Introduction: from exploration to habitation

Lunar exploration is evolving from short-term scientific missions to sustained surface residency and resource utilization. In contrast to the brief Apollo-era landings, next-generation programs, such as China’s International Lunar Research Station1 and NASA’s Artemis Program2 are designed to support continuous human presence on the lunar surface for weeks or even months. However, prolonged surface exposure poses significant risks, including intense cosmic radiation, meteoroid impacts, and extreme thermal cycling with diurnal temperature fluctuations exceeding 300 °C3. The Moon’s geological structures offer a promising solution: subsurface lava tubes and impact-generated cavities can serve as natural shelters. Thick basaltic overburden provides effective shielding against ionizing radiation and maintains thermally stable interiors, minimizing temperature variation between day and night. Leveraging these natural caves could significantly reduce construction demands while enhancing crew safety and mission resilience4. Indeed, analyses of lunar orbital radar data have suggested that large empty lava tubes exist in the maria, with roofs tens of meters thick that could protect human habitats5. We argue that identifying and characterizing these hidden refuges is now a priority for both science and mission planning.

Ground-penetrating radar (GPR) has emerged as a key tool to probe the Moon’s shallow subsurface for structure and voids4,6,7. By sending electromagnetic pulses into the ground and recording echoes, radar can reveal buried layers and cavities without excavation. Recent advances in planetary radar techniques have greatly improved our view of the subsurface on the Moon and Mars. In fact, radar sounding is increasingly recognized as indispensable for planetary exploration: over the past decade, multiple rovers and landers including China’s Chang’E-3, -4,-5, and -6 missions, the Zhurong rover on Mars (Tianwen-1), and NASA’s Perseverance, have carried in-situ GPR instruments8,9,10,11,12,13,14. Ground-based radar systems may potentially be employed in the future for the detection of subsurface cavities on the Moon15,16. These radar instruments have been employed to characterize subsurface stratigraphy and to investigate the presence of water and other indicators of potential habitability on extraterrestrial bodies. In this Perspective, we highlight how such radar investigations are laying the groundwork for a habitable Moon. We discuss how the discovery of subsurface voids by China’s lunar rover radars, combined with new analytical methods (including a “blind inversion” approach to interpret radar data), opens the door to global mapping of lunar caves. Ultimately, finding the Moon’s underground cavities and assessing their stability and environment will inform where we establish our first long-term lunar homes.

However, since the end of the orbital radar mission aboard Japan’s Kaguya mission in 2009, there has been no orbital radar currently in lunar orbit capable of effectively penetrating and detecting the shallow subsurface. Therefore, we call for the urgent deployment of a new generation of global-scale orbital penetrating radar missions to comprehensively map the lunar subsurface structure. Identifying these subsurface cavities and assessing their stability will directly determine the site selection of future human lunar bases.

Subsurface cavities as natural habitats

Lunar subsurface cavities, whether ancient lava tubes or cavities formed within impact ejecta, are widely regarded as promising candidates for human habitation. Their advantages are well substantiated: overlying rock layers several meters thick provide effective shielding against cosmic radiation and meteoroid impacts, while also maintaining a thermally stable internal environment17. Lava tubes, in particular tubular structures created by solidified basaltic lava flows, can extend for several kilometers and reach widths of up to 100 m. Evidence from gravity anomaly data and radar reflectivity indicates the presence of large subsurface caves in regions such as the Marius Hills, with some connected to the surface through collapse features known as “skylight”5,18. These natural cavities have sufficient volume to house complete habitat modules and offer inherent structural protection, reducing reliance on artificial shielding. Additionally, as geological activity within the Moon has gradually waned, both the scale of late-stage volcanic eruptions and the volume of erupted lava have significantly decreased. For instance, radar observations from Chang’E-4 reveal a basaltic lava flow layer ~3 m thick at a depth of ~27 m beneath the surface in the landing region19. This type of low-energy, late-stage volcanic activity has the potential to form small-scale lava tubes in the shallow lunar subsurface, with diameters ranging from several meters to over 10 m.

Besides lava tubes, recent findings indicate that impact processes have also created plentiful voids in the lunar crust. When meteoroids blast the surface, they churn and eject rocks chaotically, sometimes leaving empty pockets amid the jumbled debris. Using high-frequency radar on China’s Chang’E-3 mission (the Yutu rover), Ding et al.7 discovered a distinct zone of low dielectric permittivity at a shallow depth, consistent with a subsurface cavity ~3.1 m height within the Ziwei crater eject. This observation marked the first in-situ detection of subsurface cavities on the lunar surface. Although limited in spatial extent, this finding demonstrates that void structures are widespread in the shallow lunar subsurface and confirms the capability of radar technology to detect such features. More importantly, it suggests that the shallow regolith may be pervaded by similar voids and cavities, ranging in size from several meters to tens of meters. We contend that systematically searching for such cavities both lava tubes and other voids is a prudent step toward identifying safe locales for long-term lunar habitation.

Global radar sounding as the next leap

To move from chance discoveries to a comprehensive inventory of lunar caves, we advocate for global radar sounding of the Moon. A dedicated orbital penetrating radar mission could map subsurface structures across vast areas, far beyond the reach of individual rovers. The concept is analogous to the GPR orbiting Mars (MARSIS and SHARAD), which have illuminated ice deposits and layering beneath the Martian surface20,21. For the Moon, a low-frequency radar sounder orbiting at low altitude could detect the telltale reflections from void spaces essentially “seeing” the roof and floor of a lava tube or large cavity. Carrer et al. have already proposed the concept of an orbiting GPR designed to conduct large-scale mapping of lunar mare regions from orbital altitudes ranging from tens to hundreds of kilometers22. Nevertheless, the detection, verification, and assessment of these subsurface cavities remain key challenges in current lunar scientific and engineering endeavors. A coordinated global lunar exploration network, integrating orbital and in situ radar systems, should be established as the primary framework for unveiling the Moon’s subsurface architecture and evaluating its potential for sustained human habitation (as illustrated in Fig. 1).

Fig. 1: Schematic illustration of combined orbital and in-situ radar exploration of shallow subsurface cavities on the Moon.
Fig. 1: Schematic illustration of combined orbital and in-situ radar exploration of shallow subsurface cavities on the Moon.
Full size image

The orbital radar conducts global mapping of the lunar subsurface to identify potential cavity regions, while the rover-mounted radar performs high-resolution near-field imaging to resolve cavity geometry and roof thickness. Together, they provide key support for assessing the habitability of the Moon’s shallow subsurface environment.

China’s contributions: from Chang’E-3 to Chang’E-7

China’s Lunar Exploration Program is at the forefront of revealing lunar subsurface structures through radar technology. In 2013, Chang’E-3 became the first mission to carry a Lunar Penetrating Radar (LPR). Its dual-frequency radar system, operating in high and low frequency bands, achieved imaging depths of up to 40 m and several hundred meters, respectively, revealing the stratified structure of lunar mare basalts and regolith layers23,24. Critically, the Chang’E-3 LPR made the first detection of a subsurface cavity, providing a key validation case for radar identification of lunar voids7.

After landing on the far side of the Moon in 2019, Chang’E-4 used its LPR to obtain the first subsurface profile of this region: a fine regolith layer ~12 m thick, underlain by coarse-grained debris, and deeper stratified sequences of multi-stage impact ejecta and volcanic basalts, extending to a total depth of ~300 m25,26.Recent analysis reveals that in the deep potential lava tube zone, the radar signal exhibits a distinct phase inversion accompanied by near-zero electromagnetic loss. These features are consistent with a large subsurface cavity or lava tube ~130 m deep and 26 m wide27, one of the largest ever identified on the Moon.

The Chang’E-5 and Chang’E-6 missions achieved, for the first time, coordinated radar detection and lunar regolith sampling on the near side and far side of the Moon, respectively. Both landers were equipped with a High-Resolution Lunar Regolith Penetrating Radar (LRPR), capable of probing to depths of ~3 m and resolving fine stratification within the regolith10,14. However, due to the radar’s limited penetration depth and the fixed positions of the landers, their ability to detect shallow subsurface caves or lava tubes remains constrained, making direct imaging of such structures challenging.

The upcoming Chang’E-7 mission, scheduled for launch in 2026, will investigate water ice distribution and subsurface architecture in the lunar south polar region28. Comprising an orbiter, lander, rover, and hopping detector, its primary scientific objective is to identify ice deposits within permanently shadowed craters29. The mission’s advanced radar system will enable detection of ice-bearing layers and potential cavity structures. We posit that the discovery of stable voids or frozen-soil cavities at the pole could provide not only natural shelters but also serve as critical “cold traps” for volatile preservation, offering key insights for lunar resource utilization and future base site selection.

It is evident that China’s progressive deployment and enhancement of radar technology across the Chang’E-3 to Chang’E-7 missions, as well as the Tianwen-1 Mars mission12,13,14,30, are establishing a comprehensive shallow subsurface observation framework for the Moon and other planetary bodies. These contributions underscore the importance of radar for the next phase of exploration, the phase focused on making other worlds habitable for humans.

The 20-meter zone of opportunity

Treating shallow subsurface cavities (~20 m) as a primary target for future lunar exploration and base construction is justified by a confluence of geological and engineering safety considerations, as well as by the optimal balance between radar detectability and habitability requirements.

First, from an environmental standpoint, temperatures at depths of 10–20 m below the lunar surface remain stable, ranging between −20 °C and 30 °C, with negligible diurnal variation31. Additionally, ~5 m of lunar regolith can attenuate over 95% of cosmic and solar radiation, while a 20-m overburden provides near-complete shielding32. This depth range thus constitutes a “balanced layer” that simultaneously ensures thermal stability and effective radiation protection.

Second, from an engineering and structural stability perspective, a depth of ~20 m, which is equivalent to roughly 6–7 stories underground, facilitates feasible mechanical access while maintaining sufficient stratigraphic integrity. GRAIL mission data and stress state simulations indicate that a basaltic lava tube roof with a thickness of ≥2 m can remain structurally stable under lunar low-gravity conditions, even without artificial reinforcement, provided the tube spans about 1 km in width33,34.

Third, from a scientific detection viewpoint, the upper 20 m of the lunar subsurface lies precisely within the overlapping penetration depth of orbital radar systems and the high-resolution imaging capability of in-situ radar. This enables a collaborative detection framework that integrates orbital identification with ground-based validation. In polar regions, cavities within this depth may also contain deposits of volatiles or water ice35, positioning them as key targets for studies on in-situ resource utilization and long-term habitability.

Finally, from the perspective of human presence and strategic planning, shallow subsurface cavities represent an ideal shelter that simultaneously satisfies the criteria of accessibility, habitability, and detectability. These cavities are not only amenable to scientific observation and engineering development but also offer a favorable environment for sustained human activity. Critically, they support both global reconnaissance via orbital platforms and targeted verification through surface missions, enabling a scalable and integrated approach to lunar exploration.

From discovery to access: toward habitat assessment of shallow cavities

Discovery of a subsurface cavity is only the beginning; the more critical questions concern its stability, accessibility, and habitability. We propose a radar-based “blind inversion” approach for subsurface voids beneath the lunar surface, designed to reconstruct the geometry and dielectric properties of unseen structures solely from radar echoes, enabling a data-driven reconstruction of otherwise invisible space7. Numerical simulations based on Chang’E-3 radar observations revealed a subsurface cavity beneath the landing site with an arched roof and slight inclination, showing remarkable agreement with the observed echo delays7. This finding demonstrates the power of two-dimensional radar sections to reconstruct subsurface geometry, heralding a methodological transition from signal interpretation to spatial understanding across global lunar radar mapping. However, any radar-derived interpretation remains a hypothesis until validated by ground-truth observations. An ideal exploration pathway should therefore proceed in three progressive stages:

  1. (1)

    Orbital targeting: global mapping of dielectric anomalies and potential voids using orbital radar;

  2. (2)

    Landing-based imaging: in-situ measurement of roof thickness and cavity geometry via multi-frequency, near-field radar mounted on a rover;

  3. (3)

    In-cavity verification: entry of miniature robots through skylights or pit openings equipped with high-frequency radar, LiDAR, and imaging systems to map internal structures.

For small shallow cavities with diameters of only 5–30 m, we propose a directed-access strategy: the rover is equipped with a 10–500 MHz multi-frequency radar for stratigraphic imaging and thickness mapping, complemented by LiDAR to reconstruct entrance geometry and supporting features. For concealed cavities lacking natural skylights, a shallow-drilling channel can be used to create an artificial opening, enabling in-situ imaging and direct verification of the internal space and physical properties.

Challenges of detecting subsurface cavities with orbital radar

Orbital radar remains the only technique capable of probing the lunar subsurface on a global scale, yet its performance is inherently constrained by the physical trade-offs among frequency, penetration depth, and resolution. Low-frequency radars (a few MHz) can penetrate hundreds of meters to several kilometers below the surface but offer poor vertical resolution, whereas high-frequency radars (tens to hundreds of MHz) achieve meter- to centimeter-scale resolution but suffer from severe signal attenuation, limiting their effective probing depth to a few tens of meters. For shallow cavities at depths around 20 m, an optimal radar system must balance these extremes by ensuring sufficient penetration while achieving vertical resolution finer than 20 m.

For example, the Lunar Radar Sounder (LRS) onboard Kaguya operated at ~5 MHz, enabling kilometer-scale penetration and the detection of 100-m-scale lava tubes, but with a vertical resolution of only ~75 m5. In contrast, the Mini-RF radar on NASA’s Lunar Reconnaissance Orbiter operates at GHz frequencies, offering high spatial resolution suitable for surface mapping36, yet incapable of penetrating the subsurface to sense shallow voids. In essence, low-frequency radar “sees deep but unclear,” whereas high-frequency radar “sees clear but shallow”. Multi-frequency or wideband systems, such as the combined use of 5–10 MHz and 50–100 MHz frequency radar system, offer a promising compromise between deep and shallow subsurface sensing, representing a key direction for future radar development.

The lateral resolution of orbital radar further limits its ability to resolve small features. The LRS operated from an altitude of ~100 km with along-track resolutions of 75 to 750 m, which is larger than typical lunar cavity dimensions (30–100 m)37. In such large footprints, cavity echoes are easily averaged out or masked by clutter. Higher orbital altitudes also cause footprint broadening, further degrading signal-to-noise ratios. Consequently, existing orbital radars are already operating near their detection limits for cavities a few tens of meters wide and located at depths of ~20 m. Future systems will require longer synthetic apertures, lower orbital altitudes, and advanced beam-steering optimization to enhance both spatial resolution and clutter suppression.

Most lunar lava tubes lack visible skylights, making radar the only “blind-sensing” tool capable of identifying such voids4. The significant dielectric contrast between basalt (relative permittivity ≈ 6–14) and vacuum (1) generates distinct reflections. However, their detectability is constrained by signal penetration depth, surface clutter, and system sensitivity. The lava tube detected by Kaguya in the Marius Hills region was inferred from an echo-free zone followed by delayed secondary reflections5, but such signatures are only reliable under high signal-to-noise and low-noise conditions.

In summary, orbital radar detection of shallow cavities remains challenged by frequency trade-offs, insufficient resolution, and clutter interference. The next generation of instruments must adopt multi-frequency architectures, enhanced synthetic aperture processing, and robust clutter-suppression algorithms to effectively resolve cavities within 20 m depth. Overcoming these challenges will not only advance our understanding of lunar volcanic evolution but also provide critical guidance for selecting and securing future lunar base sites.

Outlook: lunar subsurface as the next frontier

We now stand at a new threshold in the history of lunar exploration. Future missions will not only unfold on the lunar surface but also extend beneath it. We believe that subsurface cavities represent a key resource for building a habitable Moon. In the coming years, a suite of techniques including orbital and rover-mounted radar, seismic arrays, and muon tomography will work together to unveil the Moon’s hidden interior. Global radar mapping is expected to reveal numerous potential lava tubes and voids, while in-situ investigations at selected skylight pits will provide the first direct evidence of accessible underground space.

By the mid-21st century, humanity may possess a comprehensive Lunar Subsurface Habitat Candidate Catalog, identifying sites that combine natural shielding, structural safety, and potential resources suitable for base construction. At the same time, technological experiments simulating life within lava tubes will begin, laying the groundwork for future habitation on both the Moon and Mars.

Just as ancient humans once sought shelter in caves, humanity’s extraterrestrial settlement may also begin beneath the lunar surface. We believe that systematic global radar mapping and targeted subsurface exploration will not only bring new scientific discoveries but also open a new era of the habitable Moon.