RETRACTED: Geo‐archaeological prospecting of Gunung Padang buried prehistoric pyramid in West Java, Indonesia

The multidisciplinary study of Gunung Padang has revealed compelling evidence of a complex and sophisticated megalithic site. Correlations between rock stratifications observed through surface exposures, trenching and core logs, combined with GPR facies, ERT layers, and seismic tomograms, demonstrate the presence of multi‐layer constructions spanning approximately 20–30 m. Notably, a high‐resistive anomaly in electric resistivity tomography aligns with a low‐velocity anomaly detected in seismic tomography, indicating the existence of hidden cavities or chambers within the site. Additionally, drilling operations revealed significant water loss, further supporting the presence of underground spaces. Radiocarbon dating of organic soils from the structures uncovered multiple construction stages dating back thousands of years BCE, with the initial phase dating to the Palaeolithic era. These findings offer valuable insights into the construction history of Gunung Padang, shedding light on the engineering capabilities of ancient civilizations during the Palaeolithic era.


| INTRODUCTION
Gunung Padang, located in Cianjur District, West Java Province, Indonesia, has been the subject of comprehensive archaeological, geological, and geophysical studies.(Figure 1, Index Map 1).Early descriptions by Veerbek (1896) and Krom (1915) described it as an ancient cemetery on top of the mound, but further investigations did not take place until local reports prompted government attention in 1979 (Bintarti, 1982).The name Gunung Padang translates to 'mountain of enlightenment' in the local language, as it has been used for religious rituals throughout history (Akbar, 2013).The National Archeological Institute conducted studies that led to the site's restoration in 1985, and in 1998, Gunung Padang was designated a Provincial-level Cultural Heritage Site.Previous studies regarded it as a significantly large megalithic site consisting of stone terraces (Bintarti, 1982;Ramadina, 2010;Sukendar, 1985), known as punden berundak, which are common in Indonesia but not on the same scale as Gunung Padang.Further archaeological studies were carried R e t r a c t e d out until 2005, including limited excavation pits that reached less than a meter deep (Tim-Peneliti, 2003).While lacking radiocarbon dating, it was assumed that Gunung Padang was a prehistoric site built between several hundred and a couple of thousand BCE, following regional megalithic cultures in Asia (Kim, 1982).Situated in the southern mountainous ranges of West Java, near the headwater of the Cimandiri River, Gunung Padang is surrounded by other megalithic sites such as Kujang 1 and 2, Cengkuk, Arcadomas and Lebak Cibedug step pyramid (Figure 1, Index Map 2).The site is located within Mio-Pliocene volcanic rocks, comprising pyroclastic, epiclastic, basaltic-andesite lava and intrusive rocks.To the north, the mountainous terrain consists primarily of Quaternary volcanic products from active volcanoes.The presence of the Cimandiri Active Fault Zone near the site poses earthquake hazards to the region (Irsyam et al., 2020;Marliyani et al., 2016).
Indonesia's tropical climate, characterized by intense weathering and sedimentation processes, combined with dense vegetation, has led to the burial and concealment of ancient cultural remains.
F I G U R E 1 (a) Aerial view of Gunung Padang taken from a helicopter.(b) Topography and site map generated from a detailed geodetic survey.(c) Geology map of the Gunung Padang region (Sudjatmiko, 1972).(d) Orthophoto map obtained from a drone survey conducted in 2014, indicating the locations of trenching sites (white rectangles) and core-drilling sites (red dots).T1, Terrace 1; T2, Terrace 2; T3, Terrace 3; T4, Terrace 4; T5, Terrace 5.

R e t r a c t e d
Furthermore, Indonesia's archipelago has undergone drastic climate changes over the past 15 000 years, resulting in rising sea levels and submerging ancient land known as Sundaland (Sathiamurthy & Voris, 2006;Voris, 2000).These dynamic natural processes have caused the disappearance of numerous ancient heritage sites in forests, underwater and buried underground.Traditional archaeological methods face significant challenges in their discovery.This study exemplifies how a comprehensive approach integrating archaeological, geological and geophysical methods can uncover hidden and vast ancient structures.
The field survey of Gunung Padang began in October 2011 and continued until October 2014, comprising multiple seasons.The survey encompassed detailed mapping, geological and archaeological observations, shallow geophysical surveys, excavation trenches and core drillings.It is one of the most extensive and integrated archaeological, geological and geophysical studies conducted on a buried ancient structure.The studies indicated that Gunung Padang is not merely a simple prehistoric stone terrace (e.g.Bintarti, 1982;Yondri, 2017) but a complex underground construction with substantial chambers and cavities.Carbon dating analysis indicates that it may have been constructed during the last glacial period in the Palaeolithic era, with subsequent modifications during the Holocene or Neolithic era.
The early publication of these findings in mass media outlets, along with public lectures and conferences (Natawidjaja et al., 2016(Natawidjaja et al., , 2018)), has garnered significant attention and popularity for Gunung Padang nationally and globally.Consequently, the Ministry of Education and Culture issued Decree in 2014, elevating the site's status from provincial to national heritage.The strength and significance of this study lie in the comprehensive and integrated use of multiple techniques to explore the buried and expansive ancient structures at Gunung Padang.

| Integrated multi-method studies
The integrated surveys at Gunung Padang were conducted for 3 years, from November 2011 to October 2014 (Natawidjaja, 2015(Natawidjaja, , 2016(Natawidjaja, , 2017)).These surveys involved a combination of detailed landscape and surface mapping, core drillings, trenching and integrated geophysical techniques involving two-dimensional (2D) and threedimensional (3D) electrical resistivity tomography (ERT), groundpenetrating radar (GPR) and seismic tomography (ST).The use of multiple methods allowed for cross-validation and enhanced interpretation of subsurface structures.
We utilized an IFSAR-5m Digital Surface Model (DSM) and Digital Terrain Model (DTM) to analyse the regional topography.These data provided a high-resolution landscape representation, enabling us to identify local features, including exposed megalithic stones and existing infrastructure.To achieve more precise mapping of these features, we conducted a geodetic survey using total stations (Figure 1b).
Additionally, we employed a small drone to perform 3D aerial surveys (Figure 1d,.The drone-captured images were used to develop a high-resolution Digital Elevation Model (DEM), a geographically referenced mosaic orthophoto and a 3D image of the Gunung Padang mound.

| Geo-archaeological trenching
The

| Core drilling
The core drilling activities were undertaken to explore the deeper rock layers.For this purpose, we employed Jacro 100 drilling equipment equipped with a diamond bit NQ measuring 2 in. in diameter and 5 ft core barrels (Figure SC.7).The collected rock cores were subjected to petrological and petrographic analysis to gain insights into their composition and characteristics.Additionally, the drilling operations allowed us to delineate the interface between the rock formations and groundwater, providing valuable information about the hydrological aspects of the site.Organic soil samples were carefully extracted from the spaces between rock fragments, which were subsequently used for carbon dating analysis.In specifically targeted locations, drilling activities aimed to explore suspected large underground cavities.
These drilling operations were executed carefully and cautiously to ensure that no megalithic stones exposed on the surface were disturbed or removed.Suitable open spaces were selected for drilling, and customized wooden constructions were utilized as foundations for the drilling equipment (Figure SC.7c).

| Radiocarbon analysis
This study represents the first comprehensive analysis of 14 C dating at the Gunung Padang site.Organic soil samples obtained from the drill cores and the trenching walls were meticulously selected for 14 C dating analysis.These organic samples were believed to contain traces of bio-organic activities during and after the construction phases.
However, it is essential to consider potential sources of contamination, such as older carbon sources or recent bio-organism activities,

R e t r a c t e d
which could impact the dating results.Notably, modern vegetation and roots were identified as the most common sources of contamination, leading to significantly younger age determinations (Natawidjaja et al., 2017).Great care was taken during sample collection to avoid modern vegetation.
Additionally, any remaining carbons derived from modern vegetation were separated and thoroughly cleaned during laboratory processing.Most samples were analysed using the accelerator mass spectrometry (AMS) dating method at the BETA Analytic Lab in Florida, USA.Some samples were analysed using the conventional carbon dating method at the National Nuclear Energy Agency (BATAN).The OxCal program was utilized to calibrate the results of the conventional 14 C ages and to conduct a robust chronological analysis (Bronk Ramsey, 2016).

| Shallow geophysical prospecting
The application of high-resolution shallow geophysical methods in archaeological studies has grown significantly over the past two decades (e.g.Tsokas et al., 1994).However, the extensive use of geophysical prospecting to investigate buried and expansive ancient structures, particularly pyramids, remains uncommon (e.g.Papadopoulos et al., 2010;Tejero-Andrade et al., 2018).
Geophysical surveys that combine extensive excavations and core drillings to validate and refine interpretations of the imaged geophysical features are relatively rare (e.g.Shi et al., 2020).Most archaeological prospecting efforts have focused on uncovering smaller buried structures or features ranging from tens of centimetres to several meters in scale, such as tombs (e.g.Sarris et al., 2007).
The research conducted at Gunung Padang breaks new ground by employing multi-high-resolution geophysical methods on a large scale, addressing the challenges associated with investigating vast ancient structures.By integrating these methods with extensive excavations and core drillings, the study offers a unique and comprehensive approach to exploring the hidden complexities of the site.This pioneering methodology provides valuable insights into the nature and construction of the structures, surpassing the limitations of traditional archaeological prospecting techniques focused on more minor features.

| GPR survey
For the GPR survey, we utilized SIR-2000 and SIR-3000 GSSI units equipped with various antennas, including unshielded multiple low frequency (MLF) and shielded 100-and 270-MHz antennas (https:// www.geophysical.com).However, the field conditions posed limitations on the GPR survey.It was not feasible to conduct high-slope survey lines due to issues with topographic correction, and heavily vegetated areas required path clearance for effective surveying.
Therefore, this study primarily focused on using GPR on flat, clear ground atop megalithic terraces.
During the survey, we experimented with different antennas and frequencies, primarily focusing on the MLF antenna that was deemed most suitable for our objectives.The MLF antenna was employed at 15, 40 and 80 MHz frequencies.However, it was found that the 40 MHz frequency provided the optimal balance between resolution and depth penetration for our study, reaching depths of up to 30 m.
The 15 MHz frequency did not achieve significant depth penetration on the site while exhibiting lower resolution.Additionally, using the 80 MHz frequency did not yield improved imaging results, leading us to solely present the outcomes obtained with the MLF 40-MHz antenna.
We employed the common-offset measurement technique to acquire data along the survey lines, incrementally capturing point data at intervals of 50-100 cm.These acquired data points were then processed using RADAN software that comes with the GSSI units.We employed various colour spectrums in the radargrams to visualize and interpret the processed data.A neutral conventional grey colour scheme was utilized to display contrasting degrees, textures and patterns, while a red-blue colour spectrum was employed to highlight the positive and negative polarities of radar waves.Bright spots observed in the radargrams corresponded to high-amplitude reflections from subsurface materials with a higher dielectric constant, whereas dark spots or negative polarity reflections indicated materials with a lower dielectric constant.GPR facies analysis was conducted to identify radar stratifications, and the results were subsequently compared to the borehole, ERT and ST data for comprehensive analysis and interpretation.

| ERT survey
ERT data played a crucial role in this study, offering valuable insights into the subsurface structures (Figure 8a

| ST survey
ST is not commonly employed in archaeological prospecting but offers several advantages over conventional seismic refraction methods (e.g.Forte & Pipan, 2008).One key advantage is the superior velocity information obtained from tomography, surpassing the layering models used in conventional seismic refraction.Additionally, the heterogeneities and complexities of the subsurface structures at Gunung Padang may limit the applicability of seismic reflection methods.
Data acquisition was performed using a 2 Â 24 channel seismograph set with a spacing of 5 m.Firecrackers or sledgehammers were utilized as seismic sources, positioned in the middle of each receiver (geophones), and fired at 5-m intervals.Data were continuously downloaded to the hard disk of a portable PC while repositioning the shot point.This configuration allowed for high-resolution ST with deep penetration.The topography of Gunung Padang, resembling an inverted boat, facilitated line configurations that covered the targets at a sufficient aperture angle (Figure SG1A), significantly enhancing the capability of ST to reveal shallow and deep targets, including large cavities.
To achieve stable inversion, we performed the inversion procedure using a wide wavepath and gradually transitioned to a thinner wavepath by sweeping the frequency component.This approach ensured stable tomography inversion, and the smoothed procedure based on the wavepath frequency was adopted in the inversion process.By incorporating a broad spectrum of frequencies for inverse modelling, multiple alternative tomography images with varying resolutions were generated .Comparing these multi-images improved the visualization and interpretation of the subsurface structures.For further details on the method and results, refer to A comparative example of stone terraces is Lebak Cibedug, a stepped pyramid close to Gunung Padang, which shares a similar size and antiquity but has not been extensively studied (Takashi, 2014).
Another example is Candi Kethek in Central Java (Purwanto et al., 2017).Stone terraces constructions can be found worldwide, such as Machu Picchu in Peru, built by the Inca civilization (Bingham, 1930).Nan Madol, a megalithic complex on Pohnpei Island,

| Multi-phase construction and hidden layers
The visual observations at Gunung Padang reveal a complex construction history, with evidence of multiple phases and diverse architectural styles.At the ground surface, known as Unit 1 (#1), the prominent megalithic stones exhibit a variety of arrangement techniques positioned on soils containing numerous andesite rock fragments (Akbar, 2013).Unit 1 comprises columnar rock arrangements of standing rocks and ramps, defining the spatial geometry and terrace spaces (Figure 2c).Notably, interwoven columnar rocks form ascending stone steps from the lower levels, while tall rock walls enclose T1 (Figure 2e).Intriguingly, this study unveils that Unit 1 stone terraces

R e t r a c t e d
extend beyond the hilltop, visible on the east slope and in other cleared areas (Figure 2i).It has been confirmed that these terraces were not recently constructed by local inhabitants for agricultural purposes (Akbar, 2013).Unfortunately, significant portions of the stone terraces on the hill slopes have not been preserved.
Beneath Unit 1 lies an older layer of columnar rocks displaying more sophisticated construction techniques.These regularly cut columnar rocks are arranged like bricks in a building, with fine-grained fillers or mortar between them (e.g. Figure 2d).This layer, called here as Unit 2, remains hidden in plain sight, with exposed parts.
Exposed areas reveal various elements of Unit 2, including (1) the stone-wall ramp between T1 and T2 (Figure 2a direction (Figure 2h).Notably, on T1, the top of Unit 2 lies just a few tens of centimetres beneath the ground surface, as uncovered by the Tango trench (Figure 2f).The interwoven columnar rocks beneath T1 and the ramp connecting T1 and T2 are aligned similarly in the N70 E direction.This evidence confirms that Unit 2 is a product of human construction, challenging previous notions that it consisted solely of natural columnar rocks (e.g.Yondri, 2017).
We have mapped a sizeable ancient flank collapse or large landslide on the west slope (Figure 1b).A part of its head scarp, Beta2 cliff, exposes not just Unit 1 and Unit 2 but also an older unit beneath them.The three layers are parallel to the ground surface (Figure 3).and decay.This layer consists of homogeneous, coarse sands with a gravelly texture, possibly originating from a variety of volcanic rock materials.It is evident that this layer was exposed to the air for a significantly more extended period before being covered by Unit 2. In contrast to the homogeneous mass, there are embedded pillar-like structures that contain columnar rocks.These columnar rocks have experienced advanced spheroidal weathering, resulting in rounded corners and edges caused by exfoliation (Figure 3e).This rock-pillar structure appears to be an integral part of the overall construction.
We designate the homogeneous mass layer as Unit 3A and the decayed columnar rocks as Unit 3B.The observed weathering profile is intriguing since natural weathering typically occurs due to exposure to air and water, resulting in weathering and decay towards the ground surface rather than the opposite.

| Results of trenching
The trenching excavations were conducted at 12 selected sites The geometry of Unit 2 is complex.The alignment of columnar rocks on T1 and the ramp and on the east and west slopes is approximately N70 E, perpendicular to the long axis of the megalithic structure.On T2 and T3 and down to their slopes, as revealed by the Charlie3 and Charlie4 trenches, the alignment of columnar rocks is approximately N55 E. It is worth noting that the stone blocks that resemble bricks are not always columnar rocks.For example, the Charlie2 trench on the east slope reveals neatly packed irregular rock slabs/fragments aligned in a similar orientation to the columnar rocks (Figure 4e).
In Trench Echo2 on T5, a steep wall constructed from interwoven columnar rocks is exposed but buried under a soil fill (Figure 4c).
According to the GP5 core drilling data described in the next section, the thickness of the soil fill is approximately 7 m (Figure 6).The

| Results of core drillings
Core drillings were conducted at seven selected sites around the hilltop (Figure 1d) with varying depths.The deepest core drilling, GP5 on T5, reached 36 m.The drillings penetrated Unit 1, Unit 2and Unit 3, as well as a deeper section composed of massive basaltic-andesite rocks referred to as Unit 4 (Figures 5 and 6).The characteristics of each rock unit are summarized in Table 1. on the west side of the columnar rock truncated line (see Figure 2h).
The GP1 borehole on T2 revealed that Unit 2 had been almost entirely excavated before the soil fills covered the remaining structures.
F I G U R E 6 Summary of all core logs showing stratigraphic units and their correlations.Descriptions of each rock unit can be found in Table 1.
It should be noted that Unit 2 does not extend to T5 and Unit 3 is buried by ancient soil fills at T5.During GP4 drilling, a significant water loss of 32 000 L (32 m 3 ) was observed between 8-and 14-m depth.Evidence of groundwater level was observed through water inflow at a depth of 17 to 20 m in GP1.

R e t r a c t e d
Drilling initially took place at GP2 to a depth of 15 m (Figure 5b).
However, further excavation was carried out at this site, referred to as the Echo1 trench, reaching approximately 11 m (Figure 5d).This excavation exposed ancient soil deposits that buried extensively weathered rocks consisting of basaltic-andesitic rock boulders exhibiting exfoliation due to spheroidal weathering.These boulders were embedded in a highly weathered matrix (#3C).This excavation highlighted the challenge of identifying rock types based solely on core samples.Fresh basaltic-andesite boulders were observed at the bottom of the Echo1 trench, while weathering gradually intensified towards the surface.
Notably, these rounded boulders are typically found in rivers, where they undergo transportation through rolling and abrasion by water streams.However, the presence of these boulders on top of the hill suggests that they were brought up to this location.
The GP4 borehole contains crucial data regarding the suspected presence of large cavities or chambers beneath the surface.During drilling, the penetration rate slowed down at a depth of 5 m.At 7 m, the drill encountered a 'blank zone' with no core samples.Subsequently, at 8 m, there was a sudden and significant water circulation loss exceeding 20 000 L. After this point, the drilling penetration accelerated and maintained a high speed.The water loss continued until the total unreturned water circulation reached approximately 32 000 L (32 m 3 ) at a depth of 14 m, prompting us to halt the drilling.The substantial water loss strongly indicates the presence of a large underground cavity.
The GP5 borehole, strategically positioned near the centre of T5, was initially selected to investigate the suspected presence of a significant cavity or chamber.However, during drilling, no such feature was encountered.To delve deeper into this possibility, we excavated in trench Echo2, situated between GP4 and GP5.Regrettably, the excavation only reached a depth of 4 m, which proved insufficient to confirm the existence of the cavity.Instead, we uncovered a steep rock wall.It is plausible that the large cavity does not extend beneath T5 but rather towards the east slope.Further exploration and excavation in this area are necessary to gather more conclusive evidence.
Boreholes GP7 were also conducted to investigate the presence of a significant primary chamber beneath the centre of the megalith site.
However, drilling was halted at a depth of approximately 21-22 m without discovering any large cavities.Nevertheless, at a depth of 10 m in GP7, we encountered a lengthy andesitic rock core measuring 1.3 m.
This core may indicate the presence of a vertical structure, such as a wall or gate, as supported by the ST profile discussed in the following section.
Considering our goal of drilling directly into the chamber, we contemplated drilling into the steep wall between T1 and T2.However, this task requires special permission since drilling without removing or disturbing the megalithic structures on the surface at this particular location is impossible.
Petrographic analysis of thin sections from the rock core samples obtained from the GP1 borehole was conducted to examine their mineralogical compositions and determine their rock types (Figure SJ).
The results indicate that the mineralogical compositions of the rock cores from #1 to #4 units are similar, consisting of basaltic-andesite rocks.The visible minerals under a microscope include feldspars, The classification of rock units based on comprehensive analysis and interpretation of the collected data obtained from boreholes, trenches and surface exposures.The groundmass consists of microlite plagioclase feldspar, very fine-grained mafic minerals and a glassy matrix, which have undergone partial alteration to chlorite and sericite.The rocks show only slight alterations, and some display 5%-7% porosity due to the presence of vesicular holes.Based on the sizes of microscopically visible vesicular holes, ranging from 0.5 to 2 mm in diameter, these basalticandesite rocks are believed to be part of shallow lava intrusions.There are no significant changes in mineralogy from top to bottom, except that rocks from Unit 4 exhibit a slight increase in more basic minerals, indicating a transition towards more basaltic rocks.Therefore, in the core log, the rocks from Unit 4 are referred to as basaltic rocks for brevity in the descriptions (Figure 5a).

| Results of carbon dating analysis
We collected a total of 12 organic soil samples, with six obtained from borehole core samples and the remaining six from trenches.Out of these samples, eight underwent analysis using the AMS radiocarbon method at BETA Analytic Lab (www.radiocarbon.com),while the other four were analysed using the conventional radiocarbon method at the Geochronology Lab in BATAN.To calibrate the carbon dating results, we employed the OxCal modelling approach (Ramsey, 2007(Ramsey, , 2016)), which utilizes Bayesian statistics and considers factors such as the material being dated, the calibration curve used (in this case, SHCAL20), and the stratigraphical and archaeological context of the dated object.Additionally, prior information and assumptions about age can be incorporated into the analysis.The detailed results of the carbon dating analysis are presented in Table 2, with further information provided in Figure SD.
The comprehensive OxCal analysis, as illustrated in Figure 7, provides invaluable insights into the chronological sequence of the constructions at Gunung Padang.According to the analysis, Unit 3 is estimated to have been constructed during the remarkable timeframe The results of carbon dating analysis conducted as part of the study.The conventional 14C ages obtained from the samples have been calibrated to calendar ages using the OxCal software.The table provides an overview of the calibrated ages for the analysed samples, allowing for a more accurate understanding of the chronological timeline associated with the studied materials.For additional information and details regarding the carbon dating analysis, please refer to Figure SD.

R e t r a c t e d
of 25 000 to 14 000 BCE.Following this period, there was a hiatus spanning from 14 000 to 7900 BCE before Unit 3 was ultimately buried between 7900 and 6100 BCE.Remarkably, approximately two millennia later, the construction of Unit 2 took place between 6000 and 5500 BCE.Another significant hiatus occurred from 5500 to The estimated ages of the units align with their respective degrees of weathering.Unit 1, characterized by relatively fresh columnar basaltic-andesite rock, is estimated to be around 4000-3000 years old, consistent with previous studies (Bintarti, 1982;Tim-Peneliti, 2003).Unit 2, which exhibits significantly more weathered columnar rocks than Unit 1, aligns well with the estimated age of 7500-8000 years.The estimated age of Unit 3, at least 16 000 years old, corresponds to the extensive decay and exfoliations observed in its rocks due to spheroidal weathering.

| Results of GPR survey
We conducted more than 30 survey lines using the MLF 40-MHz

R e t r a c t e d
between facies are often marked by strong negative or positive reflectors (Table 3, Figure 9).For further details, refer to Table 3 and Facies C is distinguished by its parallel and discontinuous reflector pattern, ranging from weak to strong.Within this facies, strong reflectors occasionally denote its upper boundary.
The primary south-north cross-sectional radargram along the long axis of the megalith reveals that the subsurface layering generally mirrors the topographic profile of the ground surface (Figure 9).In simpler terms, the underground layers appear horizontal beneath the flat ground surface and inclined beneath sloping surfaces, such as the ramp between T1 and T2.The overall geometry of the subsurface structures along the long axis is depicted in Figure 9b, with the more detailed information provided by short survey lines conducted on each terrace, as illustrated in Figure 9a.Additional radargrams can be found in the supplementary data .
The radar facies exhibit a strong correlation with the stratification of rock units (Figure 13, Figures SE2-7 and SH4).Interestingly, the expected high-resistivity anomaly of the massive andesite is not visible in the ERT sections at T5. Instead, a moderate-to low-resistivity zone is observed from 5 m down to over 30 m.In contrast, during the GP5 drilling, the massive andesite was encountered at a depth of 30 m.We propose that this discrepancy is due to a hydrological anomaly where high-pressure artesian water from below has fully saturated the highly fractured andesite lava.This saturation has transformed the andesite into a conductive or low-resistivity layer, thus obscuring the anticipated highresistivity anomaly in the ERT sections (Park et al., 2016).The water inflow observed during the GP1 drilling, specifically between 17 and 20 m deep within the weathered upper part of the massive andesite (Unit 3), supports this hypothesis.Additionally, the presence of a The results of the radar facies analysis conducted on the radargrams obtained from the GPR survey using the MLF 40-MHz antenna.The table provides a classification and description of the various radar facies identified in the radargrams.Each radar facies is characterized by distinct textures, patterns and structures, allowing for a better understanding of the subsurface features and their interpretations.The table serves as a reference for the interpretation and analysis of the GPR data.

| 3D ERT survey prospecting underground chambers
To further investigate the extremely high-resistive anomaly (EHRA) identified in the 2D ERT sections, a 3D ERT survey was conducted, covering the entire expanse of the megalithic terraces on the hilltop.
The survey employed 112 steel electrodes, strategically arranged in four parallel lines with a 5-m spacing between both electrodes and lines, resulting in a rectangular survey area measuring 15 Â 135 m (Figure 11g).By conducting a 3D survey, we were able to gather data with significantly higher density compared to the 2D survey, as the 3D approach captured volumetric information rather than just a 2D section.Consequently, the acquisition time for the 3D survey was significantly longer, spanning the entirety of the day, in contrast to the

| Results of ST survey
We conducted three survey lines (Figure 12a).The initial purpose of the ST survey is to further investigate the large cavities the ERT imaged.
However, the results also revealed other significant findings.The ST sections exhibited stratifications that correlated with the ERT and GPR

| Integrating data
This study demonstrates the effective utilization and integration of multiple techniques to explore the complex, multi-layered and exten- The GPR survey in this study employed a simplified approach by assuming a single average electromagnetic (EM) velocity for all subsurface layers due to the limitations of the software.This assumption introduces uncertainties in accurately characterizing and distinguishing subsurface features and boundaries based solely on the GPR data.
Using a single average EM velocity also limits the resolution and accuracy of the imaging results.More sophisticated approaches are required to capture the complexity of subsurface structures.
The ERT method is also subject to uncertainties, particularly in interpreting resistivity data.The relationship between resistivity and subsurface properties is complex, influenced by various factors The oldest construction, Unit 4, likely originated as a natural lava hill before being sculpted and then architecturally enveloped during the last glacial period between 25 000 and 14 000 BCE. (Figure 14).
Afterward, Gunung Padang was abandoned by the first builders for thousands of years, leading to significant weathering.Around 7900-6100 BCE, Unit 3 was deliberately buried with substantial soil fills.
Approximately a millennium later, between 6000 and 5500 BCE, a subsequent builder arrived at Gunung Padang and constructed Unit 2. Lastly, the final builder arrived between 2000 and 1100 BCE, constructing Unit 1.
It is intriguing to note that during the construction of Unit 1, Unit 2 likely remained relatively intact and well preserved.However, in a peculiar turn of events, Unit 2 was subsequently buried, possibly to conceal its true identity for preservation purposes.As a result, Unit geo-archaeological trenching activities at Gunung Padang aimed to understand better the vertical profile and lateral extensions of the buried structures near the surface.The selection of trenching sites was based on the interpretations derived from the preceding geophysical surveys.The trenching operations commenced mid-2012, with most work conducted in August-September 2014.Trench sizes varied, ranging from 1 Â 2 to 3 Â 9 m on the surface, and depths reached between 2 and 4 m, except for Echo1, which was excavated to 11 m.The trenches were manually dug using various tools, including spades and hoes.The trenches were carefully backfilled upon completing the excavations, and measures were taken to prevent erosion by replanting the surfaces.
Micronesia, utilizes similar columnar-joint rocks(McCoy et al., 2016;McCoy & Athens, 2012).Interestingly, based on oral traditions, the Saudeleur Dynasty, the newcomers to Pohnpei Island, is believed to have constructed NanMadol (McCoy et al., 2016).The pronunciation of 'saudeleur' is remarkably similar to the Sundanese word 'sadulur', meaning 'one family' in the local language of West Java, which is significant considering Gunung Padang's location.Some columnar rock arrangements observed in Gunung Padang resemble those found in Nan Madol (Figure2c).
Surface exposures of megalithic stones illustrating two units of construction, Unit 1 (#1) and Unit 2 (#2).(a) South-facing view of T1 landscape, revealing a stone floor and standing columnar rocks of #1, as well as the exposed #2 ramp and altar.(b) Northfacing view from T2 onto T1.(c) Columnar rock arrangement example of #1.(d) Unit 2 on the ramp between T1 and T2, showcasing columnar rock fragments enclosed in a fine-grained mortar.(e) Columnar rock wall of #1 construction encircling the margin of T1.(f) Tango trench on T1 exposing the thinly buried #2 columnar rock, aligned in N70 E similar to those on the ramp.(g) Southward view of T2, T3, T4 and T5.(h) Overhead photo of T2 and T3, demonstrating the N55 E alignment of #2 columnar rock with a truncation line.(i) GPR survey uncovering #1 step-stone terraces on the east slope.
,d), (2) an altar-shaped rock mound situated in the middle of T1, referred to as Masigit (meaning 'a place for praying') (Figure 2a, b), and (3) diagonally oriented tiles of columnar rocks spanning T2 and T3 in an approximately N55E Unit 1 has many rock fragments of broken columnar-joint basalticandesitic rocks.Beneath it, Unit 2 is the interwoven columnar-joint basaltic-andesitic rocks aligned in approximately N70 E direction, like those exposed on T1.The columnar-joint rocks, unlike in nature, are stacked parallel to the layer, not perpendicular, and have been regularly cut about 1-2 m long on average.Furthermore, natural columnar-joint rocks are more homogeneous in size and shape and are tightly interlocked, but Beta2 exposure exhibits columnar rocks with different shapes and diameter sizes, and their surfaces do not interlock directly but are separated by fine-grained mortar (Figure3b,c).The orthogonal sections of the columnar rocks vary in shape, including hexagonal, pentagon, rectangular, trapezoidal and trigonal.There are also distinct thin-flat rocks between the columnar blocks, probably added to tighten the structure (Figure3d).Like exposures on T1, the mortar between the columns has an average thickness of 5 cm.The columnar rocks are only slightly weathered, showing sharp corner edges.The Beta2 scarp is critical in providing evidence of Unit 2's foundation.It reveals a distinct and sudden boundary where Unit 2 meets the underlying rock layer, which has undergone extensive weathering Figure 1b,d), each with its designated name.These sites include Tango on T1, Alpha on the west side of T5, Charlie1, Charlie2, Char-lie3, Charlie4 and Charlie5 on the East Slope, Delta on the south slope, Echo1 near the southern edge of T5, Echo2 on T5 and Fanta at T2.The trenching excavations revealed that Unit 2 is a substantial construction that extends vertically and laterally along the west and east flanks.This indicates that the megalithic stones on the ground surface (Unit 1) are built upon the buried and more massive Unit 2. Unit 1 appears to be a surface re-arrangement of Unit 2. In the Fanta Trench at T2, the alignment of columnar rocks observed on the surface continues underground (Figure 4,A2,A3; Figure SB.2).Although the trench reached a depth of 3.5 m, it did not reach the base of Unit 2, which can be seen at Beta2 on the west slope (Figure3).Unit 2 extends down to the east slope, as revealed by the Charlie 1-5 trenches (Figure4,B1).The layer of columnar rock dips at an angle of about 15 on the east and west slopes, while both flanks incline 30 (Figure4,B2).The flat surfaces of the east and west hill flanks represent the top of Unit 2, which is buried beneath 1-2 m of soil containing numerous broken basaltic-andesite rock fragments.
columnar rocks in this trench have undergone extensive weathering, exhibiting spheroidal weathering similar to those in Unit 3B of the Beta2 scarp.Hence, it is considered part of Unit 3.Although Echo2 was excavated to a depth of 4 m, the base of the columnar rock wall was not reached.The Delta trench on the south slope of T5 exposes a 3-m-thick layer of homogeneous soil fill that buries decayed and unrecognizable R e t r a c t e d rocks, characterized by large rounded rock fragments instead of columnar rocks.These rocks display intensive concentric exfoliations indicative of spheroidal weathering.This rock layer, classified as part of #3, is named #3C.On top of this buried decayed rock mass, a unique stone artefact resembling a traditional Sundanese dagger called Kujang Stone was discovered (Figure 4d,D1).It was found alongside some granular quartz crystals not associated with the weathered rocks beneath it.
Figure 6).Unit 2, referred to as #2, is easily identifiable as it can be seen both on the ground surface and in the trenches.It consists of columnar basaltic-andesite rocks held together by a sandy-silt mortar of 2100 BCE, followed by the construction of Unit 1 between 2000 and 1100 BCE.Lastly, an intriguing excavation of Unit 2 and subsequent soil fills transpired between 1393 and 1499 CE.These refined chronological estimates provide a deeper understanding of the temporal development and evolution of the structures at Gunung Padang throughout its extensive history.
antenna, including the long continuous survey through the megalithic site's longitudinal axis (Figure 8b, Figure SE.2).The main findings of the survey are presented in Figure 9, where radargram stratifications are classified into distinct GPR facies based on established analysis methods ( Ekes & Hickin, 2001; Lanzarone et al., 2019; Lee et al., 2005).These facies exhibit unique textures, patterns, amplitudes, frequencies and reflector-geometrical shapes.The boundaries F I G U R E 7 The OxCal analysis of the multi-construction histories of the Gunung Padang pyramid, including soil-fill burials and time gaps between constructions.(a) Stratigraphic model and carbon-dating samples associated with each layer.(b) Results of the OxCal modelling, showing the estimated dates for each construction phase.(c) Summary of the OxCal analysis results, providing an overview of the chronology of the constructions.

Figure
Figure SD.Facies A exhibits strong amplitudes of parallel reflectors extending to a depth of approximately 5 m.Within Facies A, two subfacies can be identified.Facies A-1, found in the upper 1-2 m, are characterized by continuous and parallel solid reflectors.At a depth of 2-5 m, Facies A-2 displays more irregular and wavy reflector patterns.However, on radargrams, it may be challenging to distinguish between Facies A-1 and A-2.Strong thick negative reflectors often underlie Facies A, serving as its boundary.Facies B, spanning from 5 to at least 15 m in depth, exhibits low amplitudes and indicates smooth textures.Within Facies B, there are two subfacies.Facies B-1 is characterized by a homogenized lowamplitude smooth texture, occasionally adorned by slightly stronger, discontinuous low-frequency reflectors.As we descend, Facies B-2 becomes dominated by discontinuous wavy bands of higheramplitude reflectors.The presence of strong positive and negative reflectors often demarcates the boundary between Facies B-1 and B-2.
Figure 13).Starting from a depth of approximately 15-20 m, Facies C correlates with Unit 4, consisting of massive basaltic andesites.
Figure SF.10h-k), interpreted as the neck of a volcanic intrusion that flowed northward along the pre-existing slope of the ancient extinct volcano.The extruded magma, characterized by its thick and viscous nature, rapidly solidified, forming a lava tongue.The most intriguing feature is the presence of extremely high resistance anomalies (EHRA), exceeding 20 000 ΩÁm and even reaching beyond 100 000 ΩÁm.The most prominent EHRA is found beneath the ramp and T2 (Figure 10a, Figures SF.7, SF.8 and SF.11), providing solid evidence of a large cavity or chamber.

F
Figure 10.Similar to the 2D imaging, these large cavities or chambers were distinguished by the EHRA but in 3D (Figure 11a-c).The presence of EHRA, accompanied by surrounding layers of low resistivity, clearly outlined the chambers.The main chamber beneath the ramp between T1 and T2 was estimated to have dimensions of approximately 10 Â 10 Â 15 m (width Â height Â length).Moreover, EHRA was also observed beneath T1 and T5, further supporting the presence of additional chambers (Figure 11c-f).
Figure SG.4).It may also indicate possible large tunnels.This LVA corresponds to a location similar to the prominent EHRA observed in the 2D and 3D ERT surveys.All lines of source-receiver configurations are adequate to capture the objects of interest.Specifically, for Line-2 and Line-3, the aperture angles of the configurations are sufficiently large, enabling effective imaging of the targeted LVA within the highvelocity Layer.The top of the high-velocity layer sharply defines the surface of the massive andesite, which lies at an approximate depth of 17 m sive ancient constructions at Gunung Padang.The combination of surface observations, trenching, core drillings and geophysical surveys, including GPR, ERT and ST, has not only confirmed but also complemented each other, providing a comprehensive understanding of the site.The multi-layered nature of the constructions aligns well with the profiles of GPR facies, the layering identified in ERT and the stratifications revealed by ST (Figure13), strengthening the overall interpretation.Specifically, Unit 1 and Unit 2 correspond to GPR Facies A1 and A2, reflected in the upper low-velocity layers observed in ST and roughly correlate with Layers a and b in the ERT sections.The presence of Unit 3 is strongly supported by GPR Facies B, the intermediate seismic velocity layer and Layers b and c in ERT.Remarkably, the top of Unit 3B, located at a depth of approximately 9-11 m, closely aligns with the top of GPR Facies B2 and the top of the highresistivity Layer c in ERT.Unit 4, composed of massive basaltic andesite, strongly correlates with the high-seismic velocity layer observed in ST and the GPR Facies C. The top of the high-seismic velocity layer accurately reflects the top of Unit 4, further validating its presence and characteristics.F I G U R E 1 3 Data correlations of borehole log-ground-penetrating radar (GPR)-seismic tomogram (ST)-electric resistivity tomogram (ERT).The core log is obtained from GP1.The radargram is taken from Line-05 in close proximity.The ST profile is derived from Line-1 (Figure 12b), and the ERT profile is based on Line NS-11 (Figure 10a).Unit 1 and Unit 2 are correlated with Radar Facies A and ST's upper low-seismic layer.Unit 3 is correlated with Radar Facies B and ST's intermediate-velocity layer.Unit 4 is correlated with Radar Facies C and ST's high-velocity anomaly (LVA).The top of ERT's high-resistive anomaly (Layer-c) is aligned with the top of Unit 3's lower part (#3B and #3C) and the top of Facies B-2.R e t r a c t e d3.9 | Uncertainty of geophysical measurementsThe geophysical measurements conducted at Gunung Padang provide valuable insights into the subsurface structures and ancient constructions.However, it is crucial to acknowledge the inherent uncertainties associated with the measurements.Complex 3D features, groundwater levels, moisture content and equipment limitations can introduce uncertainties in interpreting geophysical subsurface structures and predicting their depths.For example, this study demonstrates that the geophysical layers do not necessarily align with the lithological stratigraphy as depicted from borehole data.The discrepancies highlight the need for caution when interpreting the results and emphasize the importance of considering multiple factors and approaches in the analysis.Additionally, 2D acquisition methods, compared to 3D, can introduce limitations in interpreting geophysical data.2D surveys provide valuable insights into the subsurface but represent a simplified representation of the complex 3D subsurface structures, leading to ambiguities in the characterization of subsurface features.The adoption of 3D survey and 2.5 D (2D on gridded lines) surveys, as conducted in this study, offers a better representation of the subsurface, reducing some of the uncertainties associated with 2D acquisition.
as described above.Assumptions made during inversion algorithms and the homogeneity of subsurface properties can also introduce uncertainties in accurately characterizing subsurface structures based on resistivity values.As observed in this study, the absence of a high-resistive layer of the massive basaltic andesite body beneath certain areas exemplifies the challenges in accurately interpreting resistivity data.Although powerful for subsurface imaging, the ST method is also associated with uncertainties.The inherent complexity of the subsurface, such as variations in lithology, fluid content and fractures, can introduce challenges in accurately estimating seismic velocities and F I G U R E 1 4 Simplified reconstruction of Gunung Padang.Unit 1 represents the surficial stone terraces constructed between 2000 and 1100 BCE or more recently.Unit 2 (highlighted in yellow) corresponds to a buried pyramidal-shaped layer composed of columnar rocks and was built around 6000-5500 BCE.Unit 3 (shown in green) dates back to 25 000-14 000 BCE. Unit 4 represents the sculpted massive basalticandesite lava.R e t r a c t e d imaging subsurface structures.Factors such as sensor coverage, frequency and quality of seismic waves and assumptions made in inversion algorithms further contribute to uncertainties in the derived velocity models.As demonstrated in this study, integrating multiple geophysical methods in a comprehensive survey helps mitigate the uncertainties associated with individual techniques.By comparing and confirming the results from different methods, the accuracy and certainty of the interpretations are enhanced.This comprehensive approach and integration of geophysical methods contribute to a more robust understanding of the subsurface structures, minimizing uncertainties and providing a reliable basis for further analysis.Finally, it is important to note that the borehole core logs and trenching data are superior for calibrating the subsurface geophysical interpretation, adding further confidence to the findings.4| CONCLUSION4.1 | Gunung Padang is a multi-layered prehistoric pyramidThis study strongly suggests that Gunung Padang is not a natural hill but a pyramid-like construction.The pyramid's core consists of meticulously sculpted massive andesite lava (Unit 4), enveloped by layers of rock constructions (Unit 3, Unit 2 and Unit 1).The carbon dating analysis further supports the multi-layer construction's long history, spanning successive periods.

2
now lies concealed beneath Unit 1, which comprises simple superficial stone terraces or punden berundak representing the latest visible manifestation of Gunung Padang.4.2 | Concluding remarks and further studiesThis study sheds light on advanced masonry skills dating back to the last glacial period.This finding challenges the conventional belief that human civilization and the development of advanced construction techniques emerged only during the warm period of the early Holocene or the beginning of the Neolithic, with the advent of agriculture approximately 11 000 years ago(Harari, 2014).However, evidence from Gunung Padang and other sites, such as Gobekli Tepe, suggests that advanced construction practices were already present when agriculture had, perhaps, not yet been invented.The builders of Unit 3 and Unit 2 at Gunung Padang must have possessed remarkable masonry capabilities, which do not align with the traditional hunter-gatherer cultures.The burial of these structures around 9000 years ago adds further intrigue for reasons not fully understood.Given the long and continuous occupation of Gunung Padang, it is reasonable to speculate that this site held significant importance, attracting ancient people to repeatedly occupy and modify it.To further advance our knowledge of Gunung Padang, it is essential for future research to undertake comprehensive and systematic excavations that delve into the characteristics of Unit 2, Unit 3 and Unit 4, as well as their cultural significance.Employing advanced geophysical imaging techniques and directional drilling can prove instrumental in exploring underground structures, including potential chambers.In the event of encountering a chamber during drilling operations, the use of downhole cameras can provide valuable visual documentation.Furthermore, conducting more extensive radiometric dating studies will contribute to obtaining precise age estimates for the constructions, enhancing our understanding of their historical timelines.Gunung Padang stands as a remarkable testament, potentially being the oldest pyramid in the world.Further investigation and interdisciplinary research will uncover its hidden secrets and shed more light on the ancient civilizations that thrived in this enigmatic site.