In August 2015, Islamic State militants blew up the Temple of Bel at Palmyra, destroying the central chamber of a sanctuary that had stood in the Syrian oasis for nearly two millennia. Its collapse gave an unexpected importance to photographs taken when the building was intact. Archaeologists had made measured records of its walls and decoration, but visitors had also photographed the temple simply because they wanted to remember it. Both kinds of image now preserve views that no new photograph can capture.
Those surviving images have allowed researchers to make three-dimensional reconstructions of the temple and other damaged monuments at Palmyra. By comparing a carved edge or stone joint across multiple pictures, software can estimate camera positions and build a digital representation of the missing surfaces. Earlier laser scans, excavation drawings and satellite images add measurements and dates that photographs alone cannot provide. Such models can show what stood before the explosions, help identify pieces in the rubble and inform a conservation plan. They cannot rebuild a monument by themselves.
Work at Palmyra has revealed how much digital archaeology can recover, and what it leaves uncertain. A measured surface, a proposed reconstruction and an attractive virtual display can look much the same on screen, although the evidence behind them differs. A study published in 2026 found that adding large numbers of tourist pictures made a Temple of Bel model visually richer while reducing the geometric accuracy of parts of it. To understand why that matters, it helps to know what Palmyra was, what happened to it, and which records survived.
Palmyra was a Syrian city long before it became a Roman ruin
Palmyra, known also as Tadmor, occupies an oasis in the Syrian steppe between the Mediterranean coast and the Euphrates valley. Travellers and merchants needed stopping places along routes through this dry region. The settlement appears in sources from the second millennium BCE, though most of the monuments familiar to modern visitors were built during its prosperity under Roman rule. Between the first and third centuries CE, Palmyrene merchants operated networks reaching Mesopotamia and the Persian Gulf, moving goods that had travelled from farther east toward Mediterranean markets.
Prosperity left its mark on the city. A colonnaded avenue roughly 1.1 kilometres long connected temples, public buildings and inhabited districts. The Temple of Bel combined Roman monumental forms with Syrian religious traditions, setting a rectangular sanctuary inside a much larger enclosure. Tomb monuments displayed distinctive portraits of the dead. Inscriptions in Palmyrene Aramaic and Greek name individuals, record dedications and reveal relationships within the community. The UNESCO World Heritage description of Palmyra cites this combination of local and Mediterranean influences as part of the city’s importance. Palmyra cannot be understood simply as a Roman provincial town.
The city was also a political power in its own right. During the troubles of the third-century Roman Empire, Odaenathus and, after him, Zenobia gained exceptional authority. Zenobia’s armies eventually controlled Egypt and other eastern territories, before the emperor Aurelian defeated them in 272 CE. That Palmyrene challenge to Rome helped make Palmyra famous in historical accounts. Yet its streets, inscriptions, sanctuaries and tombs preserve a much longer and more varied history than the brief period of imperial confrontation.
The Temple of Bel illustrates those successive histories particularly well. Its principal sanctuary was dedicated in 32 CE, while building work and changes to the surrounding precinct continued afterwards. Christian use altered the complex, and parts were adapted again during the Islamic period. By the twentieth century, what visitors saw reflected ancient construction, subsequent reuse and modern restoration. Calling a digital model ‘ancient Palmyra’ without identifying its date can obscure those layers. Its makers need to explain which building phase is shown and where the reconstruction remains uncertain.
The destruction of 2015 and 2017 can be dated from independent evidence
The Syrian war placed both Palmyra’s inhabitants and its monuments in danger. UNESCO put the site on its List of World Heritage in Danger in 2013, and Islamic State forces took control of the area in May 2015. During the following months they destroyed sculptures and major buildings, including the temples of Baalshamin and Bel. They also killed Khaled al-Asaad, an archaeologist whose working life had been closely tied to the site. The killings, displacement and violence against local communities were part of the same conflict that destroyed these monuments.
The Temple of Baalshamin was demolished in August 2015. On 30 August, explosives brought down the main building of the Temple of Bel and damaged nearby columns. Investigators unable to reach Palmyra safely could nevertheless compare satellite images taken before and after the attack, allowing United Nations analysts to confirm the loss. Militants destroyed the Monumental Arch, commonly called the Arch of Triumph, in October 2015. A UNESCO mission in April 2016 could still identify surviving portions of the colonnaded street, the theatre and the Tetrapylon, even though the sanctuaries and arch had already suffered severe damage.
The destruction did not end in 2015. Islamic State forces returned to Palmyra late in 2016, and the Tetrapylon, an intersection marked by four sets of columns, and the theatre’s stage backdrop were badly damaged. The UNOSAT assessment published on 20 January 2017 compared images dated 26 December 2016 and 10 January 2017. It found rubble across the theatre stage and almost all of the Tetrapylon destroyed, with just two columns still standing. Those dates are essential when assessing a reconstruction: a photograph labelled ‘after 2015’ may show monuments that remained intact in 2016 and were lost by January 2017.
Satellites could establish roughly when major structures disappeared and reveal broad changes across the site, including new pits in areas vulnerable to illicit digging. They could not show the detailed carving on a fallen capital or determine whether a cracked block could carry weight again. Those questions demand closer evidence, from photographs and measured ground surveys to direct inspection of stone. Each method supplies a different part of the record.
How photographs become a three-dimensional model
Photogrammetry turns overlapping photographs into measurements of three-dimensional form. A camera captures a flat image, but a distinctive crack or carved corner visible from several viewpoints allows software to estimate where each camera stood and where that feature lies in space. It first identifies matching points, sometimes producing a sparse point cloud of coordinates. Denser matching produces many more points, which can be joined into a surface mesh. The colours from the source photographs are then applied as a texture, making the digital surface resemble the stone it represents.
The camera-positioning stage is often called structure from motion, followed by dense matching across multiple views. For reliable results, the pictures need substantial overlap, different viewing angles and sufficient detail for software to identify the same features repeatedly. Archaeologists can plan a photographic survey to provide those conditions. The images available for a destroyed building, by contrast, may have been taken decades apart with unfamiliar cameras, lenses and little deliberate overlap.
Visitors rarely photograph a monument as a surveyor would. A famous façade may appear hundreds of times from nearly the same vantage point, while the back of a cornice, a roof surface or a joint behind a column is never shown. Hundreds of images can therefore leave large parts of a building undocumented. Software may fill empty areas with smooth polygons so a model looks complete. If those guessed surfaces are not marked, a viewer can easily take them for recorded masonry.
Wissam Wahbeh, Stephan Nebiker and Gabriele Fangi confronted this problem in a 2016 study of the Temple of Bel. They used 180 selected tourist photographs together with professional panoramic photographs taken in 2010. The panoramas, assembled from images captured at controlled camera positions, offered a more consistent geometric framework than the visitor pictures alone. Combining the sources improved their Temple of Bel reconstruction, but gaps remained: the north and south portions of the interior had to be modelled separately because the photographs did not link them well enough.
When the team joined those two interior sections using common panorama stations, it reported an average alignment error of approximately 12 centimetres. That number applies to this particular operation, not to every Palmyra reconstruction. The 2016 paper also explained that the model initially lacked exact scale until known dimensions were used to approximate it. A visually convincing digital temple may therefore remain less reliable than a measured architectural plan.
Laser scans, ground control and the difference between detail and accuracy
A terrestrial laser scanner produces another kind of record. It sends light pulses toward the surfaces it can see and measures their returns, creating a cloud of spatial coordinates without depending on features matched between ordinary photographs. Scanning still has blind spots, measurement noise and the problem of aligning observations made from separate stations. Nevertheless, a scan made before a monument’s destruction can provide independent geometry against which a later photographic model can be checked.
Surveyors also use ground control points, accurately measured targets or stable features that locate a photographic model within a known coordinate system and establish its scale. They may obtain those measurements with total stations or satellite positioning equipment. Once a monument has vanished, researchers cannot place new targets on its lost walls. Instead, they must work from older plans, surviving architecture, panoramic camera stations or pre-destruction laser surveys. Which reference they choose affects the precision of the final reconstruction.
In a 2026 study by Scott McAvoy, Wissam Wahbeh and colleagues, the researchers compared three collections made before the Temple of Bel was destroyed. These comprised laser scanning by Kiyohide Saito in 2010, Fangi’s panoramic photography from 2010, and approximately 1,580 visitor photographs collected through #NewPalmyra and dated between 1974 and 2014. They registered the datasets in successive stages, using the strongest geometric reference first, instead of presenting one combined surface without showing how it was made.
The comparison showed that added visual detail came at a measurable cost to positional accuracy. Adding the tourist images made details of the temple’s exterior look better, but agreement with the laser reference worsened: deviations increased from a baseline of around six millimetres to roughly three or four centimetres after the data were combined. Sharper-looking masonry was not necessarily more accurately positioned masonry. The researchers also cautioned that a low reprojection error, which measures how well photograph features match a software-generated camera solution, does not guarantee agreement with an independent laser survey. A model can fit the photographs well while placing stone surfaces incorrectly.
Conservators and public historians may therefore need different versions of the same reconstruction. For measuring a surviving fragment, the laser-based or carefully controlled panoramic stage is the more appropriate reference. For presenting a lost building to visitors, the photo-rich version may communicate its appearance more effectively. Both have value when their limits are stated. McAvoy and his colleagues released the component datasets and a comparative viewer, allowing others to inspect individual reconstruction stages instead of relying on the most polished image alone.
The Temple of Bel: reconstructing a sanctuary from uneven records
Bel’s sanctuary presents unusual difficulties because of its size, decoration and history of alteration. The central chamber, or cella, was approximately 39.45 by 13.86 metres according to the 2016 study, with decorated spaces at opposing ends inside a far larger precinct. Its walls, doorways, roof and surrounding columns must be related accurately in three dimensions. A photograph might capture an elaborate ceiling carving in exceptional detail yet give almost no clue to its distance from a wall outside the photographer’s view. Different kinds of evidence have to bridge those missing measurements.
The panoramic stations used in the 2016 research helped provide some of those connections. For a 2023 study, Nemeh Rihani used publicly accessible visitor photographs to create a virtual experience of the destroyed temple. The surviving views proved too uneven to model every part of the building satisfactorily. Many concentrated on the cella, leaving other architectural areas poorly documented. The project could still give viewers an impression of the lost sanctuary, but it also exposed where photographs could not support a full reconstruction.
There is a historical reason to retain more than one digital version. Bel was used and altered across centuries, so a proposed first-century appearance, photographs from 2010 and a survey of rubble in 2016 describe different states of the same site. Researchers need to distinguish original pieces from later insertions and, where possible, mark repairs undertaken in the twentieth century. A pristine-looking hypothetical temple may help an audience imagine an earlier phase. It should never be confused with a record of what conservators actually found.
The UNESCO emergency safeguarding programme for the surviving portico shows what a model can contribute to physical safeguarding. The work envisaged gathering historical documentation, scanning surviving remains in three dimensions, assessing the masonry’s condition and deciding how to stabilize vulnerable sections. Measurements can help locate cracks and plan interventions, but an engineer or conservator must still examine the stone, assess loads and decide what can be done without causing further damage. The model supports that judgement; it does not replace it.
Baalshamin: rebuilding a temple's history as well as its shape
The Temple of Baalshamin was smaller than Bel, and it benefits from a particularly valuable archaeological archive. In the 1950s, the Swiss archaeologist Paul Collart directed excavations there and left photographs, field notes and measured drawings. Those records became far more consequential after the temple’s destruction in 2015. Founded in 2017, the University of Lausanne’s Collart-Palmyre Project digitized the excavation material and worked with the heritage-technology company Iconem on a three-dimensional model.
The Lausanne team sought to reconstruct the sanctuary’s changing history, not merely its appearance before the war. Their diachronic model traces successive phases from the late Hellenistic period through later reuse, assigning architectural features and objects to their appropriate periods. In a 2023 research paper, Patrick Michel explained how an online viewer could connect places in the model to archival documents. A researcher examining a doorway or wall can consult the evidence behind its representation rather than accepting a finished digital façade without explanation.
That method matters because the same block may have been moved or reused during the sanctuary’s long life. A later wall might contain older masonry, while an excavation photograph might preserve details altered by twentieth-century consolidation. By retaining the evidence behind each modelling decision, archaeologists can revise an interpretation without erasing its documentary foundation. The ALIPH Foundation’s project record lists archive records, scans made in 2016 and public digital resources intended to assist future conservation. It does not claim that Baalshamin has already been reconstructed in stone.
The project has also involved Syrian communities and displaced people in heritage activities. Tadmor was a lived-in city before its monuments were photographed and circulated around the world. Its residents have a stake in what happens to those records. Michel discusses the danger that digital heritage could be detached from its local setting if access and control rest entirely with institutions abroad. Languages, rights of use, maintenance and long-term custody matter alongside the quality of the model itself.
The arch, the Tetrapylon and the problem of matching broken stones
Palmyra’s Monumental Arch joined stretches of the colonnaded street at an angle, giving the structure an unusually complicated plan. Pre-2015 images document its carved surfaces, profile and arrangement of blocks. After its demolition, however, identifying pieces amid the rubble became a separate problem. Digital comparison of broken edges, dimensions and carved details can help suggest where fragments belonged. Whether the damaged material can be lifted, joined and made structurally secure requires direct examination and engineering analysis.
The Tetrapylon poses another question: which version should be reconstructed? The structure visible before the recent attacks contained ancient work as well as elements replaced during twentieth-century restoration. A digital model of its appearance in 2010 accurately records that modern condition, but cannot automatically be presented as the monument in Roman times. The date and provenance of each column or other component should remain traceable, so viewers can distinguish surviving antiquity, documented restoration and hypothetical additions.
Detailed photographs of fallen stone can preserve tool marks, clamp holes, dimensions and the shapes of fractures. Matching a fragment to an older excavation image may help establish where it once stood. A photogrammetric surface alone, however, cannot reveal an internal crack or the strength of the material. Structural simulations need material testing, an understanding of damage and a properly approved engineering design. Digital geometry is one input to those decisions, not their answer.
In 2024, UNESCO’s World Heritage Committee commended the documentation and heritage-impact work undertaken for the proposed restoration of the Triumphal Arch. The committee decision regarded the methodology as a useful example for other interventions while still requiring appropriate review before construction. Approval of careful research and planning should not be mistaken for evidence that the arch had already been physically rebuilt.
What drones and satellites can show after an attack
Digital surveying can also document the site after an attack, once conditions permit safe access. Drones can take overlapping pictures of ruins and surrounding ground, while photographs from the ground capture surfaces concealed from above. Together they can produce measured orthophotographs, aerial images corrected for map-like measurement, and elevation models showing the shape of the terrain. Later surveys can then be compared with the earlier record to detect movement, fresh damage or disturbed earth.
Palmyra has been the subject of such aerial work. The Russian Academy of Sciences’ Palmyra 3D project describes campaigns in 2016 and 2019 combining drone photography, archaeological mapping and a geographic information system. Its published account includes orthophotographs, terrain models and records of damage to tombs and monuments. Those records may assist later surveys, although the project description alone cannot establish their precise accuracy or determine who should control access to the data.
Satellite imagery works on a broader scale. The 2017 UNOSAT comparison, for example, narrowed the period in which the Tetrapylon was destroyed without requiring investigators to enter Palmyra. Repeated images can reveal large disturbances around burial grounds and changes that may indicate illicit excavation. Resolution, changing shadows, camera angle and ambiguous ground markings impose limits. A dark patch or new pit may call for inspection, but does not by itself establish looting or identify precisely how a monument was damaged.
The three sources address different gaps in the evidence when their results are compared carefully. Satellite scenes supply dated overviews, drones show the relation between structures and landscape, and close-range surveys preserve individual carved surfaces. Those datasets can share mapped coordinates while still differing in date, scale and measurement error. Combining them without documenting those differences would suggest, falsely, that the whole site had been recorded at the same moment and to the same precision.
The Lion of al-Lat shows what physical restoration requires
The Lion of al-Lat demonstrates what restoration of an actual object involves. Discovered by Polish archaeologists in 1977, the large limestone sculpture stood near Palmyra Museum. It depicts a lion holding a gazelle protectively between its paws and is associated with the goddess al-Lat. The sculpture was damaged in 2015, after which recovered fragments were taken to Damascus. Restorer Bartosz Markowski worked on rebuilding the object, and the restored lion was unveiled in 2017.
According to the UNESCO account of the restoration, the sculpture stands 3.45 metres high and weighs approximately 15 tonnes. Earlier photographs helped document its original appearance, but recovering that appearance in stone required conservators to examine surviving fragments, work out how to reassemble them, provide support and distinguish new intervention from surviving material. Markowski subsequently described the project in a technical paper published in Polish Archaeology in the Mediterranean. The work depended on conservation knowledge and physical handling of a massive damaged object as well as visual records.
It would therefore be misleading to say that ‘3D technology rebuilt’ Palmyra’s ruins. A digital survey can guide a repair, document changes and help specialists compare fragments. The restored Lion of al-Lat is a material achievement involving decisions about archaeology, stability and conservation, followed by skilled work on the sculpture itself. A digital model cannot repair the fractured limestone.
The London arch was a replica, not a restored monument
In April 2016, visitors to Trafalgar Square in London saw a replica of Palmyra’s Monumental Arch. The Institute for Digital Archaeology produced the object using computer-guided carving and later displayed it in other cities. The institute’s account describes the replica as one-third scale and says that it was carved from Egyptian marble. No stones from Palmyra were reassembled in London. The arch was an exhibition replica designed to draw attention to threatened heritage.
The installation provoked a debate about the meaning of reconstruction. Supporters argued that it showed how documentation could preserve knowledge in the face of deliberate destruction. Critics questioned the material, scale, access to the data and political implications of reproducing the arch abroad. In a 2018 study of the replica’s ethics, Roshni Khunti argued that the project had not adequately reckoned with the losses suffered by Palmyra’s people and raised questions about who controlled the digital representation. Her criticism concerned the ethics of this particular display, rather than the basic usefulness of photogrammetry.
The difference is also practical. Engineers can design a newly carved, reduced-scale arch to carry its weight using modern materials. Re-erecting the original monument in Syria would involve damaged ancient stone, foundations with their own history, previous repairs, seismic risk and public safety. A convincing replica in a public square is no evidence that the historical arch could safely be reconstructed by the same process.
Who owns a digital Palmyra?
Digital records can keep fragile evidence available after the original becomes inaccessible, but they can also create new inequalities. A foreign institution might retain the raw images, camera calibration and detailed three-dimensional files, while Syrian researchers or residents are offered only a video or a viewer with limited functions. That difference matters when the information could be used to plan restoration or shape how a community’s past is represented. Being able to look at a model is not the same as being able to study, correct and reuse its underlying evidence.
The Collart-Palmyre Project shows the value of retaining archaeological archives alongside the model and linking individual architectural features to the records that support them. Syrian participants have been involved in that work. A separate 2025 UNESCO interactive map of archaeological research at Palmyra, developed with the French Ministry of Culture and the University of Lausanne, locates archaeological projects and interventions across the site, making it harder to treat conservation as the achievement of one outside institution. Yet a map alone cannot settle questions about who may download, maintain and reuse the original datasets.
Those questions concern people displaced from Tadmor as well as researchers. A three-dimensional record of the temple precinct may help conserve archaeological masonry, while surveys of inhabited neighbourhoods could be useful for housing, infrastructure and plans for return. The needs are different, and residents’ participation should not be reduced to supplying photographs for an international archive. Palmyra’s heritage includes the present city and oasis as well as the classical monuments most familiar to visitors abroad.
Generative artificial intelligence raises an additional problem: it can produce images that appear historically convincing without accurately representing the evidence. In a 2026 study, Kawsar Arzomand and Tatiana Kalganova tested an evidence-guided approach to generating features of the Temple of Bel. Their evaluators rated some architecture more favourably than other elements, but 56.2 percent considered none of the generated relief designs historically accurate. Such experiments can reveal how viewers respond to reconstructions. They cannot replace measured inscriptions or the surviving carved evidence, and the researchers did not present them as substitutes.
What Palmyra's digital records can support in 2026
In 2026 Palmyra still faced substantial conservation work. UNESCO’s World Heritage Committee, in its 2026 decision on the site, welcomed a recovery and reuse master plan and cooperation among Syrian authorities, international bodies and local communities. It also warned about continuing damage to the burial grounds and extensive illicit digging. The committee requested a conservation plan for documentation, assessment, restoration and possible reconstruction, and kept Palmyra on the List of World Heritage in Danger. Plans for the museum’s rehabilitation were still being developed. None of these decisions meant the site’s restoration was complete.
Any physical repair requires evidence beyond the visible surfaces of a digital rendering. Accurate records can help identify adjoining stones, compare damage over time, design an intervention and make its reasoning open to review. Virtual reconstructions can also give students and people unable to visit Palmyra a useful understanding of its buildings. But a model cannot measure the strength of an unseen fracture, authenticate missing sculpture or decide what should be rebuilt without the authorities and communities responsible for the place.
Photographs taken at Palmyra before 2015 now preserve details that their photographers may never have expected to become irreplaceable. Some record decorated surfaces destroyed in the attacks; others show a particular stage in the site’s long history of reuse and repair. Their usefulness depends on retaining when, where and how they were made, and how confidently a reconstructed surface follows from them. Digital Palmyra is most valuable as a collection of documented, revisable evidence that future archaeologists and conservators can test and use when responsible work on the ground is possible.
Discussion
Comments