Boulder populations and orientation trends on asteroid Ryugu: implications for rubble-pile surface processes

Abstract

Asteroid (162173) Ryugu is a spinning-top shaped, rubble-pile C-type asteroid with a surface dominated by boulders spanning a broad size-range. Using Hayabusa2 ONC-T mosaics for mapping, we present the most extensive global boulder dataset to date, representing an eleven-fold increase over the earliest survey and greater completeness than later AI-based estimates. Crucially, this work provides the first global boulder orientation dataset. The cumulative size-frequency distribution follows a power-law slope of − 2.665 ± 0.066 (boulders ≥ 3 m), indicating its fragmented nature. Boulder density is lower along the equatorial ridge, consistent with regolith accumulation during earlier YORP-driven spin-up and poleward migration of blocks during the present spin-down phase. Orientation analyses reveal systematic hemispheric trends in what we define as slope-breaker zones (i.e., regions of high slope differentials), with NW–SE alignment in the northern hemisphere and NE–SW in the southern hemisphere, but absent at the equatorial ridge and poles. We interpret these as reorientations that developed during downslope boulder migration, as despite ongoing spin-down, the current spin rate remains high enough to drive such motion. In addition, fresh craters exhibit locally heightened boulder densities, recording ongoing resurfacing by impact events that exhume buried subsurface boulders and segregate them from finer regolith via seismic shaking induced granular convection. Together, these results yield new insights into rubble-pile surface evolution, rotational dynamics, and impact-driven resurfacing, with boulder orientations providing critical evidence of spin-related downslope transport and establish a framework for comparative asteroid studies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-40370-4.

Introduction

The near-Earth asteroid (162173) Ryugu was explored by the Hayabusa2 sample-return mission operated by Japan Aerospace Exploration Agency (JAXA). Initial observations revealed that Ryugu is a carbonaceous spinning-top asteroid with an equatorial ridge and a rugged surface dominated by boulders suggesting a rubble-pile configuration^1,2. It exhibits a complex geological history involving catastrophic disruption, reaccumulation, global deformation, and rotational reshaping of its parent body^3,4. Ryugu’s bulk density (~ 1190 kg/m³) and high porosity (> 50%) further support its interpretation as a gravitational rubble pile^2,3,5. This rubble-pile structure makes Ryugu an important case study for understanding the mechanical stability, surface evolution, and dynamical history of small bodies as nearly all of the asteroids in the size range 0.2–10 km are considered to be rubble-piles⁶. Recent work by Bierhaus et al.⁷ proposed that Bennu’s interior, another carbonaceous asteroid recently explored, consists of large, competent boulders underlying a 2–6 m thick fine grained surface layer, based on variations observed in crater interiors and ejecta composition. The presence of a mechanically stronger bulk volume beneath a finer layer may also explain the formation of regional bulges. Cho et al.⁸ suggest a similar stratification for asteroid Ryugu.

Boulders are among the most diagnostic surface features on rubble-pile asteroids. Based on asteroid shape and calculated density, it can be assumed that boulders also constitute much of the Ryugu’s bulk. For rubble-pile asteroids like Ryugu, boulders are typically regarded as fragments inherited from catastrophic disruption and reaccumulation of a parent body, while subsequent small-scale impacts and cratering events have, to a limited extent, modified (e.g., disrupted) and redistributed them on the surface^9–11. Similar trends have been observed on other rubble-pile asteroids such as Bennu, Didymos and Itokawa, which also exhibit boulder dominated surfaces and high bulk porosity¹². As boulders preserve records of both catastrophic disruption and subsequent surface modification, their properties provide crucial constraints on regolith production, surface age, and processes like seismic activities, landslides, and granular convection^13–15. Previous studies on Ryugu have examined boulder populations to characterize their sizes, shapes, and distribution patterns. Michikami et al.¹⁰, for instance, analysed approximately 4400 boulders and established fundamental size-scaling relationships while Sugita et al.¹ and Watanabe et al.² linked global geomorphology to Yarkovsky-O’Keefe-Radzievskii-Paddack effect (YORP)¹⁶ driven spin evolution. However, prior studies on boulder orientations have been geographically limited to specific regions. By leveraging a larger sample size, we extend orientation analysis to a global scale for the first time, thereby establishing boulder orientation as a diagnostic of mass wasting, and spin-driven reorganization of surface materials. Global-scale boulder analyses of Bennu and Ryugu show that boulder size sorting is a function of asteroid’s spin state, with equatorward transport on Bennu and poleward transport on Ryugu^17–19. These results indicate how small changes in rotation rate can steer rubble-pile evolution along different paths. Previous studies using AI-based approaches¹⁷ demonstrated spin-state dependent boulder transport. In this study, we used manual mapping, which complements these AI analyses by providing a more reliable global dataset. Although size-frequency trends and spatial distributions are now well defined, the orientations and clustering of boulders remain poorly studied at the global scale. These properties are important for identifying downslope transport, structural modification, and surface evolution.

Here we present the first global dataset of boulder orientations on Ryugu based on 49,536 manually mapped boulders, each recorded with long-axis length, position, and orientation (Fig. 1). Comparison with established size-frequency distributions confirms reliability and enables the first global, statistically robust analysis of orientation patterns across the asteroid. The analysis aims to investigate eventual hemispheric alignment trends, potential preferred orientations in slope-breaker zones (Fig. 1e), and if there is a reduced boulder density on the equatorial ridge. Our results provide new constraints on Ryugu’s surface modification and rotational dynamics and offer a comparative framework for rubble-pile processes on other small bodies such as Bennu and Itokawa.

Fig. 1.

Global distribution of mapped boulders on Ryugu with mapped boulders (this study) and points of interest (POIs) following Komatsu et al.²⁰. (a) Global ellipsoidal projection of the ONC-T mosaic using Ryugu custom ellipsoid (semi-major axis 502.0 m, semi-minor axis 437.744 m, inverse flattening 7.8125) (b) North polar stereographic projection (radius 449.0 m). (c) Close-up view illustrating the mapping procedure. (d) South polar stereographic projection (radius 449.0 m). The orange lines are the mapped boulder long axes, dotted region refers to the western bulge, maroon lines indicate mapped craters, the semi-transparent yellow, blue, and cyan regions denote the equatorial ridge, large boulders (saxa) and troughs (fossa) respectively. (e) Slope map of Ryugu derived from the Hayabusa2 onboard LiDAR DEM (~ 1 m/pixel), with slope calculated relative to the local horizontal and arrows indicating downslope direction. The red regions indicate the slope-breaker zones while the blue region indicates the rise of the equatorial ridge.

Results

Size-frequency distribution (SFD)

A total of 49,536 boulders ≥ 0.75 m were mapped across Ryugu’s surface (Fig. 1), of which 23,553 boulders ≥ 3 m were used to construct the global cumulative SFD (Fig. 2). As the ONC-T mosaics have resolutions of 0.25 at equatorial region, 0.55 m/pixel at mid-latitudes and 1.0 m/pixel at the poles; our minimum mapped boulder sizes are ≥ 0.75 m at the equatorial ridge, ≥ 1.65 m at mid‑latitudes, and ≥ 3 m near the poles. The distribution follows a power-law with exponent α = − 2.665 ± 0.066 (95% CI 2.644–2.687; fit uncertainty from the Kolmogorov-Smirnov (KS) statistic), with the minimum boulder size for the fit (Xmin) at 3.10 m. In the equatorial region (− 40° to + 40° latitude), the SFD is steeper, with α = − 3.239 ± 0.037 (95% CI 3.202–3.275), Xmin = 3.48 m, and 38,243 boulders included in the fit. The western bulge exhibits a shallower slope, α = − 3.154 ± 0.017 (95% CI 3.094–3.215), with Xmin = 3.22 m and 4872 boulders included in the fit. Cumulative SFDs were normalized to mapped surface area and are shown in Fig. 2d–f, with representative error bars showing √N uncertainties. Histograms of boulder diameters and latitudinal/longitudinal frequency distributions are shown in Fig. 2a–c. This extensive dataset enables identification of subtle yet systematic variations in SFD slopes across regions (Fig. 3).

Fig. 2.

Size-Frequency Distribution (SFD) of boulders on Ryugu. (a) Histogram of mapped boulder diameters. (b) Latitudinal frequency distribution of boulders (≥ 3 m), showing distinct peaks in slope-breaker zones. (c) Longitudinal frequency distribution (≥ 3 m), highlighting low densities in western regions. (d) Global cumulative SFD, fitted using Clauset’s power-law method²¹, slope = − 2.665 ± 0.066. (e) Cumulative SFD of equatorial ridge, slope = − 3.239 ± 0.037. (f) Cumulative SFD of western bulge, slope = − 3.154 ± 0.017. (g) Comparison of cumulative SFDs between Ryugu (this study) and other asteroids, illustrating variations in boulder populations.

Fig. 3.

(a) Colormap of power-law slope (≥ 3 m) per 20° × 20° bin, with steeper values in the equatorial region of the western bulge and crater vicinities. The black parentheses indicate slope-breaker zones. (b) Heatmap of boulder counts per 2° × 2° bin (≥ 3 m), showing heightened counts in slope-breaker zones. The black parentheses indicate slope-breaker zones. (c) Boulder distribution in Brabo crater, mapped on a high-resolution mosaic; red dots indicate boulders and the black outline marks the crater rim²⁰. (d) Boulder distribution in Kolobok crater, mapped on a high-resolution mosaic; red dots indicate boulders and the black outline marks the crater rim (Komatsu et al., 2025). (e) Comparison of boulder density between Brabo, Kolobok, and Ryugu’s equatorial region, showing nearly double the abundance in younger craters. (f) Boulder distribution in Brabo crater, mapped over the B-x composite. (g) Boulder distribution in Kolobok crater, mapped over the B-x composite.

Orientation

This study provides the first global boulder orientation dataset for Ryugu, extending previous regionally restricted investigations on other asteroids. Boulder orientations on Ryugu are axially symmetric, as they are mapped as lines (a line pointing toward 0° is equivalent to one pointing toward 180°). For larger boulders (≥ 3 m), the global distribution is bimodal with peaks at 34.8°±2.17° and 140.2°±1.39°²². Smaller boulders (< 3 m) are predominantly aligned along a single axis at 179.0°± 0.74°. In the northern hemisphere, larger boulders show a uni-modal peak at 155.3°± 1.72°, while in the southern hemisphere two peaks occur at 36.2°±1.45° and 137.1°±1.32° (Fig. 4a–e; Table 1).

Fig. 4.

Orientation of boulders on Ryugu with Von Mises fit represented by black curve. (a) Global dataset (≥ 3 m) showing dominant NE–SW (34.8°±2.17°) and NW–SE (140.2°±1.39°) alignments. (b) Smaller boulders on the equatorial ridge (≤ 3 m) show strong N-S clustering (179.0°± 0.74°). (c) Boulders in the northern hemisphere (≥ 3 m) dominated by NW–SE alignment (155.3°± 1.72°). (d) Boulders in the southern hemisphere (≥ 3 m) showing a major NE–SW (36.2°±1.45°) and a minor NW–SE (137.1°±1.32°) alignments. (e) SBMT shape model with mapped boulders and high-resolution mosaics overlain. Rose diagrams show boulder orientations per 10° latitudinal interval. (f) Surface acceleration map (modified after JAXA) with rose diagrams of boulder orientations and SFD histograms for each zone in the northern and southern hemispheres.

Table 1.

Results of rao’s spacing test for boulder orientations on Ryugu. Orientation subsets are reported by size fraction (≥ 3 m and < 3 m), hemisphere, and 10° latitudinal bins. Reported values include sample size (n), p ≤ 0.05 indicates rejection of uniformity and orientation modality with peak azimuths as per von mises fit. Global orientations are bi-modal (NE–SW and NW–SE), while hemispheric trends differ, with the Northern hemisphere dominated by NW–SE alignment and the Southern hemisphere exhibiting both NE–SW and NW–SE peaks. Polar-region bins do not reject uniformity.

Subset	No. of boulders (n)	p_value (Non-Uniform for ≤ 0.05)	Rao U	Interpretation	Modality	Peak1	Peak2
Global ≥ 3 m	24,553	0	357.361	Non-uniform (Oriented)	Bi modal	34.8° ± 2.17°	140.2° ± 1.39°
Global ≤ 3 m	24,983	0	357.406	Non-uniform (Oriented)	Uni-modal	179.0° ± 0.74°	N/A
North ≥ 3 m	11,768	0	354.494	Non-uniform (Oriented)	Bi modal	25.9° ± 18.37°	142.5° ± 2.57°
South ≥ 3 m	12,785	0	354.932	Non-uniform (Oriented)	Bi modal	36.2° ± 1.45°	137.1° ± 1.32°
Lat − 90 to − 80° (South)	63	0.505	131.143	Uniform	N/A	N/A	N/A
Lat − 80 to − 70° (South)	241	0	170.29	Non-uniform (Oriented)	Uni-modal	158.1° ± 9.40°	N/A
Lat − 70 to − 60° (South)	535	0	249.645	Non-uniform (Oriented)	Bi modal	36.6° ± 8.13°	133.5° ± 5.92°
Lat − 60 to − 50° (South)	1065	0	299.493	Non-uniform (Oriented)	Uni-modal	171.9° ± 9.11°	N/A
Lat − 50 to − 40° (South)	1421	0	314.622	Non-uniform (Oriented)	Uni-modal	162.1° ± 11.20°	N/A
Lat − 40 to − 30° (South)	1727	0	330.459	Non-uniform (Oriented)	Uni-modal	36.7° ± 1.36°	N/A
Lat − 30 to − 20° (South)	2642	0	347.725	Non-uniform (Oriented)	Bi modal	27.5° ± 3.57°	127.5° ± 3.57°
Lat − 20 to − 10° (South)	2744	0	351.632	Non-uniform (Oriented)	Bi modal	36.7° ± 2.99°	138.0° ± 4.54°
Lat − 10 to − 0° (South)	2347	0	352.885	Non-uniform (Oriented)	Bi modal	40.0° ± 4.27°	129.2° ± 5.10°
Lat 0 to 10° (North)	2294	0	352.79	Non-uniform (Oriented)	Bi modal	12.2° ± 12.54°	143.8° ± 3.76°
Lat 10 to 20° (North)	2818	0	348.75	Non-uniform (Oriented)	Uni-modal	160.3° ± 1.99°	N/A
Lat 20 to 30° (North)	2769	0	340.203	Non-uniform (Oriented)	Uni-modal	150.6° ± 1.51°	N/A
Lat 30 to 40° (North)	1497	0	317.353	Non-uniform (Oriented)	Bi modal	37.5° ± 12.02°	147.5° ± 12.02°
Lat 40 to 50° (North)	994	0	295.792	Non-uniform (Oriented)	Uni-modal	14.9° ± 19.10°	N/A
Lat 50 to 60° (North)	719	0	273.361	Non-uniform (Oriented)	Uni-modal	168.7° ± 35.35°	N/A
Lat 60 to 70° (North)	413	0	226.187	Non-uniform (Oriented)	Uni-modal	125.5° ± 10.21°	N/A
Lat 70 to 80° (North)	200	0.213	136.8	Uniform	N/A	N/A	N/A
Lat 80 to 90° (North)	64	0.287	137.538	Uniform	N/A	N/A	N/A

Rao’s spacing test was applied at 10° latitudinal intervals to assess uniformity in the data. All bands except at the poles (90°S–80°S and 70°–90°N) reject uniformity at p ≤ 0.05 (> 95% confidence) (Table 1), indicating significant modal orientation peaks²². Boulders are mostly oriented, with bi- or uni-modal peaks largely consistent across hemispheres²².

The mean resultant length (MRL) quantifies the clustering strength of axial orientation data, where directions 180° apart are treated as equivalent. It ranges from 0 (completely uniform) to 1 (perfectly aligned) and is calculated by doubling the angles to adjust for axial symmetry.

In this study, MRL values are shown in the heatmaps and grids in Fig. 5a–d, where higher values indicate regions of strong boulder reorientation. Such areas correspond to slope-breaker zones and the rims of large craters (≥ 20% of Ryugu’s equatorial radius; Fig. 5b, d). We define slope-breaker zones as low topographic and geopotential slope regions immediately following the high slope differentials at the margins of the equatorial ridge (Fig. 1e). On Ryugu, these zones occur at approximately ± 35° latitude, more specifically, between 20–30°N and 30–40°S (Fig. 5b), where the slope transitions from steep (equatorial ridge) to gentler values. We computed the mean topographic slope as a function of latitude from the DEM-derived slope raster using 5° latitude bins and define slope-breaker zones as the latitude bands where the latitudinal gradient of mean slope reaches a maximum (largest inter-bin change), which occurs at ~ 20–30°N and ~ 30–40°S (Fig. S3).

Fig. 5.

Mean orientations of boulders (a) 0.75–3 m and (b) ≥ 3 m in 5° × 5° bins; line color corresponds to mean resultant length (MRL), with darker colors indicating stronger clustering. (c) Boulder orientation deviation per 5° × 5° bin from the corresponding 20° × 20° locale. Point size scales with angular deviation and point color reflects the MRL of the bin. (d) Heatmap of MRL, constructed from 5° × 5° bin data interpolated to a 0.5° grid. (e) Heatmap of boulder orientation deviation relative to the 20° × 20° locale, based on 5° × 5° bins interpolated to a 0.5° grid. The color scale represents angular deviation, with warmer tones indicating lower differences from the local background. Note the darker color in slope-breaker zones indicating strong orientation clustering.

Spatial density

Boulder density is heterogeneous across Ryugu’s surface. The global mean density is 9093 km⁻² (≥ 3 m), increasing to 26,072 km⁻² (≥ 0.75 m) in the equatorial region (40°S-40°N). The western bulge has a lower density, with 8627 km⁻² (≥ 3 m) and 22,060 km⁻² (≥ 0.75 m) in the equatorial region. Density estimates were area-normalized, and 2° × 2° bins were applied for the heatmaps shown in Fig. 5b, d. The equatorial ridge shows comparatively low densities, while higher values are observed in slope-breaker zones between 20–30°N and 30–40°S, representing the areas of high slope differentials (Fig. 1e). Younger craters with bluer spectral slopes²³ like Kolobok (x-b ratio < 1.0) contain more boulders than older, redder (x-b ratio > 1.0) ones like Brabo (Fig. 3c–e). The global distribution of mapped boulders and POIs is shown in Fig. 1a–e.

Discussion

Before discussing the implications of boulder SFDs, it is important to mention that the mapping was conducted using ONC-T mosaics with spatial resolutions ranging from ~ 0.25 m/pixel in the equatorial region to ~ 1.0 m/pixel at high latitudes. These differences in resolution affect minimum detectable boulder size and can bias SFDs. To minimize this all regional SFD analyses and inter-regional comparisons in this study are restricted to boulders ≥ 3 m, a size well above the resolution limit across all latitudes, and boulder frequencies are normalized by mapped surface area (km⁻²). The cumulative SFD of Ryugu’s boulders ≥ 3 m, with a steep power-law index of − 2.665 ± 0.066, suggests a boulder population dominated by catastrophic fragmentation of a parent body, closely matching the − 2.65 ± 0.05 slope of Michikami et al.¹⁰ and within the − 2.5 to − 3 range estimated by Sugita et al.¹ (Fig. 2g). Similar steep slopes on Bennu (Fig. 2g)^14,24 also indicate a catastrophic fragmentation of a parent body followed by gravitational reaccumulation, rather than impact ejecta^6,25. The high total boulder volume, together with the presence of exceptionally large boulders (> 20 m) such as Otohime Saxum, cannot be explained solely by impact ejecta from cratering¹⁰ (Fig. 1c). Ryugu’s high bulk porosity (~ 50%) and evidence of exhumed deep-seated fragments^2,15 support its rubble-pile configuration and accretionary origin with regional SFD variations reflecting localized processes such as mass wasting and resurfacing. Granular convection driven by seismic shaking from impacts can mobilize subsurface material, causing larger boulders to migrate upward and fines to accumulate at depth; analogous to the “Brazil Nut Effect” observed in asteroid regolith²⁶. On Ryugu, this mechanism likely contributes to global resurfacing of boulders by bringing buried ones to the surface during continuous micro-seismic activity²⁷.

Regionally, the equatorial ridge shows a steeper slope (α = − 3.239 ± 0.037) (Fig. 2e), indicating a relative enrichment in larger boulders compared to the global average, whereas the western bulge exhibits a relatively shallower slope (α = − 3.154 ± 0.017) (Fig. 2f), suggesting resurfacing of asteroid due to impact-induced exhumation of boulders. Local power law fits, however, show steeper values in the equatorial region of the western bulge than other equatorial parts. A steeper SFD slope typically indicates greater surface processing and in this case is likely driven more by regolith accumulation than impact events. Fine‑grained regolith below the mapping resolution can obscure larger boulders, while impact events can remove this cover and expose them²⁷. The observed pattern thus reflects regolith blanketing rather than impact‑driven resurfacing. The high global boulder volume and exceptionally large boulders suggest that Ryugu preserves deep-seated fragments from its progenitor. Both the asteroid size and the number density of boulders ≥ 20 m on Ryugu is approximately twice that of Bennu (50 km⁻²), with Bennu’s maximum boulder size of ~ 58 m (Benben Saxum) compared to Ryugu’s ~ 139 m (Otohime Saxum). This suggests that Ryugu retains a comparable or stronger record of its parent body’s structural fragments than Bennu, likely due to differences in disruption intensity or preservation during reaccumulation¹⁰.

Orientation results provide new constraints on Ryugu’s surface processes that were not accessible in earlier, regionally limited studies. Boulder orientations on Ryugu show a general global trend, with larger boulders (≥ 3 m) aligned NE–SW (34.8°±2.17°) and NW–SE (140.2°±1.39°), while hemispheric patterns are a uni-modal peak at 155.3°± 1.72° in the north and a major 36.2°±1.45° and minor 137.1°±1.32° peaks in the south (Fig. 4a, c,d). Rao’s spacing test indicates significant non-uniformity (rejects uniformity at p ≤ 0.05) across most latitudes; in the polar latitude bins uniformity is not rejected (Table 1).

Downslope motion aligns boulders in preferred directions because they stabilize in mechanically stable positions, typically with their long axes oriented across the slope, while Coriolis effects on Ryugu provide a secondary directional bias. Boulders following a NW–SE trend in the northern hemisphere and NE–SW in the southern hemisphere are interpreted to reflect downslope transport and stabilization at slope transitions, with Coriolis acceleration acting as a secondary modifier of the final resting orientation. As boulders migrate poleward under YORP spin-down dynamics, the Coriolis acceleration deflects their trajectories perpendicular to their direction of motion, imparting a preferred orientation (Figs. 4e and 5b and d)²⁸. MRL heatmaps show strong clustering along slope-breaker zones and crater rims, which correlates with regions of active mass wasting (Fig. 5b, e). The hemispheric asymmetry in their latitudinal position likely reflects differences in regional topography and the distribution of large craters and saxa. Taken together, these observations suggest that spin evolution and surface processes play a dominant role in shaping boulder orientations, though the mechanisms remain interpretive rather than conclusive. Our orientation results provide the missing directional dimension by documenting how boulders align during transport and emplacement, compared to mass-movement pathways driven by differences in spin state. Together, these records indicate that Ryugu’s surface retains evidence of both poleward size sorting from its present spin-down phase and orientation clustering linked to local slope activity and earlier changes in rotation history. Thus, while manual mapping is more time-consuming than AI-based approaches, it enables accurate identification and counting of a larger number of boulders (49,536 ≥ 0.75 m in this study vs. ~ 25,000¹⁷), making it a complementary yet more robust method.

Preferred orientations are also consistent with the physical controls described by Michikami & Hagermann²⁹, namely that the local gravity and friction angle of boulders are the principal factors governing preferred alignment on small bodies. Downslope motion forces boulders to stabilize with orientations dictated by maximum friction angles, while gravity thresholds (Fig. 4f) enhance the likelihood of clustering. The concentration of aligned boulders in slope-breaker zones (Fig. 2b) on Ryugu therefore indicates that slope transitions act as gravity and friction angle thresholds that promote reorientation and arrest of boulders²⁹. Dominant NE–SW trends in the northern hemisphere and NW–SE trends in the southern hemisphere reflect preferential stabilization along mass-movement pathways in slope-breaker zones. Secondary, non-dominant trends likely record localized mass movements associated with large craters and saxa supported by their higher presence in the southern hemisphere. The observed alignment patterns are thus in agreement with the expected mechanical behavior. Similarly, smaller boulders, being more mobile, are more reoriented than larger ones (Figs. 4b and 5a).

Orientations of boulders ≥ 3 m show a correlation with local surface acceleration. Clustering is strongest where surface acceleration is moderate (~ 0.13 mm/s²), while orientations at the poles, where gravity is higher, tend to appear more uniform. On the equatorial ridge, orientations seem largely governed by surface processes, whereas slope-breaker zones display stronger clustering (Figs. 4e and 5b and d). This correlation reflects optimal conditions for both Coriolis-driven reorientation and preservation of oriented boulders. At intermediate latitudes ( ~ ± 35, slope-breaker zones), gravity values (~ 0.10–0.11 m/s²) are sufficiently strong to permit Coriolis deflection during downslope migration, yet weak enough that arrested boulders resist re-mobilization and retain their reoriented configurations. In contrast, at the equatorial ridge (lowest gravity, ~ 0.07 m/s²), high boulder mobility prevents stable preservation, and newly reoriented boulders continue migrating before acquiring a fixed orientation. At the poles (highest gravity, ~ 0.13 m/s²), reduced Coriolis effectiveness per unit velocity diminishes reorientation intensity, resulting in weaker orientation clustering (low MRL). Additionally, sharp changes in surface slope at slope-breaker zone margins act as mechanical arrest points, transitioning from steep equatorial slopes to gentler mid-latitude slopes, causing gravitational driving force to drop below the friction threshold (~ 32°³⁰) and forcing boulders to stabilize in place. The concentration of high MRL values precisely at slope-breaker zones (Figs. 4f and 5b, d and e) therefore records this gravity and slope dynamics. In this way, Ryugu’s surface architecture can be interpreted as the combined imprint of spin-induced migration and local gravity. A sudden change in Ryugu’s spin rate could theoretically trigger surface block rearrangements; however, our dataset does not allow us to quantify the exact magnitude of rotation change required. Localized mobilization by seismic shaking and small-crater reworking may contribute to near-surface modification; however, the global patterns of size sorting and orientation clustering are primarily governed by YORP-driven migration. Testing whether the observations reflect discrete spin events or gradual YORP evolution would require physical modelling.

Small boulders (≤ 3 m) represent the most fragmented population on Ryugu, formed through both impact events and thermal fracturing (cf^31,32). Most of these boulders exhibit a strong N-S (179.0°± 0.74°) trend (Fig. 4b)³¹. We interpret this N-S orientation to be resulting from thermal fatigue rather than instantaneous thermal stress fracturing. While rapid heating/cooling produces isotropic fracturing, thermal fatigue i.e. the slow growth of pre-existing micro-cracks through repeated diurnal cycles can create anisotropic crack patterns. Although the diurnal thermal skin depth (~ cm) is much shallower than typical boulder dimensions, preferential stress accumulates over repeated day-night cycles. Since Ryugu’s rotational axis is nearly vertical (~ 172° inclination), solar irradiation produces systematic E-W heating asymmetry. During each rotation, the eastern side of boulders heats in the morning and the western side in the evening. This repeated E-W asymmetry creates cumulative thermal stress perpendicular to solar radiation, preferentially propagating N-S-oriented cracks and eventually fragmenting the boulders, producing the observed trend^31,32. An alternative explanation is that smaller impact ejecta fragments may escape Ryugu’s weak gravity, whereas larger fragments settle near the poles where gravitational force is greater. In either case, thermal fracturing plays a critical role in ongoing fragmentation with processes like exfoliation further contributing to boulder breakdown³². Boulder mobility on Ryugu is controlled by a size-dependent transport mechanism wherein smaller boulders migrate more readily than larger ones. Although internal friction angles on rubble-pile asteroids remain relatively constant (~ 32°), smaller boulders (say 1–5 m) have lower critical shear stress thresholds and thus move downslope more readily than larger boulders (> 5 m). Relatively fine regolith particles (mm to a few cm), however, exhibit higher static friction due to electrostatic and cohesive forces, reducing their mobility. This can be augmented by low gravity at the equatorial ridge creating an optimal mobility window where boulders between ~ 1–5 m preferentially migrate, while larger and finer regolith material accumulates in depositional zones. Under Ryugu’s current YORP spin-down phase, newly exposed boulders continue to undergo thermal fracturing due to diurnal temperature variations and migrate poleward. We hypothesize this to be an impact-independent cycle of fragmentation that may operate on other rubble-pile asteroids as well.

Boulder density on Ryugu is heterogeneous, averaging 9093 km⁻² for ≥ 3 m boulders but reaching 26,072 km⁻² for ≥ 0.75 m boulders in the equatorial region (Figs. 2b and c and 3a and b). The equatorial ridge exhibits lower densities, indicating that finer regolith blanketing and prior migration of larger boulders have obscured surface expression^33,34, whereas slope-breaker zones show higher densities (Fig. 2b), suggesting preferential accumulation in areas of steep slope differentials under active downslope (poleward) transport. In contrast to the higher mobility of small boulders, the absence of boulders > 3 m along the equatorial ridge (Fig. 2b) suggests that these were preferentially transported poleward during Ryugu’s present spin-down phase. This pattern indicates that boulder migration follows size-dependent mobility thresholds.

These distributions indicate a strong impact of YORP-induced spin evolution on boulder migration. During spin-up, equatorial migration formed the ridge^33,34. In the current spin-down phase, boulders migrate poleward from equatorial regions, depleting equatorial densities while enriching the slope-breaker zones. Polar regions are relatively depleted in small boulders, preserving a record of prior equatorial migration. Overall, the surface architecture of Ryugu reflects the combined effects of catastrophic fragmentation^1,10, rotational dynamics, and topography (Fig. 1a–d).

The western bulge exhibits a steep SFD slope (α = − 3.154 ± 0.017), low boulder density (8627 km⁻² ≥ 3 m), and reduced crater frequency²⁰ (Figs. 2c and 3b), suggesting that this region has experienced significant surface weathering and thermal fragmentation. The lower crater density likely reflects obscuring of craters by thermal fatigue and regolith blanketing³⁵. Its uplifted topography may inhibit regolith redistribution, while its distinct spectral properties indicate compositional differences from surrounding terrains. One plausible scenario is that the bulge represents large-sized remnants or high-cohesion aggregates of Ryugu’s parent body. In such a case, a buried block could locally modify the gravitational field and slope stability, analogous to Itokawa’s Muses Sea⁹and Bennu’s equatorial bulge⁶.

Younger craters with bluer spectral slopes exhibit higher boulder densities (Fig. 3c–e), indicating exposure of buried blocks. Older, redder craters show reduced boulder abundance due to thermal fatigue, micrometeoroid bombardment, and/or ejecta blanketing. High power-law indices in crater interiors reflect exhumation of subsurface boulders and dispersal of finer regolith as ejecta which enhance surface expression. Compared to Bennu, Ryugu exhibits slower reddening rates due to it being at an earlier stage of space-weathering³⁶; possibly due to its longer rotation period and greater heliocentric distance³⁷.

The orientation metrics, particularly clustering strength, and deviation maps, establish orientation as a sensitive tracer of surface processes, highlighting regions of mass wasting and spin-induced reorientation (Figs. 5b and c and 6e). Strongly clustered alignments, marked by high MRL values, occur along slope-breaker zones and crater walls, indicating regions of mass wasting. High angular deviations (from local trend i.e. 20° x 20°) along crater margins suggest localized disturbance (Fig. 5b, d), while low MRL in younger crater interiors reflect uniform orientations of freshly exhumed subsurface boulders. Geopotential variations corresponding to both the past high spin-rate state and the present spin-down phase of Ryugu were derived from the models of Watanabe et al.² and Shimizu et al.¹⁷ (Fig. 6a). At higher spin rates (> 5 h) the models show pronounced geopotential troughs at mid-latitudes, between ~ 20° and 40° in both hemispheres, which we identify as slope-breaker zones, a sharp transition from the steep equatorial ridge to gentler mid-latitude terrains. Under the current geopotential configuration, boulders are predicted to migrate toward these troughs from both equatorial and polar regions. However, due to Ryugu’s spinning-top morphology and the larger surface area at the equatorial regions, boulders migrating from this region are expected to dominate the slope-breaker zones¹⁷.

Fig. 6.

(a) Latitudinal distribution of Average geopotential with current, past and possible future spin rates as per Watanabe et al.² and Shimizu et al.¹⁷. The red parentheses mark the slope-breaker zones. (b) Average angular deviation of boulder long axes from the aspect (slope direction) as a function of latitude. Note the higher and lower values in the slope-breaker zones and the equatorial regions respectively. (c) Average angular deviation as a function of longitude. Note the relatively lower values in the western region. (d) Heatmap of angular deviation, constructed from 5° × 5° bin data interpolated to a 0.5° grid. Note the lower values in equatorial region (marked by blue parentheses) and higher values in the slope-breaker zones (marked by red parentheses). (e) Heatmap of angular deviation, constructed from 2.5° × 2.5° bin data interpolated to a 0.5° grid to better show local variations. Note the occasional higher values along crater walls and interior, possibly indicating mass wasting.

To evaluate the kinematic imprint of this migration, we performed a direct comparison between boulder long-axis orientations and local downslope azimuths (aspects) derived from the DEM. Angular differences were calculated using axial statistics (0–90°). Latitudinally, the slope-breaker zones exhibit higher angular deviations, which we interpret as the cumulative effect of Coriolis-induced reorientation during downslope transport, as discussed earlier (Fig. 6b). Longitudinally, the western region shows relatively smaller deviation, possibly as a result of lower crater frequency in the western bulge (Fig. 6c). The equatorial regions show lower angular deviations, consistent with minimal Coriolis influence and migration dominated by centrifugal force (Fig. 6d). At a more local scale, areas surrounding crater walls display high angular deviations, reflecting localized mass wasting and slope-controlled reorientation of boulders within crater interiors and rims (Fig. 6e). Mass wasting is therefore localized rather than uniform on Ryugu, concentrated on slopes of large boulders, crater walls, and the equatorial ridge. Higher deviation in slope-breaker zones further demonstrates that spin-induced migration has a greater imprint on surface boulders than other factors, confirming the significance of spin-induced reorientation and its implications for landing-site hazard assessment on small bodies.

Despite these insights, uncertainties remain. Boulder SFDs may be obscured by regolith cover, especially in the equatorial ridge where it accumulates the most (Figs. 1c and 2e). Orientation datasets lose statistical power near the poles due to smaller sample sizes (Fig. 4c, d). Timescales of YORP evolution are not precisely constrained, limiting the ability to link orientation trends to specific dynamical epochs. Furthermore, the origin of the western bulge as a preserved remnant or reaccumulated block requires verification through in situ geophysical methods such as seismic sounding or radar tomography.

Orientation studies on other rubble-pile asteroids have been limited to local sites. At Bennu, boulders on the Nightingale and Osprey regions showed bimodal alignments (~ 95° and ~ 163°) oriented downslope along crater rims^{18,19,38–40}, while Didymos and Dimorphos (although with limited data) display unimodal clustering near ~ 28° and ~ 125°, respectively, linked to YORP-driven regolith transport and landslide deposits^5,28,30. Itokawa’s Muses Sea exhibits weaker anisotropies (~ 45°–60°), attributed to microgravity sorting⁴¹.

Despite the regional scale, these orientation trends consistently show non-uniform unimodal or bimodal peaks, broadly comparable to the global patterns observed on Ryugu. Such consistency indicates that boulders are systematically reoriented by surface processes rather than uniformly distributed.

These alignments, observed across bodies of different size and/or history, confirm the widespread influence of spin-driven migration, slope failure, and thermal effects in reorganizing surface blocks. The observed distribution patterns indicate a link between surface organization and internal structure. Higher boulder densities in craters and slope-breaker zones are consistent with subsurface stratification, indicating that vertical layering and surface reorganization jointly govern rubble-pile evolution and the provenance of returned samples. We propose the first coherent global benchmark, and future global-scale orientation datasets on other asteroids are critical for advancing comparative asteroid studies.

Conclusions

This study presents the most extensive global-scale mapping and the first orientation dataset of boulders on Ryugu. When we combine boulder orientation patterns with the analysis of latitudinal size sorting by Shimizu et al.¹⁷, we see how Ryugu’s boulders preserve a more complete record of rubble-pile evolution than previously recognized. The two datasets together reveal that asteroid surfaces not only redistribute material under the influence of spin state but also record directional signatures that function as geological evidence of past YORP-driven change. This combined evidence is essential for interpreting returned samples (i.e., tracing back their origins) and for anticipating surface conditions on other rubble-pile asteroids. The steep cumulative SFD slope (− 2.665 ± 0.066) and abundance of large blocks support an origin from catastrophic parent-body disruption, rather than impact ejecta. Spatial variations in boulder density reflect the combined effects of YORP-driven spin evolution, impact blanketing, and localized heterogeneity. Low boulder densities along the equatorial ridge suggest poleward migration during Ryugu’s current spin-down phase, while slope-breaker zones exhibit the highest densities. Orientation patterns show hemispheric trends; NW–SE in the north, NE–SW in the south, most prominent in slope-breaker zones; indicating spin-induced reorientation during downslope migration. The western bulge, characterized by a shallower SFD and low crater density than the equatorial ridge, is likely a less impact disturbed or structurally relaxed part of Ryugu. Slower space-weathering rates inferred from young craters further highlight the influence of orbital and thermal environment on surface evolution.

These findings demonstrate that Ryugu’s boulders are not random debris but relicts of catastrophic disruption, rotational reshaping, and impact resurfacing. By studying orientation and clustering intensity, we also propose sites of possible mass wasting. These results establish the first global link between YORP-driven spin history and boulder-scale surface organization on a rubble-pile asteroid. Our global mapping shows that many boulders were likely re-exposed from subsurface layers, which suggests that the returned samples also include material recycled through stratification and impact excavation. Beyond recording past events, the dataset provides a framework for assessing surface stability, identifying hazardous zones, and guiding future landing site selection and exploration strategies on small bodies in the solar system.

Data and methods

Image and other datasets

Boulder mapping was conducted using four image mosaics prepared from images acquired by the Optical Navigation Camera–Telescopic (ONC-T) aboard the Hayabusa2 spacecraft^42,43. These are a global mosaic (hyb2_onc_mos_Global_01.tif; resolution: 0.63 m/pixel), a mid-altitude (40°N-40°S) mosaic (hyb2_onc_mos_20180801_MidAlt; resolution: 0.55 m/pixel), and a high-resolution mosaic (25°N-25°S) from the MASCOT lander module (hyb2_onc_mos_20181003_MascotSeparation; resolution: 0.25 m/pixel). For high-latitude regions (40°N-90°N and 40°S-90°S), north and south polar projection mosaics (resolution: 1 m/pixel) were used to minimize distortion induced by artifacts. The equatorial and high-resolution mosaics were projected in a simple equirectangular (Plate Carrée) projection with a central meridian of 0°, defined on a custom ellipsoidal reference body approximating Ryugu’s spinning-top shape (semi-major axis 502.0 m, semi-minor axis 437.744 m, inverse flattening 7.8125). This corresponds to a non-conformal geographic coordinate system (GCS_unnamed_ellipse; WKID 0) with angular units in degrees and a prime meridian at 0°. Polar-region mosaics were projected using north and south polar stereographic projections centered on the respective poles, adopting a spherical reference body (GCS_Ryugu; WKID 0) with equal semi-major and semi-minor axes of 449.0 m. In all cases, geographic north was defined as the direction toward the positive Z-axis of the Ryugu body-fixed coordinate frame of the SFM20180801 shape model. Boulder long-axis orientations were measured as bearings relative to geographic north in the native projection frame of each mosaic. The coordinate reference systems (CRS) in WKT format are respectively.

GEOGCS[“GCS_Ryugu_ellipse”, DATUM[“D_Ryugu”,

SPHEROID[“Ryugu”, 502.0, 7.8125]], PRIMEM[“Reference Meridian”, 0.0],

UNIT[“degree”, 0.0174532925199433]] and.

GEOGCS[“GCS_Ryugu”, DATUM[“D_Ryugu”,

SPHEROID[“Ryugu polarRadius”, 449.0, 0.0]], PRIMEM[“Reference Meridian”, 0.0],

UNIT[“degree”, 0.0174532925199433]]. We also prepared a B-x composite from ONC-T data by taking the band ratio of the b (0.48 μm, blue) and x (0.86 μm, near-infrared) bands from the multiband composites⁴⁴. This ratio is sensitive to thermal exposure and has been used as a proxy for surface age or exposure state, as bluer tones in the B-x composite generally indicate freshly exhumed surfaces with minimal space-weathering, whereas redder tones correspond to older, more weathered surfaces⁴⁵.

Identification and mapping of boulders

We defined a boulder as any isolated positive relief surface feature visible in optical imagery and discernible from surrounding regolith. Mapping was conducted manually in ArcGIS Pro (v3.3.0), where maximum pixel dimensions of boulders were digitized as straight lines representing their long axis and the midpoints were recorded as positional data. A minimum three-pixel sampling rule⁴⁶ was considered for boulder mapping, resulting in the boulder lengths being ≥ 0.75 m for the equatorial ridge, ≥ 1.65 m for mid-latitudes, and ≥ 3 m for polar regions. Features smaller than these limits were regarded as regolith rather than individual boulders. Boulders partially buried in regolith were included only if their maximum dimensions remained visually distinguishable in all mosaics. Fractured but contiguous boulders were mapped as single entities, as complete separation had not occurred³¹. Regions affected by extreme shadowing, NoData zones, or geometric distortion (~ 7% combined; mostly near poles) were excluded from mapping to maintain data reliability (refer to Data and Methods). To assess potential bias from near-circular boulders (b/a ≈ 1), we analysed 6138 boulders (~ 15% of dataset) from three representative strips (0–5°N, 30–35°S, 0–5°E). Ellipse fitting of boulders revealed only 0.6% have b/a > 0.95, confirming that they are statistically insignificant and do not influence population statistics. Furthermore, the statistically significant orientation patterns reported in this study are dominated by larger boulders (≥ 3 m), which preferentially exhibit elongated shapes with well-defined long axes (Supplementary Data S1).

Orientation measurement

The orientation (bearing) of each boulder’s visible long axis was calculated relative to geographic north in the respective image projection⁴⁷. Orientations are point-symmetrical with north defined as 0° and angles increasing clockwise. For polar regions, orientations were measured directly in the native polar map projections; preventing the geometric distortions induced in global coordinate frames at high latitudes. Although true 3D orientation cannot be determined from surface images, it was stated by Michikami et al.¹⁰ that most boulders on Ryugu lie parallel to the surface and thus orientation of apparent long axes would not deviate much or change observations drastically. Our measurement approach aligns with the a’-axis orientation framework established by Michikami & Hagermann²⁹.

Geometric verification

Geometrically corrected mosaics were used to minimize projection related artifacts associated with Ryugu’s irregular shape^2,47. Some mapped boulders were cross verified using the Small Body Mapping Tool (SBMT)⁴⁸ against the SFM20180801 shape model of Ryugu²to ensure spatial consistency and positional accuracy. To address potential projection artifacts and quantify any systematic bias in boulder orientation measurements arising from the use of 2D mosaics in ArcGIS Pro, we re-measured the long-axis length and orientation of a representative subset of 350 boulders directly in the Small Body Mapping Tool (SBMT) using the SFM20180801 shape model of Ryugu. This subset was stratified across latitude bands to capture possible variations related to projection effects near the poles and equatorial ridge. Each boulder was digitized as an ellipse in SBMT, and the resulting long-axis length and orientation were compared with the corresponding values obtained from manual line digitization in ArcGIS Pro.

The comparison yields a mean length deviation of 3.8505% (median = 3.5407%, 95% interval = [− 8.5083, 16.5386] %) and a mean angular deviation (Δθ = |θ_ArcGIS − θ_SBMT|) of 2.3469° (median = 1.4605°, 95% interval = [0.0685, 9.6153] °).

These differences are small relative to the scale of the observed global orientation signals. No systematic latitudinal increase in Δθ or length deviation is evident, indicating that projection effects in the mosaics do not introduce significant bias into the orientation dataset. The minor discrepancies are consistent with expected manual digitization variability (e.g., slight differences in endpoint placement or ellipse fitting) and do not alter the primary observations or interpretations presented in this study. Further details, including histograms of deviations and scatter plots of SBMT versus ArcGIS values, are provided in Supplementary Fig. S2.

Statistical analysis

A total of 49,536 boulders ≥ 0.75 m were identified across Ryugu. Of these, 23,553 are ≥ 3 m in length, and all global statistical analyses are restricted to this subset. This threshold enables robust characterization of global boulder populations and trends, whereas the equatorial ridge dataset, which extends to smaller size fractions, provides additional constraints on fragmentation processes. Power-law fitting of the cumulative size-frequency distributions followed the methodology outlined by Michikami et al.¹⁰.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

T.R., G.K., H.K. and R.H. planned the study. A.R. and L.M.V. curated the data. T.R. acquired the funding. A.R. and T.R. conducted the investigation. A.R. and T.R. developed the methodology. T.R. and G.K. administered the project. T.R. supervised the work. T.R., G.K., R.H., and J.O. validated the results. A.R. prepared the manuscript. All authors reviewed and edited the manuscript.

Funding

TR was supported by the KAKENHI grant no-24KK0078. The work by JO was supported by grant PID2021-125883NB-C22 by the Spanish Ministry of Science and Innovation/State Agency of Research MCIN/AEI/10.13039/501100011033 and by ‘ERDF A way of making Europe’.

Data availability

The datasets used and/or analysed during the current study are available here (https://figshare.com/s/12d284cd8a0d610861f8).

Declarations

Competing interests

The authors declare no competing interests.