against the other.

The second site

The Second Settlement Is Chosen by Decorrelation, Not by Score

Running the same screen twice produces two copies of the same site, which defeats the purpose of having two.

A weighted score and a decorrelation criterion are anti-correlated by construction, because the score rewards the same properties both times. Take the top two finishers in this set and you get two northern low-elevation plains about 2,000 km apart, sharing a water strategy, a seasonal insolation profile, an excavation problem and a node inclination. That is a second copy, not a second instance, and it buys almost none of the redundancy the second settlement exists to provide.

The correct structure runs the screen twice with different objectives. Site one maximizes expected value, or minimizes regret. Site two maximizes decorrelation subject to viability, and it will score worse — in this candidate set the most decorrelated site is Hellas, which is also the lowest-ranked, exactly as the anti-correlation predicts.

Decorrelation also propagates upward into the orbital architecture, and this cost is unpriced everywhere. A moon node is inclination-locked within about 1° of the equator, so serving a latitude-separated pair from Phobos or Deimos costs a plane change of order 0.7–1.0 km/s per sortie, both directions, every sortie, for 25–40° of separation. Low Mars orbit carries no such lock. The two-settlement requirement pushes the node down, or pushes the program to two nodes — see Mars Orbit, Phobos & Deimos.

One consequence resolves an apparent inconsistency rather than creating one. The outpost analysis favors a northern high-latitude site; the architecture review favors an equatorial low-regret pick. Read as a single choice, they contradict. Read as a pair, they are the decorrelated set, and the program should want both.

Who has picked what

Published Site Choices: NASA, SpaceX and China

Three programs, three different weightings, none of them stated as such.

NASA ran the broadest exercise: the 2015 Human Landing Sites Study collected roughly 50 proposed Exploration Zones, each a 100 km radius region containing both science targets and resource targets within crew traverse range. It was a science-and-resource screen rather than an architecture screen, and it has not been narrowed since.

SpaceX has published target areas in Arcadia Planitia for early uncrewed Starship landings, chosen for shallow ice and smooth terrain. That is the correct choice for the direct-to-surface architecture the company is building, and it illustrates the finding rather than contradicting it: the site follows the vehicle.

China landed Tianwen-1’s Zhurong rover in Utopia Planitia in May 2021 at about 25°N, a northern plain in the same family as several candidates here. For Tianwen-3, the sample-return mission flying in the 2028 and 2031 windows, CNSA ran a published down-selection from dozens of candidate regions using biosignature preservation as the primary criterion, concentrated in the northern low latitudes below about −3 km elevation. Different objective, and the geography lands in strikingly similar territory — low, northern, and above the triple point.

What would overturn this

The Measurements That Would Change the Answer

Each of these is a specific, fundable observation, and none of them is currently scheduled.

Record

Provenance and Data Sources