To date, our project has created detailed records for over 20,000 sites from a total of c.150,000 identified sites with partial records. Of the detailed records, over c.20% are previously known sites, documented from published surveys or excavations. A further c.65% are sites identified from satellite imagery and classed as having a medium or high certainty of being an archaeological ‘site’ or “feature”. The remaining c.15% are those with a low or negligible certainty of being archaeological. The database also contains over 50,000 records providing details about the sources (e.g., satellite imagery, aerial photographs or bibliographic sources) consulted by the project. This work is constantly developing, and the team is currently in the process of evaluating many thousands of potential new archaeological sites.
3.1. Case Study 1: Cyrene: The Impacts of Modern Development on a World Heritage Site and Its Immediate Hinterland
The site of Cyrene in Eastern Libya is a UNESCO (United Nations Educational, Scientific and Cultural Organization) World Heritage site (designation number 190). It faces significant problems arising from present-day activities including expansion of the adjacent town of Shahat and limited enforcement of planning regulations. While the gradual degradation of the archaeology has been an issue for many years, damage to archaeological features has accelerated because of the civil war of 2011 and the subsequent instability.
Cyrene’s monuments were first recorded by travellers in the 18th and 19th centuries [58
], with more extensive archaeological investigations over the course of the 20th and 21st centuries e.g., [59
]. Cyrene developed from a Greek colony in the 7th–4th centuries BC, with occupation continuing through the Roman and Byzantine periods [66
]. Located beside the modern town of Shahat, Cyrene has a walled circuit although much of the ancient city is outside this area, including large suburban cemeteries (Table 4
) and sanctuaries (Figure 7
). The site lies on the edge of an escarpment, 8 km from the coast, and is surrounded by arable fields and modern farms. The urban core is fenced and protected, but the suburban zones are vulnerable to a variety of threats [67
Despite current difficulties of access for foreign archaeologists, approaches that combine remote sensing and GIS survey undertaken by Libyan archaeologists have highlighted the severity of the threat (for example, [69
]). Several recent projects concerned with heritage protection have examined the risks faced by Cyrene and worked to document it; their published reports have been cited in our database where applicable and instances of damage they describe logged. The Cyrenaica Archaeological Project has undertaken a holistic approach including recording and training [70
]. They have noted specific instances of damage, for example caused by weathering and vegetation, also recorded in the EAMENA database. They worked collaboratively with the Department of Antiquities in Shahat to develop a sites and museums database. The Curious Travellers Project [24
] is also gathering data to make 3D models of Cyrene using photogrammetric methods.
3.1.1. EAMENA’s Methodology for Recording Cyrene
EAMENA has created a detailed set of records describing the site of Cyrene, the nature of the damage and the risks that are affecting it. The site consists of one “parent” record and over 30 sub-records which represent individual structures/features which are part of the overall complex and surrounding features. To build each record, we have used multiple sets of data including aerial photographs and satellite imagery and published and unpublished reports from archaeologists. Details were recorded from these data including the form, morphology, location, interpretation and condition of these features; for example, whether they were of good or poor condition according to the latest information and to what extent they had been impacted on by modern landuse.
A historical analysis of images (Table 5
) of Cyrene and its immediate hinterland highlights the impacts that development related expansion has had on archaeology, and the value of examining this over a long period of time (1949–2016). We can map the site and its immediate hinterland in detail using aerial photographs collected by Hunting Surveys dating to 1949 [65
]. A KH7 satellite image from 1967 allows further mapping. 39 images on Google Earth dating from 2006–2017 allowed detailed identification of the archaeological features and modern changes. A GeoEye-1 image from 2016 has allowed mapping and spectral analysis. Changes in the size of Shahat over time were made using Landsat images (1986 and 2000) and a Sentinel-2 image (Figure 8
). This broad dataset highlights EAMENA’s use of a range of sources to populate our database. Some features were detectable in data such as the satellite imagery, but others could only be recorded using the published data. Using both these types of information in conjunction allowed details of instances of damage to be established.
While many features are visible in the high-resolution satellite images and aerial photographs, there are features which cannot be easily recorded in this way. These include tombs of several different types and morphologies, including rock-cut structures and sarcophagi. Some are located on the slopes of the escarpment and side of wadis, making them particularly invisible to remote sensing methods. There are also specific instances of damage that cannot be identified remotely. This highlights the necessity for EAMENA to use a variety of datasets, where possible backed up by field work. In this case, several sources of published information deriving from surveys, excavations, guides and archival research have been consulted [65
] and we have worked closely with a Libyan PhD student at the University of Leicester, Mohamed Omar, who is studying Cyrene’s suburbs.
3.1.2. Antiquity to Mid-20th Century AD
Evidence for events which damaged Cyrene prior to the 20th century comes from excavations and historical texts rather than from satellite imagery. Damage to structures in the Mediterranean region were caused in antiquity by earthquakes in the mid-third century AD and in AD 365 [71
]. These are mentioned in historical sources and confirmed by archaeological and geological evidence [62
]. Archaeological excavations, ongoing since the 19th/early 20th century, have disturbed components of the site and have been logged in the EAMENA database as events which may have affected the site’s preservation. Restoration and landscaping efforts undertaken during this era have also had a deleterious impact on Cyrene, including tree planting to the north and east of the acropolis during the Italian colonial period [66
] (p. 148). The potential effects of vegetation on archaeological features have been noted in our database.
3.1.3. Mid-20th–21st Century AD
In addition to other sources, aerial and satellite images can be used to record changes from the first half of the 20th century. The 1949 aerial photographs show that the ancient city and its suburbs were relatively undisturbed by construction and development work at that time, and that Shahat was a small village. It had originally been located on the northern part of the ancient town, but on the advice of the archaeologist Goodchild, its focus was shifted to the south-east, outside the walls [66
]. The KH7 image (1967) shows that Shahat had started to expand in its new location by the late 1960s, but the cemeteries and other suburban features still appear to have been relatively unaffected by construction-related work. Since then, however, this area has been particularly at risk, and tomb robbing and vandalism has been recorded by archaeologists from the 1960s onwards [65
]. By the 1980s the expansion of Shahat had destroyed most of the Southern Necropolis [66
] (p. 151), [68
]. That this process is continuing is clearly documented on satellite imagery and confirmed by Libyan archaeologists [69
]. The recent developments include construction of houses, farms and infrastructure, with evident impacts on structures outside the ancient city walls, especially the cemeteries and the sanctuary of Demeter. Some ancient structures have been bulldozed or otherwise damaged to make way for new constructions, whilst others have been exploited for building materials [69
The expansion of the present-day town is very apparent on imagery dating from 2006 onwards. Modern roads and farms have encroached on the area of the southern and eastern cemeteries in particular. Structural robbing for building materials is part of this unregulated expansion and is recorded in our database as a source of damage and future risk. Published and unpublished records and reports have also highlighted other issues which affected the site during this period, such as pollution by sewage and rubbish dumping, which are less visible on satellite imagery [69
3.1.4. Recent Changes
Activities affecting the preservation of Cyrene and its immediate hinterland have accelerated even further over the past five years following the recent conflict, which has seen much illegal and unregulated construction work in Libya [38
]. The satellite images indicate a further massive increase in the extent of Shahat, for example, demonstrated by a Landsat 5 image (1986) and a Sentinel 2 image (2017) (Figure 8
Unsupervised classifications of Landsat and Sentinel-2 images were calculated using ERDAS (Figure 9
). Other than a detection of part of the archaeological area on the acropolis, they have picked out pixels representing modern urban activity. These show how the urban area has grown between 1986–2017. The impact of this development was recorded for the sites in our database affected by it. The core area of the town has expanded slightly, especially towards the south-west; however, more dispersed structures and associated infrastructure has spread in all directions, directly threatening the archaeological features in these areas. In the southern necropolis, in particular, there are new farms, buildings and roads. Al-Raeid et al. [69
] (pp. 8–9) reported robbing of ancient structures for building materials in this area and the looting and vandalism of tombs. They also noted the effects of continued lack of conservation on these structures. The pattern is similar in the other suburban areas. The impact of processes less identifiable from satellite imagery, including damage caused by vandalism, and water pollution, have also been noted by other sources and are listed by the World Heritage Committee in its most recent documentation [74
3.1.5. Damage Statistics
The systematic recording by EAMENA of the causes of damage and potential threats allows these problems to be measured. Table 6
presents the results from 38 site records, which were created from interpretation of aerial and satellite images, information from published reports and guides, and from discussions with our Libyan colleagues. It is worth noting that although counted only once here, some of these sites represent large areas containing multiple archaeological features. Table 7
records the proportion of sites recorded as being destroyed, damaged, or of unknown condition.
Archaeological excavations since the 19th–20th centuries have affected at least 24 sites. However, one of the most significant causes of damage to archaeology at Cyrene is modern development which comprises construction of buildings and related infrastructure/transport and utilities (24 sites affected by these categories so far—63% of the records). While the area inside the walls, including the acropolis, the main urban area and the sanctuary of Apollo, is protected from this type of damage, the features outside this zone including the cemeteries are being encroached upon by modern constructions including roads, tracks, farms and houses (Figure 10
). This problem has been mapped and recorded across the wider area of Cyrene and Shahat using satellite images showing expansion since the 1960s (e.g., see Figure 8
and Figure 9
Although difficult to identify using imagery alone, structural robbing of tombs, as well as deliberate vandalism has affected many sites. Several have been recorded as having been looted (at least seven); a figure that may rise when individual tombs can be logged.
Agricultural activity has also affected the areas surrounding the site. The hippodrome (EAMENA-0116827) has been damaged by long-term agricultural activity including planting and ploughing. Since the recent conflict, the inability to enforce regulations has led to clearing of remains to make way for new fields [67
] (p. 156). The category ‘natural’ is also a significant cause of damage (13 sites so far). This comprises recent issues such as tree growth but also known instances of damage caused by earthquakes in antiquity.
The condition of the sites was recorded using EAMENA terminologies and was assessed using the analysis of the satellite imagery and classifications and the reports of recent visitors. 23 (65%) of the sites could be described as “Good” to “Fair” (Table 7
), especially sites nominally protected by their location on the acropolis ridge. This means that they can be regarded as being reasonably stable. However sites surrounding the acropolis were less well preserved with signs of severe structural instability/missing and deteriorating features and were suffering from the consequences of ongoing activity such as structural robbing. In some cases it was not possible to identify the current condition of sites other than noting that they were likely to have been impacted by disturbances.
As described above, EAMENA also records potential threats and risks (Table 8
) which could affect archaeological sites in the future. These are recorded based on problems currently affecting sites and analysis of continuing issues in the vicinity. The urban growth identified using the multispectral satellite images (Figure 8
and Figure 9
) is an urgent issue. For example, the westward expansion of Shahat is likely to cause further damage to archaeological features in that area, including the tombs of the southern and western cemeteries. Larger features in that zone, including the Sanctuary of Demeter (EAMENA-0117108), are at high risk, for example from structural robbing or even demolition.
Cyrene achieved World Heritage Site status in 1982. The World Heritage Committee has recognised the ongoing threats to the site and have proposed satellite monitoring, field recording, additional security measures and identification of the boundaries of the designated site [74
]. However, World Heritage status has not provided tangible protection to Cyrene. Since 2011, often at considerable personal risk, Libyan archaeologists and local people have worked to protect archaeological sites and museums at Cyrene, but so far it has not been possible to enact a solution to the problems [67
] (pp. 155–156). Overall, our analysis of multiple datasets shows that while development in the vicinity of Cyrene has been taking place since the 1960s at least, it is now occurring at an especially rapid rate, one that directly threatens surviving features in the hinterland of the site including rock-cut tombs. Archaeological sites close to urban areas should therefore be monitored and recorded as a priority and regular classifications of multispectral imagery performed to track Shahat’s growth.
3.2. Case Study 2: Homs Cairns: The Benefits and Challenges of Monitoring Stone Monuments via Remote Sensing
From 2007–2010 a fieldwork project undertaken by one of the current authors mapped and analysed 525 potential burial cairns to the north-west of the modern city of Homs (Syria). This project was undertaken within the framework of the Syrian-British landscape project Settlement and Landscape Development in the Homs Region, and the field data was recorded within its GIS framework. Published overviews of the archaeology of the Homs basalt region in Graeco-Roman [75
] and earlier periods [76
] contextualise the various monuments in relation to settlement activity and the wider landscape; readers should consult these for further information. Cairns are visible on the ground as piles of stone (Figure 11
), and vary considerably in terms of size, structure and form. A well-documented form of monument found throughout the Levant and North Africa, they are also visible via satellite imagery, and can be distinguished as small circular or oval features, in many cases associated with enclosures and other archaeological traces. Additional research as part of a PhD thesis [77
] identified a further 169,000 potential cairns from an area of c.
) using remotely sensed data spanning the late 1960s to the early 2000s (Corona KH4-B, KH7, historic aerial photographs, Ikonos (panchromatic and multi-spectral). The majority of these (over 90%) were found in association with the local basalt flows to the north and south-west of the modern city of Homs [77
], whilst a much smaller percentage was found in association with lacustrine marls, limestones, clays and sands.
Details recorded during the fieldwork included the form, morphology, location and interpretation of these features. These were all collated in a project database, alongside basic information about levels of preservation. For example, a rough measure of “percentage intactness” was recorded for each cairn surveyed in the field (less than 50% intact; more than 50% intact; 100% intact), and notes were made about the potential causes and effects of any identifiable disturbances. A preliminary assessment of recent, pre-conflict, land-use practices (Figure 13
), carried out in 2010 using Ikonos panchromatic imagery (from 2002), indicated that over 60% of the archaeological features, including cairns, enclosures and other features, identified from the Corona satellite imagery have been either partly or totally destroyed by clearance or ‘de-rocking’ operations using heavy machinery, often bulldozers, noted in the field by the authors during fieldwork, with the intention of increasing the cultivable area [79
]. The irony is that a practice that was originally supported by development organisations to increase agricultural productivity has been widely adopted at a local level, often on a ‘freelance’ basis and with little technical or administrative oversight, and now poses a serious risk to the preservation of cultural heritage. Assessments carried out using this imagery, however, also indicated that areas of the study region were still being used for grazing activities and had, as yet, not been cleared or bulldozed.
In 2016 these field records (525 in total) were loaded into the EAMENA database, and updated using the EAMENA methodology. Preliminary disturbance and threat assessments were also recorded for a sample (6975) of the 169,000 potential cairns, identified from satellite imagery, bringing the total recorded from this area in the EAMENA database up to 7000.
In the case of the surveyed cairns, disturbance assessments were generated from the field survey records, which recorded landuse and landcover at the time of data collection. This information was then double checked against the most up to date imagery in Google Earth (2014–2016). For those cairns not visited in the field, a remote characterisation assessment was made, by assessing groups of cairns in relation to their association with different types of landuse.
As the original 2010 study had indicated, this work demonstrated that whilst 59% of potential cairns showed “No Visible/Known” disturbance causes (Figure 14
), nearly 40% were affected by bulldozing or clearance activities (Figure 15
). Clearance destroys even substantial surface and sub-surface archaeological features, and creates a “cleared” field, bordered by newly constructed field walls composed of huge basalt boulders, which can easily be identified from satellite imagery. In total 2683 (38%) of digitised cairns were recorded as “Destroyed”, while 4159 (59%) were recorded as being in ‘Good’ condition.
This preliminary and basic assessment has a number of limitations. For example, whilst some cairns identified during ground survey were recorded as being in either a “Fair” or “Poor” condition, the resolution of the imagery means that, more often only two basic condition states can be identified in remote sensing analysis: “Destroyed” or “Good”. Using this “broad brush” approach also limits the range of disturbance causes and effects that can be identified. In particular, clearance activities appear as a major disturbance factor. Moreover, the size of the features (generally between 2 m–20 m in diameter), means that the different types of disturbance causes which can be identified from satellite imagery alone are limited. As a result, the number of features affected by other disturbance causes, such as looting, construction and dumping is probably a significant under-estimate.
For example, out of the 104 cairns which were recorded in field survey as having identified disturbance causes, over 70% were affected by recent construction activities, such as the erection of small hides or shelters or modern dumping of cleared material or rubbish. A much lower percentage (c.
17%) were recorded as having been disturbed by illicit excavations, whether recently or in antiquity, although dumping might have concealed earlier looting activity. Based on the field notes, just under 8% of recorded cairns were associated with a disturbance cause of clearance/bulldozing activities. This low figure is due to a number of factors; firstly, as one of the field survey’s main aims was to understand the morphology, chronology and location of these features, the research specifically targeted cairns in areas where they were better preserved, based on 2002 imagery. Thus, while c.
8% of cairns were categorised as damaged by clearance or bulldozing activities in the intervening five-eight years, based on satellite imagery analysis, this number probably significantly underestimates the overall impact of this disturbance cause at the regional level. It is also apparent that a number of disturbance causes and effects identified by the field survey cannot be identified from satellite imagery alone. By way of example, EAMENA-0059581 was recorded in 2006 as a fully intact, large cairn. The survey returned to the same location in 2007 to find that the feature in question had been illicitly excavated and was most likely not a cairn, but instead a mausoleum dating to the Roman period. The excavation exposed the internal structure of the monument and material, mostly consisting of pottery sherds, was strewn across the area, but no evidence of this disturbance is visible via Google Earth (Figure 16
Despite these limitations, recent remote sensing analysis allows us to identify broad-scale changes and while the ongoing conflict has rendered these features inaccessible on the ground, we have taken our analysis further. Using imagery from Google Earth, we have been able to revise overall condition assessments for the original surveyed cairns. Out of a total of 525 field-recorded cairns, 127 (24%), required an updated re-assessment, whilst the remaining 398 (76%) showed no significant changes over the seven years since the original study was completed. Unfortunately, the majority of cases involved updates to the disturbance extent and overall condition state. Most required a re-classification of the overall state from “Good”, “Fair” or “Poor” to “Destroyed”.
Overall, based on these re-analyses, the percentage of cairns listed as ‘91–100% disturbed’ increased to 66% of the sample, with the number of cairns with an ‘Unknown’ or ‘1–10% disturbance’ extent also increasing (Table 9
). This reveals fairly significant changes, with the total number of cairns recorded as showing “91–100% disturbance” increasing from 0 to 85 (16%). Due to the poor resolution of some of the latest available imagery in Google Earth, the number of cairns for which assessment was not possible (e.g., disturbance extent or condition recorded as ‘Unknown’) also increased.
The pattern for the overall condition state is very similar (Table 10
), with the total number of cairns identified as being in ‘Good’ condition decreasing from 368 (70%) to 272 (52%) of surveyed cairns. Conversely, the number of ‘Destroyed’ cairns has increased from 0 (0%) to 84 (16%) during this period.
This updated analysis also allowed us to confirm, and quantify, a number of disturbance causes which were not originally recorded in the field, although were noted as possibilities. These included evidence for flooding, a disturbance cause that was identifiable through the use of multi-temporal imagery which showed that a number of cairn clusters found in the vicinity of seasonal lakes were likely to have been affected by flooding.
Despite the limitations of using satellite imagery to record and monitor disturbances such as the illicit excavation of stone monuments, this case study illustrates the benefits of using EAMENA’s simple remote sensing techniques to continue to monitor monuments and update records in currently inaccessible areas. As the most recent imagery available for this area in Google Earth dates to 2015/2016, it is likely that the disturbance patterns identified here have continued since then.
3.3. Remote Sensing and Field Survey
Field-based validation for many archaeological features in the database may be possible in the long term: the EAMENA database is being made available to individuals and institutions with responsibilities for cultural heritage throughout the MENA region. Its uptake is being facilitated by dedicated training courses and collaborative working. As with any monuments record, database entries can be revisited and updated in the future as necessary.
Over a more immediate timescale, however, we need to ensure that our image interpretation methodology is producing viable data which will help, rather than hinder, the protection efforts of archaeologists in the MENA countries. Ultimately, each filled-in record needs to be a starting point for future detailed recording of site location, ideas and interpretation, and the identification of potential threats. As a cross-check on our methodology we are systematically comparing field-based and remote-based interpretation for select samples of sites. EAMENA is actively collaborating with several projects conducting field survey, for example the Middle Draa Project [80
] and Koubba Coastal Survey [81
]. Ground survey allows further details about many sites to be added to the database. However, most significantly, it allows us to assess the accuracy of EAMENA’s remote-sensing methodology of standardised interpretations and terminologies.
Validation of remote sensing methods by comparing results to interpretations made on the ground is a well-accepted process in the wider field, and there are established statistical and descriptive methods in remote sensing for assessing the accuracy of data such as image classifications [82
]. Accuracy assessments need clear plans, an unbiased and consistent sampling procedure, and a process of analysing the data. “Classes” assigned to the site from both ground collected data (fieldwork) and image interpretation can be compared, for example by using an error matrix [82
] (p. 3) [83
]. Although adopted by some projects [34
], the process of quantifying the accuracy of image interpretation and remote-sensing has not been widely used by archaeologists, and can be challenging when dealing with multiple levels of image interpretation. In many cases, some archaeological information simply cannot be known without field-based investigation.
While it is beyond the scope of this article to discuss this at length, we outline our field-based validation strategy here. The data comparison below was made by getting an analyst not familiar with the areas concerned, but trained in the EAMENA methods, to identify sites and damage threats in two sample areas for which we have ground data in Morocco and Lebanon. We compared the site records made separately using Google Earth images with interpretations made on the ground, using a simple table to reflect key terminology from our database. We counted the number of sites which matched, had a full or partial match, or did not match. The concordance between the numbers identified using each method is then established (Table 11
The results of this exercise are presented in Table 11
. This shows the number of exact matches between image and ground interpretations and, given the difficulties of making detailed interpretations from imagery alone, we also counted correlated matches, including instances where the image interpreter simply made a broader interpretation (e.g., “building”) than the field-based interpreter (e.g., “house”). The morphology, form and functional interpretation of a site were often easy to identify using imagery. While we were often correct in recognising that a site had been damaged in some way, it was much more challenging to identify the type of damage which had affected it.
A key factor to note is that the EAMENA site terminology extends beyond what may be visible on satellite imagery, as it incorporates categories that derive from ground survey, but are meaningful in an archaeological sense. By the same token, analysts are trained not only in the EAMENA methodologies, but also in the regional archaeological typologies and dating frameworks, which are generally derived from a long history of ground based investigation of sites. In our blind tests, we required an analyst with expertise in the field archaeology of Lebanon but unfamiliar with Morocco to look at that area and one with experience in Morocco to look at the Lebanese data. This probably accounts, in-part, for the lack of exact matches and emphasises the importance of local knowledge, and so highlights some of the significant challenges faced by crowd mapping and automated methods.
As we increase the number of samples used for this validation process, we will be able to refine our methodology based on these results and so identify error thresholds appropriate for application to assessments of archaeological remote recording. Remote classification of modern land-use, for example, can be relatively straightforward and its accuracy easily assessed. Given that many archaeological sites cannot be fully interpreted without ground-based work, especially excavation, EAMENA will seek to establish a more nuanced methodology that is attuned to assessing accuracy of archaeological interpretations. Whilst the details of this are beyond the scope of the current paper, the project is developing its field and imagery validation methods via further blind tests. We will explore the different factors affecting our ability to accurately identify and categorise site types, disturbances and threats, and determine whether our methodology needs to be adapted or refined as a result of this. As archaeological work is likely to rely increasingly upon remote sensing for making interpretations, robust assessment of accuracy is necessary, especially for the large-scale data collection undertaken by our project.