Lunar Surface Environment

The lunar surface environment as characterized for engineering use. Design values are taken from the Cross-Program Design Specification for Natural Environments (DSNE, SLS-SPEC-159) [1] and the Lunar Sourcebook [2], with DSNE section numbers given for traceability. Terramechanics, thermal and illumination values are cited to the measurement or modeling documents that establish them.

Orbital and illumination geometry

The sidereal rotation period is 655.73 h. The synodic period, one full day and night cycle relative to the Sun, is 708.73 h, or 29 d 12 h 44 min, varying slightly through the year with orbital eccentricity (DSNE §3.4.9) [1].

Rotational obliquity is 1.54 degrees relative to the ecliptic, against 23.4 degrees for Earth, and the orbit is inclined 5.16 degrees to the ecliptic [1]. The small obliquity is what produces the polar illumination regime: the Sun never rises far above the horizon at high latitude, and never sets far below it. The solar disk subtends about 0.53 degrees, so the penumbral edge of any topographic shadow is narrow [3].

Equatorial sites are illuminated for approximately 15 Earth days followed by 15 days of darkness [1]. Polar illumination admits no general statement, because local topography dominates.

Polar illumination

Illumination at the poles is computed by ray-tracing the Sun against a digital terrain model rather than by any latitude rule. The reference calculation uses Lunar Orbiter Laser Altimeter models at 20 m/pixel covering 400 by 400 km at each pole, evaluated hourly over 1 January 2017 to 1 January 2037, a span chosen to cover the 18.6 year lunar precessional cycle [4]. New LOLA south polar products extend this to 5 m/pixel over selected sites [5].

Averaged over that twenty-year span at ground level, the best sites receive between 69.5 and 82.9 percent illumination, the maximum being the equator-facing rim of Whipple crater at the north pole [4]. Six named sites carry the high-illumination clusters: Hinshelwood, Peary and Whipple in the north, and Shackleton crater and two locations on the Connecting Ridge between Shackleton and de Gerlache in the south.

Illumination is a strong function of height above the surface, which makes it a solar array mounting parameter rather than a site property; power-system studies of the poles treat array height as a design variable for that reason [6]. Raised 2 m above ground, the same sites receive 77.1 to 88.0 percent, the maximum at Connecting Ridge C1 near the south pole, where solar visibility reaches 92.1 percent [4]. Maximum continuous darkness at 2 m height, from the same model:

Site	Pole	Max. average illumination at 2 m (%)	Longest shadow at 2 m (h)	Longest shadow at surface (h)
Hinshelwood	North	81.5	203	392
Peary	North	77.1	124	334
Whipple	North	85.1	101	101
Connecting Ridge C1	South	88.0	112	268
Connecting Ridge C2	South	78.1	165	166
Shackleton	South	85.5	66	221

Raising the observation point by 2 m shortens the longest shadow interval by up to 70 percent at Hinshelwood, Peary, Connecting Ridge C1 and Shackleton, and does nothing at Whipple and Connecting Ridge C2 [4]. The locations of maximum average illumination at 2 m do not coincide with those at ground level, so a site chosen on a surface illumination map is not the site a mast-mounted array would choose. Energy storage must still cover the longest shadow interval, which for the best south polar cluster is 112 h, and thermal storage in sintered regolith has been proposed as an alternative to carrying that energy as batteries [6].

The high-illumination clusters are small and sit next to permanently shadowed ground. The best south polar cluster, Connecting Ridge C1-0, spans 135 200 m² and lies 100 m from a 5200 m² permanently shadowed region whose average and maximum internal slopes are 8.4 and 12.8 degrees [4]. That proximity is what makes these sites operationally interesting and is also why traverse planning between illuminated and shadowed ground is a power-management problem on the scale of hundreds of meters.

Thermal environment

Surface temperature extremes, from DSNE Table 3.4.6.3-1 [1], after Williams et al. (2017), with the PSR value after Vasavada et al. (2012):

Location	Mean (K)	1σ extreme (K)	Condition
Equator, maximum	391	394	Local noon
Equator, minimum	96	94	Before sunrise
45° latitude, maximum	350	357	Local noon
45° latitude, minimum	89	83	Before sunrise
85° latitude, maximum	182	224	Local noon
85° latitude, minimum	61	41	Approx. 03:00 local
Coldest permanently shadowed crater	30		No direct illumination

The equatorial range spans 295 K within a single cycle, and space sink temperature is 3 K [1]. The bounding solar flux values for lunar thermal analysis are 1310 W/m² for the cold case and 1426 W/m² for the hot case, and neither includes the ±5 W/m² measurement uncertainty that enveloping analyses are told to add [3]. Lunar albedo inputs to the long-wave radiance calculation span 0.07 to 0.20, with surface emissivity taken as 0.95 in the cold case and 0.98 in the hot case. Unilluminated surface temperature is taken as 80 K in the cold case and 120 K in the hot case [3].

Polar shaded temperatures are set by how completely a spot is screened from direct sunlight and from the infrared reradiated by nearby illuminated terrain, so they are a property of local geometry rather than of latitude. Maximum annual surface temperatures inside mapped south polar PSRs are 100 to 112 K [7]. No shaded floor cools without limit: heat conducted up from the interior, taken as a geothermal flux of 0.018 W/m² in the Diviner-calibrated regolith model, is the lower boundary condition on any polar thermal calculation [8].

Subsurface temperature stabilizes rapidly with depth. The diurnal thermal skin depth is about 4 to 10 cm for typical upper regolith [8], so diurnal variation is confined to the uppermost few tens of centimeters and below that the regolith acts as a thermal buffer rather than a load. Diviner-derived thermal conductivity of the regolith fines rises from 7.4 × 10⁻⁴ W/(m·K) at the surface to 3.4 × 10⁻³ W/(m·K) near 1 m depth, tracking a density increase from 1100 to 1800 kg/m³, and the density and conductivity scale height H has a global mean of 0.068 m, equivalent to a thermal inertia of 55 ± 2 J/(m²·K·√s) at 273 K [8]. Ground heat flux for analysis is 0.021 W/m² at Hadley Rille from the Apollo 15 heat probe, 0.016 W/m² at Taurus-Littrow from Apollo 17, and 0.018 W/m² as a global average adjusted from the two probe sites for subsurface potassium, uranium and thorium content. Polar values are lower: 0.0112 W/m² for Cabeus and Haworth, 0.0074 W/m² for Shoemaker, 0.01 W/m² at the LCROSS impact site and 0.009 W/m² at the south pole, all four from Chang’e-2 microwave radiometer data [3].

Thermal conductivity of the fines measured in place and measured in a laboratory on returned samples do not agree, and the gap is a factor of about five. The Apollo 15 and 17 heat flow probes give 0.9 to 1.3 × 10⁻² W/(m·K) for undisturbed regolith below the borestem contact zone, over the 15 to 186 cm depth range the annual thermal wave resolves, from the attenuation and phase shift of that wave through multi-year subsurface temperature records [74]. The laboratory value on Apollo 12 returned fines is 0.207 × 10⁻² W/(m·K) at 304 K, measured by transient line heat source at 10⁻⁶ torr on a 5.49 g sample poured to 1300 kg/m³. Both are measurements, and the source plots them together against porosity on one axis, so the comparison is the paper’s own rather than assembled here; the laboratory half is quoted from Cremers and Birkebak rather than remeasured, which makes this a comparison of published values and not a controlled pairing.

The difference is density and porosity, not measurement error. The poured laboratory sample sits at 1300 kg/m³ against 1.8 to 2.0 g/cm³ for regolith in place below 20 cm, which is the depth below which conductivity is near its maximum attainable value, having risen rapidly through the upper 10 cm [74]. A conductivity measured in a laboratory on poured returned fines is therefore not a lunar surface value.

The heat flow figures quoted above carry their own conditions. Converting the measured diffusivity to a conductivity uses a bulk density of 1.75 to 1.90 g/cm³ at Apollo 15 and 1.83 to 2.09 g/cm³ at Apollo 17, taken from drill core and drive tube samples, values the source notes are probably minimum estimates because drilling disrupted the soil [74]. Uncertainty on both site fluxes is about ±15 percent, dominated by the resolution with which the annual wave fixes the diffusivity, and the Taurus-Littrow probe value of 0.016 W/m² is corrected downward for topography to 0.014 W/m² as the regional estimate. Two sites and four probes, on a body whose regolith properties are set by local impact history.

Effective radiative sink temperature is a terrain quantity, not a latitude quantity. Surface temperature is set by the solar incidence angle that local slope produces and by radiation exchange with the surrounding regolith, so flat mare terrain with its higher solar absorptivity runs hotter than highland terrain, which is steeper and more reflective. Minimum temperatures, reached just before local sunrise, depend on the adjacent terrain and on the platform the hardware is mounted to. Regolith thermal inertia is low enough that even a brief shadow drops the surface to near nighttime temperatures, and a briefly illuminated shadowed patch heats just as fast [9]. A closed-form surface temperature function fitted to Diviner observations, peaking at 392 K at the subsolar point, is available for analyses that need temperature as an analytic function of solar incidence rather than a lookup [10].

Surface temperature at the subsolar point follows a Stefan-Boltzmann balance, and variation about that point is approximately ±20 K. As solar incidence angle increases toward 80 degrees the temperature range widens to about 60 K, though the corresponding irradiance variation is smaller because of the fourth-power dependence [1]. Thermal buffering of the regolith itself has been proposed as night-survival infrastructure, with sintered regolith thermal wadis storing daytime solar energy for release through the 14 Earth day equatorial night, of which about 350 h is the cryogenic period; a constant 25 W/m2 from the wadi holds a rover at its 243 K minimum operating temperature [6].

Regolith physical properties

Regolith is the fragmental layer produced by prolonged meteoroid bombardment. It behaves mechanically as a silty sand, with mean grain size between 45 and 100 µm [2]. Grains are angular: they are produced by impact comminution rather than fluvial or aeolian transport, so no rounding mechanism operates. Much of the fine material consists of agglutinates, glass-welded composite particles formed by micrometeoroid impact melting. Grain size statistics compiled for mobility testing give a mean particle size of 40 to 800 µm with the majority between 60 and 80 µm [11], and the reference characterization against which simulants are graded scores lithic fragments, mineral grains, glasses and agglutinates as separate components of material composition [12].

Bulk properties representative of intercrater areas, from DSNE Table 3.4.2.3-1 [1], after Carrier et al. (1991):

Property	0 to 30 cm	30 to 60 cm
Bulk density ρ (g/cm³)	1.58 ± 0.05	1.74 ± 0.05
Relative density D_R (%)	74 ± 3	92 ± 3
Porosity n (%)	49 ± 2	44 ± 2
Void ratio e	0.96 ± 0.07	0.78 ± 0.07

Specific gravity is 3.1 [1]. Density increases with depth following a curve fit to Apollo 15, 16, and 17 core data (DSNE §3.4.2.3.1):

    ρ = 1.92 (z + 12.2) / (z + 18)   [1]

with ρ in g/cm³ and depth z in cm. These are minimum values; larger in-place densities occur. Terramechanics practice quotes 1.53 to 1.63 g/cm³ over the first 30 cm [11], consistent with the DSNE figure and its stated dispersion.

The upper few centimeters are loose, but relative density reaches 92 percent by 60 cm [1], which is dense sand in terrestrial terms, and any subsurface access mechanism sees increasing resistance over its first half meter.

That density profile is not self-generated. Soil at 50 cm would have had to be compressed above 100 kPa, more than 100 times its present overburden stress, to reach the measured density, so a post-deposition mechanism is required [13]. Vibration is the traditional candidate and does not fit: compacting JSC-1A to the asymptotic subsurface value of 1.95 g/cm3 requires accelerations around 0.23 g [13], and laboratory vibration raises relative density essentially independently of overburden weight, so it cannot produce an increase of density with depth [14]. Thermal cycling does: five cycles between ambient and 250 °C compacted simulant by 0.263 mm at zero overburden rising monotonically to 1.3812 mm at 0.9 m simulated depth [14]. Hypervelocity impact experiments, scaled by ejecta volume against compressed volume, produce net loosening rather than net compaction [13].

The engineering consequence is latitudinal. If thermal cycling is what compacts the regolith, compaction should fall with insolation, and Diviner data compared against thermal modeling shows exactly that: soil within about ±60 degrees latitude is homogeneous, and beyond ±60 degrees it becomes systematically looser [13]. Material inside a permanently shadowed region has never been thermally cycled at all [14]. The only ground truth is the LCROSS impact into Cabeus, where a 300 ms delay before the infrared flash and a suppressed visible flash imply a target of greater than 70 percent porosity [13]. Bulk density values quoted from Apollo cores are therefore upper bounds for polar work, not design values.

Regolith mechanical properties

Shear strength follows a Mohr-Coulomb relation combining a cohesion term independent of applied stress and a frictional term proportional to normal stress [2]. Shear strength governs bearing capacity, slope stability, and trafficability.

Estimates evolved substantially. Pre-Apollo values derived from the Surveyor 3 and 7 soil mechanics surface samplers proved to lie near the lower bound of actual strength, and the Apollo Model of Mitchell et al. (1972, 1974), drawing on Apollo and Lunokhod data, supersedes them [2]. The LRV design specification, written before Apollo 11 data was available, carried cohesion 0 to 0.34 kPa, internal friction angle 35 ± 4 degrees and bulk density 0.80 to 1.60 g/cm³ [15]. Current mobility practice quotes cohesion up to 1 kPa and friction angles of 30 to 50 degrees [11]. The parameter set used in the NASA Lunar Terrain Vehicle simulation takes a single representative point in that range: cohesion 170 N/m², friction angle 30 to 40 degrees, weight density 2470 N/m³ [16]. Higher shear strength corresponds to higher relative density, so strength and depth increase together [1].

The published values do not converge, and they separate by method and by the stress range each was fitted over rather than by the soil. The Apollo surface synthesis, drawn from penetrometer readings, core tubes, trench walls and footprint and track depths at landing sites 11 through 17, gives cohesion 0.1 to 1.0 kPa and friction angle 35 to 50 degrees, both rising with density and depth [43]. The Lunar Sourcebook recommends a single typical pair for the top 60 cm of intercrater ground, 1.6 kPa and 49 degrees, obtained by inverting penetrometer curves against simulant shear data [2]. Summary values across the Luna 16, Luna 20, Lunokhod 1 and Lunokhod 2 datasets, from a Coulomb shear apparatus and a vane, are cohesion 0.04 to 0.06 kg/cm2 and friction angle 20 to 25 degrees [69]. Those are 3.9 to 5.9 kPa, four to sixty times the Apollo in situ cohesion, and the Sourcebook attributes the disagreement to the confining stress a laboratory apparatus applies rather than to a different soil [2]. Friction angles above 55 degrees are quoted for medium to high density material from early mission results [45]. Three vacuum direct shear tests on a 200 g sieved and recompacted Apollo 12 sample at 30 to 70 kPa normal stress returned 28 to 35 degrees with cohesion 0 to 0.7 kPa, weaker than a basaltic simulant because the agglutinates crush [2].

Bearing capacity separates further, because the figure scales with footing width. A reference envelope for a simulant program states ultimate bearing capacity of 25 to 55 kPa in intercrater areas and below 25 kPa at crater rims, alongside a calculated 6 to 419 kPa for a 0.1 m footing on level ground and 6000 kPa for a 1 m footing [70]. The 6000 kPa figure recurs on its own, from plasticity theory applied to a 1 m footing at in situ density and shear strength [2], and the allowable value against a 1 cm settlement limit at 95 percent confidence for a footing narrower than 0.5 m is 2 kPa, derived from 776 bootprints [2]. A bearing capacity quoted without its footing width and its settlement criterion spans more than two orders of magnitude.

Cohesion is small in absolute terms but not negligible at lunar gravity, and is sufficient to hold near-vertical trench walls in undisturbed material [2]. It is also site-dependent. The Chang’e-6 far-side sample has an average static angle of repose of 52.9 degrees, larger than every Chang’e-5 and Apollo simulant, and an average dynamic angle of repose of 70.4 degrees at 0.1 rad/s, which is read as stronger cohesion and attributed to high plagioclase abundance and impact reworking [17]. A single cohesion figure carried across sites is therefore a modeling convenience rather than a measurement. Where no lander data exists at all, bearing capacity can be inferred from boulder tracks in orbital imagery; the same method applied to asteroid Didymos returns about 0.07 kPa against 100 to 1000 kPa for Apollo lunar regolith, which sets the dynamic range the technique has to span [18]. Mitchell and Houston’s stabilization study is the source that treats cohesion, friction and the depth dependence together as a single engineering model [19].

Terramechanics

Terramechanics is the mechanics of vehicle interaction with deformable ground. It converts the soil properties above into predicted mobility.

Pressure and sinkage

Sinkage under a loaded wheel is conventionally represented by a pressure-sinkage relation of the Bekker form,

p = (k_c / b + k_φ) z^n

where p is contact pressure, z is sinkage, b is the smaller dimension of the contact patch, n is a deformation exponent, and k_c and k_φ are cohesive and frictional moduli [16]. The parameters are fitted from plate-sinkage measurement rather than derived from material constants, and they depend on plate size and on wheel loading, so they must be measured at a scale relevant to the vehicle [20].

No true Bekker parameters exist for lunar regolith, because obtaining them would have required in-situ plate-sinkage measurement on the surface. The Apollo-era values were estimated from Surveyor data to support Apollo simulation, should be treated as rough estimates, and do not span the range of possible lunar regolith behavior; almost no data exists for South Pole soil at all [20]. The values in circulation are therefore design conventions with a measurement pedigree, and differ between sources:

Parameter	LRV design set [15]	LTV simulation set [16]
Exponent of sinkage n	1.0	1.0
Cohesive modulus k_c	0 to 0.70 kN/m^(n+1)	1400 N/m²
Frictional modulus k_φ	81.4 kN/m^(n+2)	820 000 N/m³
Shear deformation modulus K	1.8 ± 0.8 cm	0.018 m
Cohesion c	0 to 0.34 kPa	170 N/m²
Friction angle φ	35 ± 4 deg	30 to 40 deg
Wheel-surface friction coefficient µ	0.6	not used

Measurements on lunar soil simulant at the Waterways Experiment Station returned k_c of 0.07 to 0.16 (kN/m)(cm⁻ⁿ), k_φ of 58.2 to 74.6 (kN/m²)(cm⁻ⁿ) and n averaging 0.49 to 0.51, an exponent half the design value of 1.0 [21]. The fitted k_c is not reliably positive. Bevameter plate penetration on the crushed basalt simulant used to qualify the Apollo LRV wheel averaged k_c of 0.42 lb/in^(1+n) over four tests in the loose condition LSS1, with individual fits from minus 0.76 to plus 1.28, and minus 1.58 lb/in^(1+n) over three tests in the dense condition LSS3, where all three fits were negative, from minus 2.44 to minus 0.94 [72]. A negative cohesive modulus of sinkage has no physical meaning, and its appearance in a fit to a real soil is what the analytical objection predicts: k_c, k_φ and n combine geometric parameters with the plasticity constants c and φ and change with the form and size of the test plate, so they are regression coefficients for one plate on one soil state and not material constants [71]. The exponent n from the same four conditions is 0.90, 1.15, 1.48 and 1.18, and does not track density monotonically [72]. Bearing capacity in the LTV model is closed with Terzaghi factors N_q = 32.23, N_c = 48.09 and N_γ = 33.27 evaluated at the design friction angle [16].

Shear and thrust

Available thrust follows from the shear stress the soil can develop, described by a Mohr-Coulomb criterion, τ_max = c + σ tan φ, with the mobilized stress rising with shear displacement j approximately as

    τ = τ_max (1 - e^(-j/K))   [16]

where K is a shear deformation modulus [16]. Shear displacement accumulates with wheel slip, so thrust is a function of slip rather than of torque alone. K for lunar soil is taken as 1.8 cm [15][16].

The K = 2.54 cm that recurs in the terramechanics literature as one inch entered it without a measurement behind it [71]. It was adopted for a rigid wheel on dry quartz sand and was explicitly not measured for the sand it was applied to, on the stated grounds that determining K is too subjective, and the accompanying bow wave factor was reduced from 0.65 to 0.5 for convenience while a round tan phi of 0.5 was used in place of the 0.6 and 0.65 of the two sands actually tested [71]. A comparative study of the same sand in a linear shear box and an annular ring shear device found K differing between the two by a factor of 1:2 to 2:1, with the absolute value depending further on whether the curve was fitted by Bekker’s, Janosi’s or Adams’s method; device size had no discernible effect. The relation itself describes an intermediate elastic-plastic stress state that does not exist immediately beneath a rigid wheel in sand, where the soil is fully or partly plastic, and testing it against measured mean shear-to-normal ratio on a 71 cm rigid wheel at 3, 14 and 33 percent slip reproduced one of five qualitative features of the measured curves [71]. Direct measurement of that ratio by pressure transducers set in the wheel rim was abandoned because scatter exceeded 50 percent of the mean, and the substituted mean-ratio method carries plus or minus 25 percent [71].

The shear strength that closes these relations is apparatus-dependent in the same way: four methods run on one condition of the LRV qualification simulant spread the friction angle over 19 degrees, and the vacuum triaxial value alone fell from 38.5 degrees at 0.85 psi normal stress to 37.0 degrees at 1.65 psi [72]. The full comparison is in regolith simulants.

Drawbar pull, the net tractive force available for climbing or towing, is the developed thrust less the motion resistance from compaction and bulldozing [16]. It is measured as a coefficient, longitudinal traction force divided by normal load, against slip, and paired with a sinkage-versus-slip curve; the two together define the performance envelope of a wheel [20]. Standardized procedures for that measurement, including the zero-slip definition that determines where the curve starts, are documented for off-road vehicle testing [11]. Peak drawbar pull and peak tractive efficiency occur at different operating points: in NASA Glenn testing, 60 percent travel reduction returns a net tractive force of 27 percent of vehicle weight, while peak tractive efficiency occurs at 9 percent travel reduction [11].

Tractive efficiency against travel reduction [11]. Efficiency peaks near 0.058 at about 9 percent travel reduction and falls below 0.01 by 83 percent travel reduction. Public domain (NASA / US government work).

Slip and slip-sinkage

Wheel slip is the fractional difference between wheel circumferential speed and actual forward speed. Thrust increases with slip up to a maximum, then falls.

Drawbar pull coefficient and sinkage against wheel slip for a single wheel in dry granular terrain [20]. Normalized traction rises to about 0.19 by 30 percent slip and is then flat to 70 percent, while sinkage climbs monotonically past 50 mm. Public domain (NASA / US government work).

For the LRV wire mesh wheel the 20 percent slip point was fixed empirically as the operating ceiling: the point of maximum drawbar pull without excessive power consumption [22].

The critical behavior on loose regolith is slip-sinkage: at high slip a rotating wheel excavates material from beneath and behind itself, so sinkage increases with continued rotation. Increased sinkage raises motion resistance, which raises slip further. The loop is positive and terminates in immobilization. This is why the gradeability consequence is expressed in slip rather than in torque. Meeting the LRV 25 degree slope-climbing requirement demanded 59.5 percent slip on average across the four simulant conditions tested, and 72.0, 79.5 and 84.5 percent at one, two and three standard deviations of an approximately 500-point slope-versus-slip fit [22]. At those values, 0.5 km of progress up a 25 degree slope corresponds to as much as 3.2 km of wheel travel. The design was accepted on the basis of a thermal analysis showing the drive motors could sustain the torque for the extra cycles.

Wheel performance is not a single curve. Single-wheel testing of the Boeing LRV wheel in lunar soil simulant separated the effects of translational speed, wheel load and soil condition on the drawbar pull and torque coefficients [23], so a drawbar pull figure quoted without its load and soil condition is not usable.

Apollo 15 measured the benign end of this behavior. The navigation system was calibrated with a constant 2.3 percent slip bias, and the associated average wheel sinkage was 1.25 cm, with general sinkage ranging from imperceptible to 5 to 7 cm and the highest values on the soft rims of small fresh craters [15]. Traverse closure error was under 200 m per EVA, and the computed median slip at zero slope was 2.1 percent. Slopes encountered did not exceed 12 degrees [22]. The single observed excursion into the slip-sinkage regime was a wheel spinout at Station 8 near the ALSEP site, which buried the wheels about 13 cm, to the lower part of the rim [15].

Heading matters as well as gradient. On 15 degree slopes, lower angles of attack reduce wheel slip and sinkage while direct uphill driving maximizes the efficiency of upward progress, so the two objectives trade against each other [24].

Vacuum

Interstitial gas contributes to apparent cohesion in terrestrial soils, and its absence changes pressure-sinkage behavior, so simulant testing at ambient pressure does not reproduce lunar response without correction. Isolating the effect requires vacuum testing, which is dominated by the off-gassing transient rather than by the final pressure: simulant disturbance from escaping gas occurs mainly between 2.5 and 10 Torr, the regime in which the gas mean free path is comparable to the interstitial voids, and no disturbance at all has been observed below 2.5 Torr regardless of pump rate [25]. Reaching 10⁻⁵ Torr takes about a week because the soil continues to off-gas. Soil bin preparation and strength characterization at vacuum are correspondingly slow, and cone penetrometry is the practical instrument, returning cohesion, friction angle, bulk density and shear modulus in one measurement [25][26].

Reaction forces and low gravity

Traction available to any vehicle is bounded by the product of normal force and an effective friction coefficient. Normal force scales with weight, which on the Moon is one sixth of terrestrial for the same mass.

The standard terrestrial workaround, building a test article at one sixth of flight mass so that its weight matches lunar conditions, is not a valid substitute. Reducing the mass places a light vehicle on granular material still governed by terrestrial gravity, and yields overoptimistic performance for the nominal vehicle in lower gravity [27]. VIPER was tested this way: its design mass grew from about 440 kg to about 520 kg, and the MGRU3 test articles were built at 73 kg and 88 kg accordingly. Physics-based simulation of the same cases [27] found that the nominal-mass vehicle produced the same slip versus slope curves under both gravities, and that matching the scaled power versus slip relation requires the Earth-test wheel angular velocity to be about √6 times the lunar value, 0.8 rad/s against 0.33 rad/s, a granular scaling law rather than a mass offset. Continuum and discrete-element methods can now resolve these cases directly, at the cost of running 30 to 150 and 3000 to 15 000 times slower than real time respectively, against under real time for a semi-empirical Bekker-Wong model [20]. The rigid wheel case in dry granular media has been reduced to a continuum plasticity formulation that reproduces drawbar pull and sinkage without wheel-specific fitting [28].

For excavation the gravity problem is disqualifying rather than merely awkward. Digging reaction forces do not scale down with gravity, since they are set by soil strength, while the weight available to resist them does. Percussive excavation reduces the required steady force by substituting impulse for reaction mass: impact energies of 13.6 to 30.5 J at 0 to 700 blows per minute cut the average maximum excavation load by up to about 50 percent on a 0.76 m bucket, and a much smaller Honeybee Robotics scoop reached about 70 percent with a 2.5 J spring at 1000 blows per minute [29]. The counter-rotating bucket drum arrangement used by RASSOR and the ISRU Pilot Excavator opposes one drum’s reaction against the other’s, reacting excavation forces through the structure rather than through traction. Measured excavation forces in lightly compacted GRC-3B bound the reaction the structure must carry, at a load cell range of 1334 N horizontal and 3892 N vertical for a 21.6 cm bucket at 10 cm depth and 5 cm/s [30].

Simulants

Simulants are matched to a target property, not to lunar regolith in general. GRC-1 and GRC-3 were developed to reproduce cone index gradients for mobility testing at one sixth mass, GRC-3b-DST adds a 0.7 percent by weight dust-suppressing additive to cut respirable silica, which leaves the particle size distribution and the well-graded silty sand classification unchanged but introduces cohesion the base material lacked, so ASTM D4253 no longer applies and maximum density has to be set by modified Proctor test; unmodified GRC-3b has minimum and maximum densities of 1581 and 1972 kg/m3 [31], and simulants are graded against a lunar reference material by Figure of Merit algorithms that score mineralogy, chemistry, particle size, particle geometry, density, shear strength and magnetic susceptibility separately, each on a 0 to 100 scale [12]. The particle-geometry target is measured: three-dimensional shapes of Apollo and Luna regolith samples determined by X-ray microtomography identify mechanical abrasion during impact gardening as one of the two processes that set them [42]. That per-property grading is the point: no simulant scores well on everything. Producers have improved their reproduction of impact-shattered crystalline material, but the geometry of glass particles, and of agglutinate-like particles in particular, has not routinely reached satisfactory results, and terrestrial feedstocks carry hydrated and weathered minerals, quartz and carbonates that lunar regolith does not have. Simulant results bound rather than reproduce lunar behavior [12]. Bulk density and friction angle vary enough within a single simulant to change the answer: GRC-1 preparations at 1660, 1730 and 1760 kg/m³ give internal friction angles of 33.4, 37.2 and 38.4 degrees, and GRC-3 at 1734 kg/m³ gives 42.0 degrees [27].

Terrain and geomorphology

A slope is defined only against a base length, and the base length is the part usually dropped. The pre-Apollo design model gives a mean slope of 4.5 degrees with a standard deviation of 1.2 degrees at the Sinus Medii mare site over a base length of about 8.5 m, the distance between the lunar module footpads, derived from about 50 cumulative slope frequency distributions [73]. Change the baseline and the number changes, so a requirement of the form “slopes under N degrees” carries no content until it names the length over which N is measured, which for a rover is its own wheelbase rather than a lander footprint. That model is remote sensing plus a handful of unmanned lander points [73], issued as a guide rather than a requirement, and its photoclinometry is stated to be valid only above about 1 m, so it does not support slopes at wheel-scale baselines at all. Crater and block hazards in the same document are cumulative frequency distributions per unit area, N = K D^n, with n about -3 on a young production surface and about -2 on a saturated one, not single worst cases.

Crater density accumulates with surface age and follows a production function until saturation. Craters below roughly 100 m diameter lie in the equilibrium part of the distribution, where formation and destruction balance, and most are reduced to less than 50 percent of initial depth (DSNE §3.4.1.1) [1]. Surfaces with appreciable topographic slope retain fewer craters, because downslope transport degrades them faster. Sub-kilometer crater morphology, after Basilevsky (1976) as given in DSNE Table 3.4.1.2-2 [1]:

Class	Population fraction	Depth/diameter	Max wall slope
Freshest, blocky ejecta, optically immature	0.5%	0.12 to 0.2+	35°+
Steep slopes, blocks common	2.5%
Moderate slopes, blocks mostly on rim	17%
Gentle slopes, rim mostly eroded	30%
Very gentle slopes, no rim	50%	0.19 to 0.22 (degraded)

Half the sub-kilometer crater population is degraded to the point of having no discernible rim, and because small craters degrade rapidly, most slope hazards at candidate landing sites derive from large craters rather than small ones [1].

Polar terrain statistics are established differently, because the hazards lie in shadow. Machine-learning denoising of LRO narrow angle camera frames resolves craters of 4 to 320 m diameter inside PSRs, average 18 m, at spatial densities from 0.1 to 68 craters per km² with an average of 15, and resolves boulders down to about 3 m [7]. Shackleton crater, the reference south polar case, is 21 km across and 4.1 ± 0.05 km deep, with an average wall slope of 30.5 degrees and a maximum of 35 degrees, and a depth-to-diameter ratio of 0.195 ± 0.025 [32]. Its floor is less reflective at 1064 nm than its walls, 0.43 ± 0.02 against 0.46 ± 0.03 [32]. Wall slopes of that magnitude are beyond any wheeled vehicle, so PSR access is a floor-entry problem via the shallower approaches, not a wall-descent problem. Coupled illumination and thermal modeling constrained by LOLA topography is the method by which shadowed floor conditions are predicted before arrival, and it shows that shadowing persists well below the scale of the mapped PSRs: roughly 10 to 20 percent of the total cold trap area for water ice, about 40,000 km² in all, sits in micro cold traps on scales from 1 km down to about 1 cm [33]. ShadowCam images PSR interiors directly at 1.7 m/pixel, using only light reflected from nearby topographic highs. It is over 200 times more sensitive than the LRO Camera Narrow Angle Camera and saturates on sunlit terrain, so it is a shadow-only instrument; block identification and trafficability assessment are the two objectives that set its pixel scale [34]. Normal albedo at 1064 nm from LOLA supplies the reflectance baseline for interpreting those images: about 0.3 for highlands, 0.15 for the low-albedo mode within 30 degrees of the equator, and about 0.36 in northern PSRs [35].

Plasma and surface charging

The near-surface plasma environment is not electrically neutral, owing to surface potentials and a photoelectron population. The non-neutral region is typically 0.5 to 1 m thick (Poppe and Horanyi, 2010, cited in DSNE §3.4.3) but can extend to 100 m within the terrestrial plasma sheet, and on the nightside to kilometers [1].

In full sunlight, the surface and objects on it hold low positive potentials, because photoemission dominates plasma currents. In shadow, and possibly in sunlight during plasma sheet crossings, negative potentials develop whose magnitude depends strongly on material properties [1]. The terminator separates these regimes, so a vehicle crossing it transitions between charging states.

Charging is not solely a plasma effect. Triboelectric transfer between insulating surfaces and soil simulant charges both, with sign and magnitude set by the material pair rather than by the plasma environment [36]. Whether that charge is a hazard depends on how fast it can bleed away, and the dissipation path is the ambient plasma. Inside a polar crater the plasma flux is so reduced that charge accumulated by contact electrification is not fully dissipated between steps of a walking astronaut, and a modeled conducting suit and boot on a 2 second cadence reaches below -2 kV in about 30 seconds of movement, while the same object on the crater topside dissipates each step’s charge before the next [37]. Electrostatic charging is the mechanism behind dust adhesion: charged grains adhere to surfaces including optics, radiators, and seals, and the adhesion is not removed by mechanical means alone [38].

Dust

Dust is the environment’s principal wear mechanism. Grains are angular, comminuted and, in the fines, chemically reactive.

Optical and thermal degradation is disproportionate to coverage. Eleven percent areal dust coverage doubles the solar absorptance of a thermal control surface, and fine dust below 34 µm at only 4.5 percent coverage raised measured absorptance to 0.245, of which a nitrogen jet recovered 2 percent, to 0.239 [38]. The size dependence of removal is the governing fact: particles above 50 µm are easily removed, while particles below 2 µm dominate residual contamination and resist mechanical cleaning. The consequence was observed on Apollo 12, whose magnetometer electronics ran 68 °F hot because of dust on its radiator [38].

Abrasion is likewise disproportionate to exposure. Apollo 12 suits showed more abrasion after about 8 hours on the surface than after more than 100 hours of training wear, and the Apollo 12 suit pressure leak rate rose from 0.15 psi/min after the first EVA to 0.25 psi/min after the second, against a 0.30 psi/min limit [38]. Abrasion rate is grain-size dependent, and testing candidate fabrics against simulant fractions of different mean size is how that dependence is quantified: LHS-1D at 7 µm mean behaves differently from OB-1A at 72 µm mean [39]. Dust exposure behavior is qualified against simulants using a defined deposition methodology [41].

The respirable fraction is also a crew health hazard, and the toxicity risk posed by the fines is assessed separately from their mechanical effects [40].

Ionizing radiation

The surface receives galactic cosmic rays and solar particle events without atmospheric attenuation, shielded only by the solid angle subtended by the Moon itself [1].

Secondary albedo neutrons are produced when energetic particles interact with regolith. These are primarily a crew dose concern but must also be accounted for in avionics design because of displacement damage in semiconductors (DSNE §3.4.7.3) [1]. Albedo neutron flux is anticorrelated with solar activity, since GCR flux is highest at solar minimum; design should use the solar minimum flux.

Meteoroid and ejecta environment

The surface is exposed to primary meteoroid flux and to secondary ejecta raised by nearby impacts (DSNE §3.4.8) [1]. Secondary impact of that material on optical and thermal surfaces produces the same absorptance degradation as deposited dust [38].

Neutral atmosphere

The neutral atmosphere is tenuous enough that DSNE states it is not anticipated to affect the design or operation of lunar surface systems (§3.4.10) [1]. It is not tenuous enough to be irrelevant to soil mechanics testing, where the residual gas pressure of a terrestrial chamber is the dominant experimental variable [25].