PaperlayerView exact revision on arXiv

Source-linked AI summary

LunarRF: A Terrain-Aware Propagation Engine and Pole-Wide 3GPP Surface Coverage Atlas of the Lunar South Pole

Christopher J. Santorelli, Alexander M. Wyglinski

arXiv:2610.34554v2cs.NIastro-ph.EPastro-ph.IM

TL;DR

Terrain relief and the nearby radio horizon make lunar south-pole coverage a geographic problem, but pole-wide coverage evidence has been limited. LunarRF enumerates candidate mast sites and computes a terrain-aware coverage atlas. The atlas’s coverable fractions are sensitive to a 3–4 dB engine optimism, which shifts reported coverage by 6.7 percentage points at 1.8 GHz and 16.5 at 5.855 GHz.

  • Problem

    Crater rims and massifs obstruct line of sight over tens of kilometers, making local topography rather than transmitter range the governing factor in surface coverage.

    Source: §I · pp. 1–2

  • Method

    LunarRF enumerates candidate mast sites and computes their footprints for a coverability map, serving redundancy counts, and per-candidate coverage data.

    Source: B. Candidate Mast Sites · p. 5

  • Results

    A 3–4 dB atlas optimism shifts coverable fraction by 6.7 percentage points at 1.8 GHz and 16.5 at 5.855 GHz.

    Source: H. Discussion and Limitations · p. 24

  • Takeaways & Limitations

    The coverability map shows where any candidate site can serve, while per-cell redundancy and per-candidate data support site selection by planners or optimizers.

    Source: B. Candidate Mast Sites · p. 5

    The atlas is optimistic by 3–4 dB on visible paths; the resulting coverage correction is larger at higher bands as more coverable cells lie near threshold.

    Source: H. Discussion and Limitations · p. 24

Abstract

from arXiv · show

Artemis surface operations will depend on 3GPP cellular links, yet no public product shows where such links are physically possible across the lunar south pole. We present LunarRF, a GPU system for terrain-aware surface coverage, and the pole-wide atlas produced with it. A footprint engine casts 8192 radial profiles from one of 11,827 candidate masts enumerated by stated rules on the orbit-adjusted LOLA elevation model, takes the viewshed, and evaluates ITU-R P.1812 Delta-Bullington propagation on every profile, continuous through the radio horizon. An order-independent resolver merges the footprints per 20 m cell, making the run shardable and resumable. The atlas covers the disc south of 80{\deg}S at 20 m for four SFCG-recommended surface bands and the IM-2 demonstration band, 1.8 to 25.375 GHz, under a stated link budget, storing per cell the best-server power and identity, redundancy count, line of sight and coverability. The engine is validated against a parabolic-equation solver, ray tracing and analytic solutions. Swapping Deygout for Delta-Bullington, or adding terrain noise at the level of the published height-error map of the DEM, moves the coverable class on under 1% of a mast's disc in the mean. With every candidate active, 90.5% of the disc is coverable at 1.8 GHz, 59% at 5.855 GHz and 10% at 25.375 GHz, or 84, 43 and 5% after the measured 3-4 dB engine optimism. A mast covers about 90% of its visible area within its budget-set range, and no band is terrain-limited. The nine current Artemis regions are 85-99% coverable at 1.8 GHz and 22-85% at 5.855 GHz. The atlas and candidate table are released with a DOI.

I. INTRODUCTION

Lunar surface coverage is governed by terrain and has not been quantified in a public pole-wide product. LunarRF enumerates candidate sites, computes terrain-aware footprints, and merges them into an atlas for screening and later network-design work.

I. INTRODUCTION

  • 61% of permanently shadowed-region area is in line of sight of at least one permitted site, and 59% is reachable at 1.8 GHz.The unserved area is concentrated on inner crater flanks and in small craters without a candidate site in view, making coverage a geographic problem to resolve with a DEM.Source: §I · pp. 1–2
  • No public product quantifies achievable coverage across the polar region, so LunarRF enumerates plausible sites and merges their footprints into pole-wide coverage layers.The layers record coverability, link quality, serving-site identities, and alternative-site counts per 20 m cell for screening and subsequent network design.Source: §I · p. 2
  • LunarRF combines a GPU radial viewshed with ITU-R P.1812 Delta-Bullington propagation and an order-independent merge to produce the pole-wide atlas.The candidate sites are enumerated by stated rules rather than hand selection, and the atlas includes best-server power and identity, redundancy, and coverability at 20 m.Source: §I · p. 2, Figure 1 · p. 3

A. Lunar surface propagation models

  • Earlier work modeled lunar propagation or coverage at individual sites, while no published study provided full-polar analysis with systematic candidate-site enumeration.The closest antecedents include ray-traced and DEM-based site studies and a tower-placement analysis; BICTR captures multipath but omits diffraction, which dominates behind terrain obstructions.Source: A. Lunar surface propagation models · p. 3

B. Terrain-based lunar visibility products

Prior lunar visibility products address Sun or Earth visibility, while LunarRF evaluates visibility and propagation from surface base stations. Its architecture separates per-mast footprint computation from pole-wide accumulation and uses orbit-adjusted LOLA terrain.

B. Terrain-based lunar visibility products

C. Ray tracing

  • Sionna RT cross-checks line of sight and reflection, but its wedge diffraction on faceted DEM meshes is not a reference for gentle terrain.Parabolic-equation methods serve as the numerical reference for propagation over irregular terrain with lossy ground.Source: C. Ray tracing · p. 3

D. Terrestrial coverage prediction

  • Delta-Bullington is adapted to lunar radius, absent atmosphere, regolith electrical properties, and a sphere-drop term applied to every profile.Its use at 25.375 GHz extrapolates beyond the recommendation’s stated 6 GHz limit; the retained knife-edge integral and spherical-surface term are frequency-scaled physical constructions.Source: D. Terrestrial coverage prediction · p. 4

III. SYSTEM ARCHITECTURE

  • The system separates one-mast-at-a-time footprint computation from pole-wide accumulation, allowing processing to be sharded, interrupted, and resumed.The footprint engine and resolver use separate state; validation reads the same profiles and accumulators without writing to the store.Source: §III · p. 4

1) Elevation:

  • The atlas uses the orbit-adjusted 20 m LOLA elevation product; the comparison PDS product has orbit-track bumps at about one in ten of its local maxima.Those bumps are 20–257 m high and more than 140 m wide; each candidate site records its elevation difference between the PDS and PGDA products.Source: 1) Elevation · p. 4

2) Permanent shadow:

LunarRF treats permanently shadowed-region floors as receiver locations but excludes them as mast sites. Candidate-site rules, region data, and regolith assumptions define the scope of the resulting coverage analysis.

2) Permanent shadow:

  • Masts are excluded from permanent shadow, but shadowed crater floors remain valid receiver locations when they have line of sight and margin.The rationale given is that a solar-powered lander or mast would not be placed in permanent shadow, while rovers or astronauts may operate there.Source: 2) Permanent shadow · p. 4

3) Landing regions:

  • The study uses nine current Artemis candidate landing regions and three additional 2022 regions with public footprints; five other 2022 regions are excluded from figures and tables.The 2024 list retained five 2022 regions and added four; five 2022 regions without public grids remain as unverified Gazetteer centers.Source: 3) Landing regions · p. 4

4) Regolith:

  • Regolith is modeled with relative permittivity 3.7 and loss tangent 0.01, with 0.1 retained as a high-loss sensitivity case.No small-scale roughness term is applied, and roughness below DEM resolution is among the stated omissions.Source: 4) Regolith · pp. 4–5

B. Candidate Mast Sites

  • The atlas enumerates plausible mast sites rather than choosing placements; their union defines coverability, while per-cell counts measure serving redundancy.The per-candidate table lets a planner or optimizer select a subset of sites.Source: B. Candidate Mast Sites · p. 5

1) Rules common to both classes:

  • Candidate sites are extracted in 40 km tiles with a 10 km halo, exclude locations north of 80°S or in permanent shadow, and are thinned to 1 km spacing.Re-extraction at 500 m spacing admits only 7% more sites; candidates within 22 km of verified region footprints are processed first.Source: 1) Rules common to both classes · p. 5

2) Ridge class: where the terrain offers a high point:

LunarRF defines ridge and landing-site candidate classes using terrain-based rules, then examines exclusions and DEM artifacts that affect site selection. Applying the rules to the orbit-adjusted DEM avoids selecting artifact-raised ridge sites.

2) Ridge class: where the terrain offers a high point:

  • 10,499 ridge-class sites meet terrain high-point or crest criteria, with local slope at most 12°; the class asks what a 4 m mast could see from favorable terrain.The 150-per-tile cap binds on no tile; rims and peaks are not screened for landability.Source: 2) Ridge class: where the terrain… · p. 5

3) Landing-site class: where a lander can stand:

  • 1,328 landing-site-class sites lie within verified USGS footprints and meet a 100 m-scale slope limit of 5° after smoothing, with permanent-shadow cells excluded.Mons Mouton Plateau contributes 766 sites, while Peak near Cabeus B contributes four.Source: 3) Landing-site class: where a lander… · p. 5

4) Criteria deliberately not used:

  • Candidate selection excludes illumination, Earth visibility, mast-set optimization, and terrain properties below the 20 m DEM resolution.Overlapping footprints are not optimized or deduplicated, and bearing strength, boulders, and finer-scale roughness cannot be assessed from the available terrain data.Source: 4) Criteria deliberately not used · pp. 5–6

5) DEM artifacts and site selection:

  • 10.0% of ridge candidates selected on the PDS product sat on features 20–257 m higher than in the adjusted product.The final adjusted-product candidate set has a 3.8 m ridge-class RMS height difference; 77 sites (0.7%) are more than 20 m higher in the adjusted product than in the PDS product.Source: 5) DEM artifacts and site selection · pp. 6–7

C. Propagation Engine and Link Budget

The propagation engine uses a stated small-lander link budget and evaluates coverage against the more restrictive link direction. The budget sets a common ray radius from the lowest carrier frequency, while free-space range declines at higher bands.

1) Link budget and thresholds:

  • The base budget uses 30 dBm base-station power, 23 dBm UE power, isotropic antennas, a 4 m mast, and a 1.5 m receiver.It assumes a 20 MHz downlink carrier, 7 dB UE noise figure, 3 dB base-station noise figure, 150 K antenna temperature, and a 0 dB required SNR without fading, implementation, or body-loss margin.Source: 1) Link budget and thresholds · pp. 7–8
  • Coverage is assessed in both directions, using the more restrictive direction; the stated base budget is downlink-limited at −94.4 dBm.The uplink uses reciprocal path loss with UE power 7 dB below base-station power and a −105.2 dBm threshold over a 5 MHz cell-edge allocation.Source: 1) Link budget and thresholds · p. 8
  • The shared ray radius is 22.0 km at 1.8 GHz; free-space ranges at the other bands are 15.4, 10.9, 6.8, and 1.6 km.The five evaluated carriers are 1.8, 2.58, 3.65, 5.855, and 25.375 GHz; the ray radius uses the lowest carrier and is capped at 100 km when applicable.Source: 1) Link budget and thresholds · p. 8

2) Terrain profile and viewshed:

For each mast, LunarRF derives line of sight and terrain profiles from radial DEM rays, then evaluates obstructed receivers with Delta-Bullington propagation. This yields receiver classifications along each ray, including diffraction-served cells beyond geometric line of sight.

2) Terrain profile and viewshed:

  • 8192 azimuthal rays sample each mast’s DEM window, with lunar sphere drop applied before horizon or diffraction tests.A cumulative-maximum pass determines every receiver’s line-of-sight state and the horizon-forming obstruction; the batched computation runs on a GPU.Source: 2) Terrain profile and viewshed · p. 8, Figure 5 · p. 8

3) Obstructed receivers: Delta-Bullington:

  • Delta-Bullington evaluates every receiver on the actual profile and a smooth-Earth profile, with a spherical-Earth diffraction term, and remains continuous through the radio horizon.Deygout, Bullington, and single knife-edge alternatives are implemented; model-choice sensitivity is quantified separately.Source: 3) Obstructed receivers:… · p. 8

4) Visible receivers: the ground reflection:

LunarRF applies Delta-Bullington to visible as well as obstructed receivers, accounting for ground reflection on clear paths. Its stored layers and order-independent resolver consolidate per-mast footprints into a resumable pole-wide atlas.

4) Visible receivers: the ground reflection:

  • On flat ground, Delta-Bullington reproduces the analytic two-ray power to 0.85 dB RMS; adding a separate two-ray term would double-count grazing-incidence reflection.The atlas therefore uses Delta-Bullington on every receiver, visible or not, as its terrain-and-ground model.Source: 4) Visible receivers: the ground… · pp. 8–9

5) Products:

  • For every 20 m cell and carrier, the atlas stores best-server received power and identity, threshold-exceeding candidate count, and the best server’s line-of-sight flag.It also records band-independent line-of-sight counts and nearest-visible-candidate distance; stricter thresholds can be applied to stored power, but stronger budgets require a new run.Source: 5) Products · p. 9

1) Tiling and accumulation:

  • Footprints merge into per-tile accumulators holding best power, identity, threshold count, and best-server line of sight; associative and commutative operations make merging order-independent and exact across shards.Per-tile layers also include the number of visible candidates and distance to the nearest visible candidate.Source: 1) Tiling and accumulation · p. 9

2) Cluster run:

  • The 11,827-candidate run used 16 GPU shards and took 40 GPU hours; a CPU merge produced a 1.7 GB release bundle.The run could resume after interruption, and the merge rejects overlapping candidate sets or differing engine versions.Source: 2) Cluster run · p. 9

3) Statistics and areas:

LunarRF’s validation shows that Delta-Bullington is optimistic on visible paths, while diffraction errors depend on crest shape. The reported statistics use exact 20 m accumulators, not the coarser release mosaics.

3) Statistics and areas:

  • 160 m release mosaics overstate coverable fractions by 2–8 percentage points, so pole-wide statistics are computed on exact 20 m accumulators.Areas and fractions are reported in polar stereographic map space unless stated otherwise.Source: 3) Statistics and areas · p. 9

IV. RESULTS

  • The mast covers 227.9 km2 of its 1526 km2 disc; its footprint includes 226.8 km2 in line of sight and 1.1 km2 by diffraction.The figure shows the horizon obstruction, direct and diffracted paths, and received power against the coverage threshold along one terrain profile.Source: Figure 6 · p. 10

1) Parabolic-equation reference on real profiles:

  • Delta-Bullington differs from the soft-wedge reference by 4.1 dB at 5° faces, 1.4 dB at 7°, −4.3 dB at 10°, and −11.0 dB at 15°.The PE reference is taken as reliable for shadowed paths only through 5° slopes; the soft wedge is a bounding case, not regolith.Source: 1) Parabolic-equation reference on… · p. 11
  • 3.9 dB at 1.8 GHz and 3.3 dB at 5.855 GHz are the visible-path optimism of the atlas; applying those biases changes the coverable class on 6.7 and 13.1 percentage points of the disc.The PE-corrected fractions are recommended where that bias matters; the correction is applied to line-of-sight cells.Source: 1) Parabolic-equation reference on… · p. 11

2) Ray tracing: line of sight and reflection:

The engine’s terrain-propagation results were cross-checked against parabolic-equation and ray-tracing references, while sensitivity tests found limited coverability-class changes from diffraction-model choice and DEM error at the published error-map level.

2) Ray tracing: line of sight and reflection:

  • Ray tracing matches free-space LOS power to 0.003 dB, but its real-terrain result lacks the grazing ground reflection included by Delta-Bullington.That reflection difference amounts to 11–14 % of one mast’s covered area at 1.8 GHz and 1.6–2.5 percentage points of pole-wide coverability.Source: 2) Ray tracing: line of sight and… · pp. 11–12

3) Ray tracing: wedge diffraction on DEM meshes:

  • On DEM meshes, first-order wedge diffraction yields energy in fewer than 6 % of shadowed cells, 41–57 dB below Delta-Bullington where present.The ray tracer reproduces knife-edge and 60°-wedge diffraction on canonical obstacles, but returns no diffracted path for 31° wedges or a Gaussian ridge.Source: 3) Ray tracing: wedge diffraction on… · p. 12

4) Sensitivity to the diffraction model:

  • Swapping Deygout for Delta-Bullington changes 0.55–0.75 % of the 22 km disc at 1.8 GHz and 0.07–0.13 % at 5.855 GHz.Plain Bullington instead changes 1.4 % of the disc and adds 22 % of covered area at 1.8 GHz because it omits the spherical-earth grazing-reflection term.Source: 4) Sensitivity to the diffraction model · p. 12

5) Sensitivity to the DEM:

  • Using PDS rather than orbit-adjusted PGDA terrain flips coverable class on 0.8 % of the disc at 1.8 GHz and 0.2 % at 5.855 GHz on average.Ridge-class masts in the 82°–85°S band lose 12 % of covered area, while landing-site masts change by 3–4 %.Source: 5) Sensitivity to the DEM · p. 12
  • At the published DEM error-map level, under 1 % of cells change coverable class; peak-picked sites lose 2–2.5 % of covered area north of 85°S and gain slightly south of it.At three and ten times the map amplitude, class flips rise to 0.6–1.7 % and 2.3–5.8 %, respectively.Source: 5) Sensitivity to the DEM · pp. 12–13

6) Effective resolution of the DEM:

The DEM’s interpolated track gaps make outer-latitude coverage estimates likely optimistic, while candidate-density tests show limited changes at most bands. The atlas run is internally consistent, and its coverage falls with increasing frequency and toward the pole.

6) Effective resolution of the DEM:

  • The 20 m DEM interpolates between widely spaced LOLA tracks; outer-latitude coverable fractions are likely optimistic rather than bounded.Candidate mast cells are better constrained than surrounding receivers, but unresolved terrain can either hide a receiver or lift a mast.Source: 6) Effective resolution of the DEM · pp. 13–14

7) Candidate density:

  • Halving candidate spacing from 1 km to 500 m changed coverable fractions by 0.2–0.8 points at 1.8 GHz and 0.3–2.9 at 5.855 GHz on three test tiles.Changes were always upward; at 25.375 GHz each added site contributes its own 1.6 km disc.Source: 7) Candidate density · p. 14, Table IV · p. 15

8) Internal consistency of the pole-wide run:

  • Four consistency checks found exact agreement with the earlier atlas layer, matching free-space ranges, expected band-to-band power steps, and no visible shard seams or doubled count bands.The checks support consistency of the merged run; the paper treats physics validation separately.Source: 8) Internal consistency of the… · pp. 14–15

B. The whole disc

  • 90.5% of the disc is coverable at 1.8 GHz, 59.1% at 5.855 GHz, and 10.0% at 25.375 GHz; PE-corrected fractions are 83.8%, 42.6%, and 5.4%.All 11,827 candidates are active; the correction re-thresholds visible cells by the measured engine bias.Source: B. The whole disc · pp. 15–16, Table V · p. 16
  • Within its free-space range, a mast covers about 90% of its visible area whatever the band, while that range shrinks from 22.0 km to 1.6 km.Source: B. The whole disc · p. 16
  • At 1.8 GHz, 59.1% of the permanently shadowed area is coverable, but 69% of the uncoverable area lies outside permanent shadow.The result is conditional on the siting rules: no candidate stands in permanent shadow or on ground steeper than 12°.Source: B. The whole disc · p. 15, Table VI · p. 17
  • Coverage declines poleward at 1.8 GHz from 92.9% to 84.5%, while mean serving candidates fall from 10.3 at 1.8 GHz to 1.2 at 25.375 GHz.Coverage also falls toward the pole at 5.855 GHz, from 66.2% to 49.8%.Source: B. The whole disc · pp. 17–18, Figure 11 · p. 18, Figure 10 · p. 18
  • Only 2.7% of 1.8 GHz coverable cells are served without line of sight, compared with 0.9% at 5.855 GHz and 0.3% at 25.375 GHz.Diffraction changes the coverage class for only a few percent of the disc; geometry and budget account for the rest.Source: B. The whole disc · pp. 15–16, Table VII · p. 19

C. Per-mast footprints

Individual mast footprints vary by candidate class, but each class covers most of the terrain visible within its band-specific free-space range.

C. Per-mast footprints

  • At 1.8 GHz, median covered area is 228 km2 for ridge-class masts and 56 km2 for landing-site masts; at 5.855 GHz, the respective medians are 39 km2 and 19 km2.Median geometric line-of-sight areas are 269 km2 and 62 km2, respectively.Source: C. Per-mast footprints · p. 19
  • Within each band’s free-space range, ridge-class masts cover a median 89% of visible terrain and landing-site masts 93–94%.The ridge-class 10th–90th percentiles are 75–96%; landing-site values are 76–113%.Source: C. Per-mast footprints · p. 19
  • Within the 2 km EVA disc around a landing-site mast, median coverage is 49% at 1.8 GHz and 35% at 5.855 GHz.The paper identifies taller lander-mounted equipment as the strongest argument for EVA planning.Source: C. Per-mast footprints · p. 19

D. The Artemis candidate landing regions

Current Artemis candidate regions have high coverability at 1.8 GHz, while their higher-band coverage varies substantially with terrain and nearby high ground.

D. The Artemis candidate landing regions

  • At higher bands, footprint geometry relative to nearby high ground separates plains overlooked by rim sites from high ground with shadowed flanks.Peak near Cabeus B is 85.4% coverable at 1.8 GHz but 1.9% at 25.375 GHz; its steep flanks are shadowed or beyond higher-band range.Source: D. The Artemis candidate landing regions · pp. 20, 22, Figure 15 · p. 22
  • At 1.8 GHz, mean serving candidates per coverable cell range from 7.2 at Malapert Massif to 31.9 at Mons Mouton; every region has 1.2–2.1 at 25.375 GHz.The count scales with candidate density and is a relative measure between regions.Source: D. The Artemis candidate landing regions · p. 22
  • At 1.8 GHz, every current region is at least 85.4% coverable; Slater Plain reaches 98.5% and Nobile Rim 2 reaches 97.7%.The 2022-only Faustini Rim A is the least coverable of the twelve regions at 82.1%.Source: D. The Artemis candidate landing regions · p. 19, Table XI · p. 22, Figure 16 · p. 23
  • Across regions, coverable fraction spans 22.4% (Peak near Cabeus B) to 85.3% (Slater Plain) at 5.855 GHz, and 1.9% to 45.0% at 25.375 GHz.At 5.855 GHz, coverable fraction equals the footprint fraction in line of sight of a candidate within the 6.8 km free-space range to within two percentage points in each of twelve regions.Source: D. The Artemis candidate landing regions · pp. 19–20

E. Throughput implications

The atlas reports SNR rather than throughput, so the paper uses Shannon capacity as a scale and estimates practical rates from that bound.

E. Throughput implications

  • The atlas gives SNR, not throughput; median coverable-cell Shannon capacity for a 20 MHz carrier is 85 Mbit/s at 1.8 GHz, 47 Mbit/s at 5.855 GHz, and 35 Mbit/s at 25.375 GHz.These are bounds, not measured practical rates.Source: E. Throughput implications · p. 23
  • Practical LTE/NR radios are estimated at 60–75% of the Shannon bound, or 50–75 Mbit/s near a mast and 5–10 Mbit/s at the coverage edge.Rates scale with carrier width and MIMO order; n78 permits 100 MHz carriers.Source: E. Throughput implications · p. 23
  • The uplink, which carries EVA video, has the smaller share; FDD uplink rates use the budget’s 5 MHz cell-edge allocation rather than a TDD time split.The TDD split applies to the listed TDD bands, while Band 3 and n7 are FDD.Source: E. Throughput implications · p. 23

F. Power-only control: terrain or budget?

The power-only control finds that coverage is budget-limited rather than terrain-limited at every band. A reference lander scenario uses a taller tower but is uplink-limited; the discussion qualifies the atlas with its sensitivity, validation limits, assumptions, and scope.

F. Power-only control: terrain or budget?

H. Discussion and Limitations

  • The atlas is an envelope of what permitted sites could serve, not a deployed network; turning it into a design requires a separate placement and interference problem.The per-candidate table and best-server identity layer are inputs to that design task.Source: H. Discussion and Limitations · p. 24

G. A reference budget

  • The reference scenario is uplink-limited at 16.4 km and covers less of every footprint than the base atlas at 1.8 GHz.The comparison uses landable-site towers only in the reference scenario, versus all candidates in the base atlas; the taller tower approaches base-atlas fractions at higher bands in several regions.Source: G. A reference budget · p. 24, Table XIII · p. 25

H. Discussion and Limitations

  • At 1.8–2.58 GHz, diffraction, reflection, and DEM choices barely change coverable fractions, while visible-path optimism shifts 6.7 percentage points at 1.8 GHz and 16.5 at 5.855 GHz.The atlas is 3–4 dB optimistic against the parabolic-equation reference on visible paths; the released line-of-sight flag supports a read-time correction.Source: H. Discussion and Limitations · p. 25
  • Validation lacks measurements on obstructed lunar-like paths, and the parabolic-equation comparison checks one model against another on the same DEM.The PE terrain treatment is trusted only to 5° segment slopes; on steeper crests the reference is the exact wedge solution.Source: H. Discussion and Limitations · p. 25
  • The model assumes antenna temperature, receiver noise figures, lander mounting heights, and regolith properties, and omits multipath, interference, roughness below DEM cells, and time dependence.Five 2022 regions without public footprints are excluded rather than reconstructed.Source: H. Discussion and Limitations · p. 25
  • Local-maximum selection on the unadjusted PDS 20 m DEM picks orbit-track artifacts at about one ridge site in ten.The orbit-adjusted product does not select them; per-site height differences can filter artifacts where the PDS product must be used.Source: H. Discussion and Limitations · p. 25
  • Next steps include matching the reference scenario’s uplink, coupling coverage to illumination and Earth visibility, and running 5 m regional analyses and measurement validation.The reference scenario leaves its 20 dB of extra base-station power unused because it is uplink-limited.Source: H. Discussion and Limitations · p. 25

V. CONCLUSION AND FUTURE WORK

LunarRF produces a pole-wide potential-coverage atlas from 11,827 candidate sites, with an envelope of possible service rather than a deployed network. The work releases the atlas and identifies accuracy and terrain-data boundaries, with further validation and analysis left for future work.

V. CONCLUSION AND FUTURE WORK

  • LunarRF’s atlas covers the 80°–90°S disc at 20 m from 11,827 candidate sites; 90.5% is coverable at 1.8 GHz, 59.1% at 5.855 GHz, and 10.0% at 25.375 GHz.The atlas is a potential-coverage envelope, not a deployed network; the engine is 3.3–3.9 dB optimistic on visible paths against a parabolic-equation reference.Source: §V · p. 26
  • Future work includes set-cover mast placement, 5 m regional runs, coupling to illumination and Earth-visibility maps, and measurement validation on obstructed lunar-analog paths.Source: §V · p. 26

DATA AVAILABILITY

  • The atlas bundle, exact-statistics tables, and validation records are released under CC BY 4.0; full 20 m tiles are available from the authors on request.The release is about 3 GB; the full tiles are 60 GB.Source: DATA AVAILABILITY · p. 26

Related papers

Loading 2610.34554v2…