“Spiders” on the Moon: Morphological Evidence for Geologically Recent Regolith Drainage into Subsurface Voids

On the Moon, the surface morphology at the scale of meters and tens of meters is typically smooth and subdued due to regolith gardening. Sharp, “crisp,” meter-scale morphologic features are observed only where the regolith is either thin or recently disturbed. Such crisp morphologies are typically created by geologically recent meteoritic impacts of different scales. The prominent exception is so-called irregular mare patches (IMPs), rare small features of debated origin. We report here on the discovery of previously unknown crisp immature morphological features (named “spiders” due to their central circular region and radiating “legs”) not related to impacts and even more rare. The spiders are meters-deep depressions with near-radial chutes open toward the center which make an incipient dendritic pattern 50–80 m in diameter. All spiders found thus far occur in clusters in the same region in Mare Tranquillitatis in the immediate proximity to small IMPs. We interpret spiders as the result of an energetic granular flow of the regolith draining into shallow subsurface voids following the sudden collapse of the roofs of the voids. Regolith gardening destroys the spiders’ legs rapidly, on a timescale of a million years. If the entrance into the subsurface void remains unclogged, a spider appears to evolve into a pit; otherwise it evolves into a gentle depression and finally disappears. Our interpretation of spiders provides a consistent explanation of all of their features, occurrence settings, and associations.


Introduction
The surface of the Moon is typically covered by a metersthick layer of regolith (e.g., Lucey et al. 2006).The surface morphology at the scale of meters and tens of meters is smooth and subdued, being predominantly controlled by regolith gardening, which acts as a topographic diffusion process (e.g., Soderblom 1970;Fassett & Thomson 2014), eroding convex topographic features and filling concave ones, and thus smoothing the regolith surface.Sharp, "crisp," meter-scale morphologic features are observed only where regolith is absent or very thin (e,g., Basilevsky et al. 1977Basilevsky et al. , 2020)), or where it was recently disturbed.Such crisp regolith morphologies and corresponding rough decameter-scale topography occur predominantly in geologically young impact craters of different sizes and their continuous ejecta (e.g., Kreslavsky et al. 2013).Other than young impact craters, features with no regolith, or very thin or recently disrupted regolith, are extremely rare.The largest class of such features (in terms of area) is the set of rocky slopes and solidified melt ponds in the area approximately antipodal to the Tycho crater (Robinson et al. 2016).These deposits have been interpreted as a result of focusing of Tycho ejecta at the antipodal point (Artemieva 2013; Bandfield et al. 2016) and are therefore also of impact origin.
In addition to impact-related features, crisp morphologies are observed in so-called irregular mare patches (IMPs), rare features found in the lunar maria (e.g., Braden et al. 2014;Qiao et al. 2020a and references therein).The largest IMPs are several kilometers in size, while the smallest identifiable ones are tens of meters.Figure 1 shows a typical example of a small IMP; other examples can be seen in Figures 2 and 3. Stooke (2012) also used the term "meniscus hollows" for these features, which appropriately describes their morphology: they consist of two subunits: bright hummocky lower subunits ("hollows") and dark smooth convex-upward slopes surrounding them and forming the upper subunits ("menisci").The boundary between the subunits is sharp and crisp; it is associated with a very sharp slope break between the rough, but generally horizontal lower subunit and the rather steep slope of the upper subunit.The transition between the upper subunit and mare area surrounding the IMP is gradational (Figure 1); the contact between them cannot be accurately identified.In some of the largest IMPs (Class #1 IMP according to Qiao et al. 2020a) in addition to the "hollow" surroundings, the upper subunit also forms islands, elevated domes surrounded by the lower subunit, such as in the most well-known IMP examples of Ina (e.g., Qiao et al. 2017Qiao et al. , 2019)), Sosigenes (e.g., Qiao et al. 2018), and Cauchy 5 (Qiao et al. 2020b).Reflectance spectra suggest that the relatively higher albedo of the lower subunit is caused by surface immaturity (Grice et al. 2016).Small superposed craters are hard to identify in the hummocky bright lower subunit; the upper subunits of the largest IMPs have measurably sparser populations of superposed small impact craters than the surrounding mare surface, suggesting very young crater retention ages of that subunit (<100 Ma for Ina, Sosigenes, and Cauchy 5; Braden et al. 2014).Braden et al. (2014) cataloged 70 IMPs and their clusters, and Qiao et al. (2020a) added 21 more.
The formation mechanism of IMPs is debated.Braden et al. (2014) suggested that IMPs are extremely young volcanic features with the "hollows" being collapse depressions often observed in terrestrial inflated lava flows.This mechanism has Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.some difficulties explaining the observed features of the IMP upper subunit.First, it is inconsistent with the observed gradational boundary between the upper subunit and IMP surroundings.Flows of segregated impact melt in large young impact craters have sharply defined boundaries and characteristic lobate planforms (e.g., Denevi et al. 2012); it is reasonable to expect the same relationships for lava flows of comparable young age and dimensions.Second, the morphology of small craters superposed on the upper subunit is identical to the morphology of craters of the same size in the surrounding maria, as documented by Basilevsky & Michael (2021), which suggests a thick regolith layer, and thus an ancient age for the upper subunit.In contrast, small craters superposed on young flows of segregated impact melt have distinctively different morphologies, suggesting very thin regolith.Qiao et al. (2019) suggested that IMPs are old volcanic features made of highly vesicular lavas; they envisioned recent sifting of the regolith into shallow subsurface voids as being responsible for the observed immaturity of the lower subunit.This mechanism requires some special conditions (see Wilson & Head 2017) to reconcile the interpreted ancient emplacement age with the young crater retention age; see Qiao et al. (2019) for a detailed discussion.Kreslavsky & Head (2023) argued that IMPs could have formed by the collapse of the roofs of showing the lower unit (L; "hollow"), upper unit (U; "meniscus"), the sharp contact between the lower and upper units (solid line), and gradational contact between the upper unit and surroundings (dashed line).Local equirectangular projection, north is at the top.shallow subsurface voids: the lower subunit is the place of the roof collapse and regolith drainage, while the upper subunit is made of precollapse regolith shaped by topographic diffusion due to regolith gardening.They explained how the apparent young crater retention age of the upper subunit could be an observational artifact.
An additional IMP formation mechanism was suggested by Schultz et al. (2006).They envisioned the removal of regolith by the recent sudden release of gas from the subsurface.This mechanism provides a plausible explanation of IMP morphology.The apparent young crater retention age of the upper subunit could be explained by partial obliteration of small craters due to infill by regolith entrained by the released gas stream.However, the accumulation of a significant amount of gas under pressure in the shallow subsurface, and its sudden release, seem unlikely.In addition, this mechanism would predict a halo of immature regolith around "hollows," which is not observed.
In the process of a reconnaissance analysis of mare surface textures and IMPs in LROC NAC images (Robinson et al. 2010), we discovered a new class of small features in Mare Tranquillitatis (Figure 2) that are crisp, bright (thus apparently immature), and therefore young.We then broadened our search, finding a few more examples in adjacent areas of Mare Tranquillitatis (Figures 3, 4, and 5).We informally refer these new rare features as "spiders," due to their arachnoid-like central circular features and radiating "legs" (also similar to an asterisk; Figures 2 and 3).We first describe spiders (Section 2) and outline their spatial association with IMPs and other features.We then consider a range of candidate interpretations for their formation (Section 3), concluding that they are most likely to have formed by the recent collapse of shallow subsurface voids and the drainage of regolith into the voids.We further discuss this scenario (Section 4) and conclude (Section 5), focusing on questions that still need to be answered.

Morphology
Spiders are small asterisk-shaped depressions (Figures 2 and  3).They are only distinguishable in high-resolution LROC NAC images (∼1-2 m pix −1 sampling).Even in the highestresolution images the details of their morphology are sometimes poorly resolved.The asterisk shape is composed of 5-10 "spider legs," narrow reentrants, or chutes, widening from their outer tips downward, and open toward the depression center.The chutes are a few meters wide and narrower at their tips.Some chutes have short tributaries, thus forming incipient dendritic chute systems.A typical diameter of a spider is about 70 m, and is very consistent among different examples: the measured size variations (50-80 m) are comparable to the accuracy of the size definition.Three spiders, including the most pronounced one (Figure 2), have a single, small (∼5 m in diameter) sharp rimless depression (pit) in their centers.These central pits differ from relatively fresh impact craters of the same size, which have poorly resolved elevated rims.

Location, Geological Settings, and Associations
We identified four sets of spiders in Mare Tranquillitatis: set A at 8°.80 N, 22°.00 E (Figure 2), set B at 8°.85 N, 21°.75 E (Figure 3), set C at 8°.43 N, 21°.91 E, and set D at 7°.90 N, 22°.00 E. Figures 4 and 5 show their locations and the regional context in the images and topographic data.Each set consists of 3-5 spiders; all four sets are in the same region and the distance between the sets is 7-14 km.The region is a typical Late-Imbrian-age mare; its absolute model age of emplacement is ∼3.7 Ga according to Hiesinger et al. (2011), and it is characterized as a high-TiO 2 mare basalt (Qiao et al. 2021, their Figure 1(c)).
Each spider set is located in a gentle depression 10-40 m deep, ∼0.5 km wide, ∼1-2 km long, and elongated in an ∼east-west direction.These depressions are barely resolvable in LOLA topographic data (Zuber et al. 2010;Figure 5(a)), but are resolved in SLDEM2015 global topographic data set (Barker et al. 2016;Figure 5(b)).The origin of these depressions cannot be unambiguously determined from their morphology; however, they appear to be ancient pits of possible volcanic origin, rather than clusters of degraded impact craters or secondary crater chains.
All spiders are in the immediate proximity of small IMPs (within hundreds of meters, Figures 2 and 3).Given the extreme rarity of IMPs on average (e.g., Braden et al. 2014;Qiao et al. 2020a), this clear spatial association suggests that the spider formation may in some manner be related to the formation of small (Class #2 according to Qiao et al. 2020a) IMPs.The albedo of the spiders is as high as the albedo of the brightest (presumably, the most immature) parts of the IMPs.However, spiders differ from IMPs in their morphology: they lack generally flat hummocky floors and the characteristic convex-upward slopes seen in the IMPs.
In addition, there are bright boulder fields in the proximity of all spider sets (Figures 2 and 3).These boulder fields are not associated with ejecta of any young impact craters.Since the lifetime of a boulder on the surface is geologically short (e.g., Basilevsky et al. 2013), such fields are geologically young, which is also consistent with their high albedo caused by their immaturity.Similar boulder fields have been documented by Valantinas & Schultz (2020) in association with some wrinkle ridges and interpreted as a consequence of recent tectonic deformation and movement of regolith, exposing boulders and fresh, immature soil.There are, however, no wrinkle ridges in the immediate proximity to the spider sets.A well-pronounced wrinkle ridge is located a few kilometers to the east from the spiders (Figure 5), and in some of its sections there are also boulder clusters.
In sites A and B, the individual spider locations are aligned in a general east-west direction, tilted slightly toward a WNW direction (∼280°azimuth).The spider-bearing depressions in these two sites are elongated in the same direction; this directional trend has the same azimuth as the line from site A to site B (Figures 5(b  All these directional trends related to the spiders are comparable to a number of other WNW-ESE trends observed in Mare Tranquillitatis: elongation of the depression containing the large Sosigenes IMP structure (see Figure 4 for its location), pit chains along its strike (Qiao et al. 2018, their Figure 2), several linear rilles, elongation of the summit small shield pit crater of the Cauchy 5 IMP (Qiao et al. 2020b, their Figure 3), and elongation of numerous other small shield pit craters (Qiao et al. 2021, their Figures 4 and 6).In a regional context, Mare Tranquillitatis is the location of the highest concentration on the Moon of (1) small shield volcanoes (mare domes) (Qiao et al. 2021 and 14), small mounds in the lunar maria with a surrounding moat, often associated with small shields and IMPs, and interpreted to represent the effusion of magmatic foams from highly vesicular lava flow surfaces during their final emplacement and cooling (e.g., Wilson et al. 2019).
We undertook an analysis of the regions surrounding almost all of the cataloged IMPs (Qiao et al. 2020a) and found no other examples of spiders.Our search, however, cannot be considered exhaustive for the Moon as a whole, and we cannot claim that there are no additional spiders on the Moon.Searches for other examples of spiders elsewhere on the Moon and assessing their associations and host unit ages should be a high research priority.

Age Constraints
The high albedo of the spiders is apparently caused by immature soil, suggesting a young age.The unusual morphological sharpness of the small chutes also indicates their recent formation, because all sharp topographic forms are rapidly smoothed by topographic diffusion due to regolith gardening (e.g., Soderblom 1970;Fassett & Thomson 2014).
We numerically modeled degradation of spider topography as a topographic diffusion process.We numerically solved the linear diffusion equation ¶ where h(x, y, t) is the surface elevation as a function of time t and spatial Cartesian coordinates x and y, K is the topographic diffusivity, and ∇ 2 is the two-dimensional Laplacian in (x, y) space.We modeled regolith drainage into the pit in the spider center (see the next section) by imposing the boundary condition h(0, 0, t) = −10 m at the model center x = y = 0. We also used stationary boundary conditions h = 0 at the edges of the 80 m × 80 m spatial model domain.We created an artificial relief resembling a 10 m deep spider with ∼70 m diameter and ∼5 m wide legs and used it as the initial condition h(x, y, 0). Figure 6 shows an image of the spider with a central pit from Figure 2, which was used as a basis for the initial modeled relief, and four solutions of the diffusion equation that correspond to four progressive degradation states Kt (Fassett & Thomson 2014).Figure 6 shows that the spider becomes poorly recognizable at greater than 2-3 m 2 .The absolute timescale of diffusive degradation of the relief is defined by the topographic diffusivity K. Fassett & Thomson (2014) reported a diffusivity of 5.5 m 2 Ma −1 derived from analysis of a population ∼1 km size craters, and Fassett et al. (2018Fassett et al. ( , 2022) ) noted that the effective diffusivity scales with crater size D as ∼D 0.9 .We assumed an effective diffusivity ∼ 0.5 m 2 Ma −1 , approximately corresponding to D ∼ 70 m.We found that the legs become topographically indistinguishable in a few millions of years (Figure 6); they lose their morphologic sharpness over a 1 Ma timescale.Thus, our modeling shows that it is likely that the sharpest spiders are hundreds of thousands of years old or even younger, and all observed spiders are not more than several millions of years old.

Formation Mechanisms
Spider-like features are known on Mars (the so-called araneiforms; e.g., Kieffer 2007;Portyankina et al. 2017;Mc Keown et al. 2023).Some of them have somewhat similar morphology (central depression and a near-radial dendritic set of channels).However, they are larger, and the dendritic pattern is more developed in the Martian examples.The Martian araneiforms result from soil displacement by a gas flow channelized beneath an impermeable slab of seasonal dry ice (solid CO 2 ).An analogous formation of lunar spiders by a nearsurface gas flow in the regolith would be consistent with the gas discharge mechanism of IMP formation (Schultz et al. 2006).However, in the absence of an impermeable ice slab on the Moon, channelizing of the gas flow in the regolith would require the presence of a stronger, less permeable layer in the upper decimeter above a weaker, more permeable regolith.Such arrangement, if were typical in lunar maria, would be easily detectable by astronauts and in observations with robotic landers, however, it has never been reported (e.g., Lucey et al. 2006).Thus, we do not believe that this "Mars-like" mechanism of spider formation is plausible.
We suggest that spiders were formed by regolith draining into shallow subsurface voids (Figure 7).Collapse of a part of the void roof would cause regolith flow into the void and the formation of a funnel-shaped depression in the regolith.Weak but nonnegligible cohesion of regolith and its fluidization by the funnel-forming granular flow could cause the initiation of flow instabilities, leading to flow channelizing and the formation of the chutes observed as spider legs.When the flow ceases, it leaves a spider structure as we observe it (Figure 7).In this scenario, the spider-forming flow removes the mature regolith veneer and exposes immature regolith, which makes spiders brighter than their more mature surroundings.
Shallow subsurface voids are known to be present on the Moon.High-resolution images show skylights and pits that may be entrances into spacious caves (e.g., Haruyama et al. 2009;Robinson et al. 2012;Wagner & Robinson 2014, 2022).Voids are also known to form in terrestrial volcanoes: when lava drains from lava tubes, magma leaves shallow reservoirs and magma conduits, and where volatiles are exsolved in the latter stages of an eruption and gas bubbles and coalesced gas  slugs are trapped below a solid lava lake floor flow surface.
From a theoretical point of view, the formation of shallow voids is likely to occur due to the degassing of lava during its emplacement, which is linked to flow inflation and/or cooling and volatile release associated with "second boiling" (e.g., Wilson & Head 2017, 2018;Wilson et al. 2019).Similarly, the propagation of dikes from the lunar mantle to the lunar surface not only causes effusive eruptions, but dikes are known to stall in the shallow crust, producing a variety of features, one of which forms linear crater chains from both gas venting and passive collapse into gas-produced voids in the dike tip (Head & Wilson 2017; Figure 8).These multiple mechanisms are consistent with the observation that the spiders are observed inside mare pits that are likely to be of volcanic origin.The pronounced linear trend associated with spider sets A and B (Figure 5) suggests a role of a single dike in the formation of the voids responsible for these two spider sets.The association of spider sets C and D with two other dikes is not excluded.
Lower lunar gravity, low levels of seismicity, and the absence of erosion all favor the preservation of voids for a long time following their formation.Thus, we suggest that the voids related to spider formation are old, contemporary with mareforming volcanism several billion years ago, while the spiders themselves formed recently by partial roof collapse of the voids.
The subsurface void needs to be sufficiently large to accommodate all drained regolith.The missing volume of a spider can be approximately estimated as a half of the volume of a cone of diameter D equal to the spiderʼs diameter (D ≈ 70 m) and a height H equal to the maximum depth of the spider (H ≈ 10 m), that is ∼ (1/8)D 2 H ∼ 6 × 10 3 m 3 .This volume in the void would make a conic pile (Figure 7) with radius r ∼ (1/ 2)(D 2 H cot θ) 1/3 ∼ 20 m, where θ ∼ 35°is the angle of repose and the height r tan(θ) ∼ 14 m.These estimates provide an approximation of the size of the smallest single void necessary for spider formation.
Granular flows are known to produce flow instabilities of different kinds due to their strongly nonlinear rheology (e.g., Aranson & Tsimring 2006, and references therein).We suggest that the initiation of chute formation was caused by such flow instabilities breaking the axial symmetry of the forming funnel, similarly to the formation of fingers in a granular flow (e.g., Aranson & Tsimring 2006); the enhanced fluidization in the fingers then caused enhanced erosion of slightly cohesive regolith and the incision of chutes, which in turn increased slopes and entrained new portions of regolith at the tips of growing chutes.From everyday experience we know that sometimes the formation of chutes occurs in the terrestrial environment during the collapse of decimeters-and metershigh banks of slightly wet (therefore, cohesive) sand.However, such terrestrial chutes are shorter; more accurately, they have smaller length/depth and length/width ratios than the lunar spiders' chutes.We are not aware of any literature analyzing the formation of such chutes in the terrestrial environment.Figure 9 shows a small terrestrial sinkhole that formed in slightly wet sand.The sand surface is shaped as a funnel complicated with various small chute-like features.Of course, this example cannot be considered as a scaled analog of lunar spiders, however, it demonstrates the formation of a funnel with irregularities during drainage of slightly cohesive granular material into a subsurface void.Under lower lunar gravity, granular flow fluidization would be more effective in general, and we thus suggest that the formation of longer chutes on the Moon is reasonable.Additional knowledge of the dynamic properties of lunar regolith flows is needed to understand such a process in more detail.
Our suggested spider-forming scenario readily explains the depth of the spider central depressions: it is approximately equal to the regolith thickness.Bart et al. (2011) have reported a median regolith thickness of 4.4 m for the central part of Mare Tranquillitatis and a wide (∼3 m to ∼7 m) interquartile range of individual thickness measurements.The spiders are located in depressions that accumulated a substantially thicker regolith layer due to topographic diffusion; the ∼5-10 m thickness value seems very reasonable for those locations.In our model, spider diameters are controlled by regolith thickness and the mechanics of chute incision.The very similar settings and proximity of the four spider areas are predicted to imply the thickness and mechanical properties, which appear to account for the similar and consistent spider diameters.
Spider formation under our scenario is a rapid process.The upper limit of its duration can be derived from the velocity of the slowest granular flow that potentially can be self-sustained, which is approximately equal to q l g sin , where l is the characteristic particle size, g is gravity, and θ is the slope angle.This scaling estimate results from a simple consideration of particles sliding downslope one by one, when the next particle moves after the space is vacated by the previous particle.For sand in Earth conditions this velocity is a few centimeters per second; it can be observed in sand avalanches on dune slip faces.The velocity of uphill propagation of the avalanche upper front is on the same order of magnitude (e.g., Aranson & Tsimring 2006).For a coarse fraction of lunar soil (l ∼ 0.1 mm; e.g., Lucey et al. 2006) under lunar gravity this velocity is about 1 cm s −1 .Chute lengthening cannot be slower than this velocity, which, when combined with spider radius, gives an upper boundary for the duration of spider formation of about 1 hr.It is likely that a chute-forming flow capable of incising slightly cohesive regolith is more energetic, and that the chute tip propagates by progressive regolith slumping, in contrast to the grain-by-grain propagation of sand avalanche fronts.
Therefore, a realistic duration of spider formation seems to be substantially shorter, of the order of minutes or tens of minutes.The flow is predicted to stop when the slopes of the actively forming funnel and chutes becomes sufficiently gentle; alternatively, the flow might also stop soon after the void is filled and/or the sinkhole is clogged.The central pits observed in three spiders are likely open entrances to the subsurface sinkholes and shallow voids.In the other spiders the absence of such pits may mean that the void entrances were clogged and the latest stages of downhill regolith flow covered the sinkhole openings.
This rapid formation process is followed by a much slower spider modification and degradation process driven by regolith gardening acting as topographic diffusion.The nature of spider modification strongly depends on whether the void entrance is clogged or remains open.In the clogged case the diffusion softens and smooths the spider.The legs become indistinguishable, and the spider turns into a rimless depression not identifiable as a spider, which further smooths and fills with regolith, becoming indistinguishable from an old impact crater, and finally disappears in a manner similar to an impact crater of comparable size (e.g., Fassett et al. 2022).If, however, the entrance into the subsurface void remains open, regolith transported by diffusion down to the spider center drains into the void, and the depression does not fill and smoothen (Figure 6), while the legs degrade and become indistinguishable at approximately the same time as in the At the sinkhole edges the upper surface of the regolith is pinned at the regolith base level, since regolith transported there would immediately drain.Away from the sinkhole edges, the diffusion smooths the surface, thus producing convex regolith slopes.Thus, spiders with open sinkholes are predicted to evolve into pits (e.g., Wagner & Robinson 2022) geologically rapidly.

Discussion
According to the scenario outlined above, the formation of spiders requires a concurrence of several conditions and events: (1) the presence of shallow subsurface voids of sufficient size (tens of meters), (2) the presence of a sufficiently thick regolith layer to produce resolvable chutes, and (3) some trigger causing the formation of a hole in the void roof.The first setting may occur only in some types of volcanic units with lava properties favoring the formation of shallow subsurface voids (e.g., Head & Wilson 2017, 2020;Wilson & Head 2018;Qiao et al. 2020aQiao et al. , 2020b)).This could explain spider occurrence in localized regions on the Moon, as well as their occurrence in tight clusters.A wider region in Mare Tranquillitatis is known to host a number of pits (e.g., Wagner & Robinson 2022) that are thought to be related to subsurface voids.This wide region also contains the majority of IMPs (Qiao et al. 2020a) and RMDSs (Zhang et al. 2020) that, according to some suggested scenarios, also require subsurface voids for their formation.This suggests that typical lavas in Mare Tranquillitatis indeed produced abundant shallow voids.The tight local association of the spiders with IMPs is also consistent with the suggestion that both types of young features require voluminous subsurface voids for their formation.In addition to the IMPs and RMDSs, we also note the presence of the elongated depressions, their parallelism with the chains of spiders contained within them, and their similar directional orientations (WNW).These three factors are similar to the orientations and associations of the two large IMP pits Sosigenes and Cauchy 5, and to the orientation of many summit pits on the population of summit craters on Tranquillitatis mare domes (e.g., Qiao et al. 2018Qiao et al. , 2020aQiao et al. , 2020bQiao et al. , 2021)).
The need for a sufficiently thick regolith layer for the formation of discernable spiders is consistent with the relatively old age of Tranquillitatis lavas (e.g., Hiesinger et al. 2011).The location of spiders in old shallow depressions that are likely to contain a thicker regolith layer is also consistent with this.In addition, the occurrence of spiders in depressions may be related to the possibility that the depressions are the sites of old lava vents or the surface manifestation of shallow dikes, where the formation of voluminous shallow voids is more likely (see Head & Wilson 2017, their Figure 16).
Somewhat puzzling is the cause of the predicted very recent (∼1 Ma or less), apparently sudden collapse of the roofs of the voids, given that that the void-containing substrate has survived intact for such a long period (longer than 3 Ga).Individual, roof-penetrating meteoritic impacts are unlikely to be the cause: they would be inconsistent with the observed spiders forming in such close proximity to each other and over such a short period.
Void roofs were likely fractured soon after lava emplacement due to temperature variations (cooling and thermal contraction), seismicity, and small impacts before and during the accumulation of a protective regolith layer.Some roofs may have partly or wholly collapsed soon after lava flow solidification.Despite fractures, however, some of them were mechanically stable, like vaults, and survived for several billion years, protected from the effects of impacts and temperature contrasts by an accumulating regolith cover.Occasionally, seismicity or relatively larger impacts might break the mechanical stability of a vault-like roof and cause roof collapse.We suggest seismic events as a cause of the partial collapse of the void roofs and the formation or rejuvenation of sinkholes.The close proximity of spiders to each other is consistent with seismicity as a candidate sinkhole formation trigger.The proximity of spiders to boulder clusters is also consistent with increased seismicity of the region, since boulder clusters are thought to form due to geologically recent seismic events (Valantinas & Schultz 2020).
We suggest that roofs of different voids in the region are likely to have collapsed from time to time throughout the long period of their existence, and only the consequences of the most recent collapse events are observable as spiders.In other words, many spiders may have been formed and then geologically rapidly evolved into pits or degraded into featureless smooth depressions not identifiable as spiders.The very small number of identifiable spiders is caused by their very short life span in comparison to the mare ages (four orders of magnitude difference).The more abundant small IMPs may have a longer life span.It is also possible that even smaller spiders formed in places with a thinner regolith layer cannot be identified in the available images due to resolution limitations.
Large lunar pits (Wagner & Robinson 2022) have associated funnels that exhibit a gradually increasing slope approaching the pit, from nearly horizontal at some distance from the pit.The outer edges of the funnels are poorly defined; the apparent width of the funnel walls is difficult to measure, however, its order of magnitude (∼30 m; Wagner & Robinson 2022, their Figure 10) is the same as the spider radius.The funnels of the large pits are not incised by chutes similar to the spider "legs."It is probable that those funnels were originally incised by chutes, and the chutes degraded and disappeared with pit aging due to regolith gardening.The large pits likely are much older than the spiders.The same is applicable to the dimple craters on the Moon (e.g., Greeley 1970), which are even older.
The formation and degradation of spiders may have implications for studies involving the analysis of crater sizefrequency distributions.On the one hand, former spiders may be misinterpreted as degraded impact craters, mimicking a higher crater density.On the other hand, the formation of spiders erases small (meter-size) superposed craters, mimicking a younger surface.

Conclusions
We have described the occurrence of spiders, extremely rare, small, crisp, extraordinarily young (<∼1 Ma) features on the Moon.We outlined a scenario for their formation by regolith drainage into shallow subsurface voids.This scenario naturally and logically explains the range of observed characteristic of spiders and the settings of their occurrence.The observed spatial association of spiders with IMPs is consistent with those IMP formation hypotheses that also involve ongoing drainage of regolith into shallow subsurface voids (Wilson & Head 2017;Qiao et al. 2019Qiao et al. , 2020aQiao et al. , 2020b;;Kreslavsky & Head 2023).The colinear occurrence of the elongated degraded pits and spider rows, and the orientation similarities to other apparently dike-related features in Mare Tranquillitatis, strongly suggest that voids associated with shallow and linear voids at the top of shallow dike tips are both plausible mechanisms.
Two logical points in the suggested scenario, although plausible, remain speculative: (1) we assume that selfsustaining granular flows of slightly cohesive poorly sorted lunar regolith can incise chutes with incipient dendritic pattern.The ability of lunar regolith to produce such flows has yet to be demonstrated.Laboratory studies of dynamic fluidization of lunar regolith analogs would be valuable not only for the understanding of spiders, but also for geoengineering applications.(2) We assume that lunar seismicity in the spider region is capable of triggering roof collapse of shallow voids and subsequent spider formation.Further seismic observations of the Moon as well as further photogeological observations of potentially related morphologies, especially boulder clusters and small IMPs, may help in better understanding this aspect of modern geological processes in general and spider formation in particular.In addition, future missions could deploy ground penetrating radar investigations on rover traverses to test for the presence of subsurface voids, and document their nature, geometry, and abundance.
The existence of spiders as the result of the recent collapse of void roofs indicates that shallow subsurface voids are also a hazard for operations at the lunar surface.As we discussed above, void roofs are almost certainly heavily fractured, and it is likely that a weak disturbance would be sufficient to trigger roof collapse.Operations on top of a roof of an unrecognized void could be such a trigger, and collapse could lead to disastrous consequences.

Figure 1 .
Figure 1.Typical small IMP: one of the IMPs forming an IMP cluster listed as "Carrel-1" in the Braden et al. (2014) catalog.The scene is located in Mare Tranquillitatis at 9°. 84 N, 25°.56 E. (a) Low-Sun image, illumination from the left, with a 71°incidence angle.Portion of Lunar Reconnaissance Orbiter Camera (LROC) narrow-angle camera (NAC) frame M1096351025L.(b) Low-Sun image, illumination from the right, with a 72°incidence angle.Portion of LROC NAC frame M1096329585R.(c) Contrast-enhanced high-Sun image, with a 15°incidence angle.Portion of LROC NAC frame M139741390R.(d) Sketch map of this IMPshowing the lower unit (L; "hollow"), upper unit (U; "meniscus"), the sharp contact between the lower and upper units (solid line), and gradational contact between the upper unit and surroundings (dashed line).Local equirectangular projection, north is at the top.

Figure 2 .
Figure 2. Spider cluster A. Long arrows point to spiders, short arrows point to examples of small IMPs.(a) Low-Sun image, illumination from the left, with a 71°i ncidence angle.Portion of LROC NAC frame M1096351025L.(b) Contrast-enhanced high-Sun image, with a 14°incidence angle.Portion of LROC NAC frame M1195277193R.(c) Sketch map of the scene: thin black lines are spider chutes, the black dot is the spider central pit, the bold black lines are the outlines of IMP lower units, the gray dashes are the outlines of spider-bearing depressions, with the short line pointing downslope, and the raster (dotted) patterns are boulder fields.Local equirectangular projection, north is at the top.

Figure 3 .
Figure 3. Spider cluster B. Long arrows point to spiders, short arrows point to examples of small IMPs.Local equirectangular projection.(a) Low-Sun image, illumination from the right with a 65°incidence angle.Portion of LROC NAC frame M1108139411L.(b) Low-Sun image, illumination from the left with a 69°i ncidence angle.Portion of LROC NAC frame M1234109579L.(c) Contrast-enhanced high-Sun image, with a 14°incidence angle.Mosaic of portions of LROC NAC frames M1343424827R (right part) and M185741082R (left part).Local equirectangular projection, north is at the top.
) and (c)).Moreover, the same line contains a number of small IMPs(Qiao et al. 2020a) between sites A and B and to the west of Site B. In sites C and D the alignment of individual spiders is not as pronounced as in sites A and B.

Figure 4 .
Figure 4. Locations of the four spider sets (white dots) in western Mare Tranquillitatis.Arrow points to a deep depression containing the Sosigenes IMP (Braden et al. 2014; Qiao et al. 2018), one of the largest IMPs.Portion of an LROC WAC image mosaic, local equirectangular projection, illumination from the left, where north is at the top.
see their Figure 3); (2) IMPs, including two of the three largest ones (Sosigenes and Cauchy 5; Braden et al. 2014), and the majority of the smaller ones (Qiao et al. 2020a their Figure 2); and ringmoat-dome structures (RMDSs; Zhang et al. 2020 their Figures 1

Figure 5 .
Figure 5. Location and topographic context of the four spider sets in western Mare Tranquillitatis.Simple cylindrical projection, north is at the top.(a) Detrended topography (Kreslavsky et al. 2017) derived from LOLA data (Zuber et al. 2010), where brighter shades denote locally higher elevations.Local depressions containing spiders are seen as small black spots at the tips of arrows (A-D).Wide dark (locally low) areas around the north-south trending elevated (white) wrinkle ridge and mare domes in the southern part of the scene are detrending artifacts: there are no actual depressions there.(b) Slopes with an ∼100 m baseline derived from SLDEM2015 topographic data (Barker et al. 2016), where brighter shades denote steeper slopes.The mare surface is very flat, where slopes steeper than ∼10°are associated with inner crater walls and the wrinkle ridge.Arrows A-D point to gentle depressions containing the sets of spiders.(c) Sketch map of the area: black ellipses denote the location and orientation of elongated gentle depressions containing spider sets A-D; a wrinkle ridge (WR) is also shown, while shaded areas are mare domes.

Figure 6 .
Figure 6.Model of spider degradation due to topographic diffusion.(a) Image of a real spider with a central pit from Figure 2; the image size is 80 m × 80 m. (b)-(e)Artificially shaded modeled relief corresponding to degradation states Kt of 0.6 m 2 , 1.9 m 2 , 3.8 m 2 , and 9.4 m 2 , respectively, which correspond to ages of 1.2 Ma, 3.8 Ma, 7.5 Ma, and 19 Ma, assuming a topographic diffusivity of K = 0.5 m 2 Ma −1 .

Figure 7 .
Figure 7. Schematic block diagram showing the suggested scenario of spider formation by the drainage of regolith into a shallow subsurface void of sufficient depth (>∼15 m) and width (>∼40 m).

Figure 8 .
Figure 8. Block diagram showing the characteristics of dikes intruding into the lunar crust and the several features that can be produced at the surface, ranging from no evidence (a) through explosive venting and collapse pits (b), to surface lava flow eruptions (c)-(f) (from Qiao et al. 2020b; after Head & Wilson 2017).The inset illustrates a dike penetrating the shallow subsurface with gas exsolution near the tip, leaving a linear void space (from Head & Wilson 2017).

Figure 9 .
Figure 9. Oblique view of a sinkhole that formed in slightly wet terrestrial beach sand.The outer diameter of the funnel is ∼0.8 m.