Between the Wars

In a year-long forensic investigation,  Caucasus Heritage Watch  (CHW) used high-resolution satellite imagery and other visual sources to document changes to the condition of Azerbaijani cultural heritage in territories controlled by an ethnic Armenian separatist government after the First Nagorno-Karabakh War (1991-1994). For nearly three decades, the unrecognized Republic of Nagorno-Karabakh (NKR) administered a territory that roughly included the Nagorno-Karabakh Autonomous Oblast of Soviet Azerbaijan, as well as the seven surrounding regions of Aghdam, Fuzuli, Jebrayil, Zangelan, Qubatly, Lachin and Kalbajar (Map 1). After decades of unsuccessful negotiations, regular skirmishes, and a 4-day war in 2016, Azerbaijan regained control of the seven territories and the historic city of Shusha (Arm. Shushi) in a bloody 44-day war in 2020. This investigation concerns the treatment of Azerbaijani cultural heritage in the unsettled decades between the First and Second Nagorno-Karabakh Wars.

To summarize our main finding, CHW’s research reveals that the treatment of mosques, mausolea, and historic cemeteries under Armenian control varied widely and changed over time, reflecting the complexities of the disputed region’s shifting political, geopolitical, and economic realities. Of the 109 cultural heritage sites we were able to assess for this investigation, 42 (38%) remained structurally unchanged since the late Soviet period, 9 (8%) sustained minor structural damage, 39 (36%) sustained major structural damage, 16 (15%) were destroyed, 2 (2%) were renovated, and 1 (1%) was restored. The adverse impacts on Azerbaijani cultural heritage were significant; the total of heavily damaged (n=31) and destroyed (n=16) sites represents a majority of our dataset (51%). Armenian authorities of NKR failed to prevent both widespread looting and several instances of targeted destruction. At the same time, the forensic evidence we present below shows no attempt to systematically erase the material traces of Azerbaijani history and cultural life in the troubled lands that Armenians controlled from 1994-2020. This invetigation is thus an effort to clearly define what has been lost and what remains.     

 Although most of the damage and destruction to cultural heritage documented here occurred many years ago, and Azerbaijan has since regained control over most of the territories under examination, a full account of the fate of Azerbaijani cultural heritage while under the administration of de facto Armenian authorities is important. A politicized discourse of heritage and its abuses pervades the Nagorno-Karabakh conflict. Exaggerated reports and unsupported claims of heritage destruction not only inflame the conflict, but they have so obscured the facts on the ground that it is almost impossible to come by an objective accounting of impacts to cultural heritage. Azerbaijani officials  have claimed  the near total destruction of mosques in the region. The  official factoid , frequently recited by President Ilham Aliyev, holds that 65 out of 67 mosques were destroyed. Our investigation shows that this is false. Dismissive Armenian rejoinders (most recently  voiced  by Prime Minister Nikol Pashinyan) have placed the blame for all abuses to religious monuments in Nagorno-Karabakh on bygone Soviet authorities. This is also incorrect. 

This investigation provides the first systematic, fully transparent assessment of the condition of Azerbaijani mosques, mausolea, and historic cemeteries that were still extant at the end of the Soviet period using satellite imagery and other visual evidence. We undertook this investigation in order to establish an empirically verifiable record of what did and what did not happen – one that can cut through the veritable “flood of obfuscating messages” ¹  permeating this conflict. This research provides Azerbaijanis and Armenians, as well as all outside observers, with painstakingly detailed evidence that can support deeper understanding and honest reckoning. This study also advances CHW’s larger goal to remove cultural heritage from the crosshairs of conflict; it is only with full and accurate information, informed publics, and dialogue that Karabakh’s diminished, endangered, and centuries-old Azerbaijani and Armenian cultural heritage can be spared new abuses and manipulation in the culture war between Armenia and Azerbaijan.

This study concerns the treatment of cultural heritage in the period between two wars, amidst decades shaped by sustained violence, profound loss, forced displacement, and abandonment. The First Nagorno-Karabakh war was an ethnic territorial conflict that arose out of the collapse of the U.S.S.R. In February 1988, the parliament of the Nagorno-Karabakh Autonomous Oblast (NKAO) voted to unite the enclave with Armenia, and in a referendum that month, the majority Armenian population of NKAO voted in favor of independence from Azerbaijan. Pogroms of Armenians in Sumgait (February 1988), Kirovabad (Ganja) (November 1988), and Baku (January 1990), and of Azerbaijanis in Gugark (March 1988) sparked a cascade of violence, leading to full-scale war in which the Armenians of Nagorno-Karabakh were supported by the newly independent Republic of Armenia. Waves of ethnic cleansing ensued. Although figures are disputed, an estimated 1.2 million people were displaced between 1988 and 1994. ²  This estimate includes the roughly 360,000 Armenians who fled regions of the Azerbaijan SSR outside Nagorno-Karabakh and the roughly 205,000 Azerbaijanis who fled the Armenian SSR between 1987 and 1990. ³  But of immediate relevance to our investigation is the estimated 520,000-550,000 Azerbaijanis displaced from cities, towns, and villages in and around Nagorno-Karabakh between 1991 and 1994. ⁴  It is this forced exodus and the attendant takeover of the abandoned landscapes by the unrecognized NKR that left numerous historic mosques, mausolea, and cemeteries vulnerable to harm. Amidst the tragic human toll of the conflict, cultural heritage, the region’s repository of memory, history, and meaning has also suffered. 

Our analysis reveals several key findings that are elaborated further below. First, a plurality of Azerbaijani heritage sites remained unchanged from the late Soviet period to the 2020 war, demonstrating that neither the de facto authorities nor the general population sought to systematically eliminate the physical testimony of Azerbaijani cultural life in and around Nagorno-Karabakh.  

Second, a great many sites sustained some form of damage (44% including both major and minor damage). Major damage consisted mainly of large-scale looting, which took place predominantly in the decade immediately after the 1994 ceasefire, judging by the satellite evidence. Opportunistic stripping of cultural heritage sites occurred in the context of weak legal, political, and economic institutions that also allowed for the well-documented ruination of towns and cities once inhabited by Azerbaijanis displaced by the first war. As we discuss below, this decade of looting must also be considered in the context of post-Soviet economic collapse, which led to enormous flows of metal onto the scrap metal market driven both by the personal enrichment practices of the privileged and the subsistence practices of the poor. Be that as it may, de facto Armenian authorities of NKR did not safeguard abandoned heritage properties from looters who, in most cases, damaged monuments in the course of a wider search for fungible economic resources.

Third, our findings show that while the rate of outright destruction was low (n=16, or 15%), the deliberate targeting of mosques and mausolea increased over time, with most destruction episodes occurring after 2011. Although scattered widely, we identified three locations where destruction episodes were concentrated: in Shusha, in Papravend along the Aghdere/Martakert highway, and in the large Garaaghach cemetery, west of Aghdam. These three locales represent 71% of the discrete destruction events. CHW is unable to identify the agents or proximal causes of the destruction. But most took place at a time when Armenian perceptions of the seven territories had changed. Immediately after the first war, the seven territories were widely understood as a bargaining chip in negotiations over the final status of Nagorno-Karabakh. But early in the new millennium, this pragmatism had given way to a new geopolitical culture that dissolved the distinction between Nagorno-Karabakh and the surrounding territories. This shift may in part account for the increase in the rate of destruction after 2011, perhaps an effort by certain individuals or groups to purge the landscape of the material traces of Azerbaijani cultural life.

Although destroyed monuments cluster in the Aghdam and Shusha regions, there is no correlation between the proximity of heritage sites to places of Armenian settlement and the incidence of destruction. Moreover, there is a negative correlation between proximity to places of Armenian settlement and the intensity of damage. Our assessment of the relative condition of heritage sites compared to the surrounding residential and civic properties that Azerbaijanis were forced to abandon reveals that in most cases heritage sites were either less severely impacted than their surroundings or suffered the same impacts. Heritage sites suffered more severe treatment than surrounding buildings in only a minority of cases.

The Methodology section of this website details how we geolocated 116 of the 146 sites in our database using a wide range of photographs, videos, scholarly publications, topographic maps, and satellite imagery. We also describe our methods for determining the condition of cultural properties and define our seven condition assessments: unchanged, major damage, minor damage, destroyed, renovated, restored, and indeterminate. Our assessments rely heavily on declassified high-resolution American satellite imagery from the 1980s, namely the KH-9 Hexagon platform, which provides a ‘baseline’ for the condition of sites in the late Soviet period. It is important to emphasize that our evaluations refer to the structural condition of heritage sites, which can be determined from satellite images and the other image sources available to us. Certain forms of damage, such as vandalism of surfaces, bullet holes, decay, and erosion are beyond the scope of this investigation as they require in-person inspection. 

It is also important to note that this analysis does not undertake an examination of the wider landscape of cities and towns in the region that have been devastated by decades of conflict, animosity, and extractive economic practices. Our research mission focuses squarely on cultural heritage – public sites of collective history and memory – and does not undertake assessments of impacts on residential buildings, administrative buildings, or commercial buildings. Our primary concern is to understand what happened to Azerbaijani heritage between 1994 and 2020. Azerbaijan’s official list of heritage sites ( supplement No. 1 to decision No. 132 of the Cabinet of Ministers of the Republic of Azerbaijan  dating to August 2, 2001) provided the primary source for developing our dataset.

The threats to Azerbaijani cultural heritage after the First Nagorno-Karabakh War represent not an isolated instance, but a permutation of a broader phenomenon of heritage endangerment in the aftermath of ethnic conflict. In the report version of this investigation ( CHW Special Report #2 ), we compare in broad brushstrokes the fate of Azerbaijani cultural heritage from 1994-2020 with two other cases of large-scale population displacement, namely, the 1948 Arab-Israeli War and the 1974 Turkish invasion of Cyprus. This comparison offers instructive analogies and contrasts that can advance our understanding of how war and displacement affect cultural heritage. The most notable difference among the three cases under examination is the relatively limited reuse, renovation, or restoration of Azerbaijani heritage structures by the Armenian population of NKR. The most salient factor to explain this difference is the degree to which the victors in each of these armed conflicts sought to consolidate a new state through settler colonialism and the appropriation of properties and landscapes. Unlike in the cases in Cyprus and Palestine, over the course of three decades Armenians did not settle in large numbers in the seven territories surrounding Nagorno-Karabakh. In that sense, limited Armenian settlement and the attendant lack of development inadvertently protected Azerbaijani heritage from more severe impacts.

After two wars, it is significant that most of the mosques and mausolea in and around Nagorno-Karabakh are sufficiently intact to allow for restoration, raising broader questions about the future of the region’s cultural heritage writ large. Even as the threat has receded from Azerbaijani heritage (as of this writing, only two monuments remain under the control of the de facto authorities of Nagorno-Karabakh, the  Malibeyli Mosque  and  Khojaly Mausoleum ), scores of Armenian churches, monasteries, and historic cemeteries in territories now under Azerbaijani jurisdiction face threats of appropriation, damage, and destruction. CHW continues to closely  monitor  Armenian medieval and early modern heritage currently at risk in Azerbaijan, maintaining our  interactive dashboard  and releasing regular  reports . In the summer of 2022, we documented the destruction of an 18th-19th century Armenian church in the village of Susanliq/Mokhrenes. There is considerable danger that the next decade in Azerbaijani-controlled Karabakh will witness the juxtaposition of Azerbaijani heritage restoration with Armenian heritage destruction. Such an unraveling of the region’s cultural landscape would only make the heritage politics of the region worse, further jeopardizing the fragile remains of the region’s past.

In addition to this online environment, which presents the visual evidence for each site in Story Maps, our findings can also be accessed in our comprehensive  Special Report #2 . This report and supporting StoryMaps follow on CHW’s Special Report #1 on the ‘ silent erasure ’ of Armenian cultural heritage in the Nakhchivan Autonomous Republic of Azerbaijan. That study documented a policy of state-sponsored cultural erasure, resulting in the destruction of almost the entire repertoire of Armenian cultural heritage sites in that region. In this study, we document a complex pattern of heritage impacts that led to some destruction and much damage, but also left a plurality of Azerbaijani cultural heritage unchanged since the late Soviet period. The two cases are thus incommensurable in scope and organization. It is our hope that these two investigations will be read and considered together as chapters in an account of heritage caught up in a prolonged ethnic conflict. By presenting an empirical, objective account of heritage impacted by war, our goal is to ground debate in facts rather than propaganda and to contribute to the truth that must precede any long-term hope of reconciliation.

Charts 1-5 summarize the overall findings that are presented in exacting detail in the Site Inventory of this report and in the online StoryMaps. As Chart 1 shows, for this investigation we examined the records for 146 mosques, mausolea, and historic cemeteries. One of the challenges of this research was simply locating the monuments on the landscape, which we were able to do in the case of 116 sites (see Methodology-Geolocation). We have included in the report version of this investigation a Directory of Unlocated Sites in the hope that stakeholders with access to the region might provide new information that would allow us to update these findings.

After geolocating 116 heritage sites, we used high-resolution satellite imagery, photographs, videos, topographic maps, and scholarly publications to assess their condition. As Chart 2 shows, of the 109 sites whose condition CHW was able to assess, 42 (or 38%) remained unchanged from the late Soviet period through 2020, 39 (36%) suffered major damage, 16 (15%) were destroyed, 9 (8%) suffered minor damage, 2 (2%) were renovated, and 1 (1%) was restored. A list of all the heritage sites within each of these condition categories can be found in the Site Directory and Map 2 below of their location. Charts 3-5 and Maps 3-5 provide a breakdown of our findings for each type of site (mosque, mausoleum, cemetery).

What is immediately evident from these basic statistics is that the fate of Azerbaijani heritage under Armenian control was highly variable, showing no evidence of a singular governmental policy. As visible in Chart 2, a plurality of Azerbaijani heritage sites remained unchanged from the late Soviet period to the 2020 war and the rate of outright destruction was low (15%). But a great many sites sustained some form of damage (44% including both major and minor damage), and the total of heavily damaged (n=31) and destroyed (n=16) sites represents a majority of the dataset (51%). CHW delved deeper into these findings, analyzing them along several variables in order to reveal any patterns that might explain the treatment of Azerbaijani heritage under the control of the de facto Armenian authorities of Nagorno-Karabakh from 1994-2020. 

Maps 2 through 5 and Table 1 present CHW’s condition assessments of Azerbaijani heritage sites by region. Major damage or destruction took place in all regions except Khojaly, where the single mausoleum in our database remained unchanged. In most regions, our findings show that the treatment of monuments ranged widely to include destroyed, damaged, and unchanged sites.

The public availability of satellite imagery determines our ability to define date ranges for impacts on Azerbaijani heritage sites. The current lack of available high-resolution satellite imagery prior to 2002 greatly limits our understanding of the timing of impacts in the years immediately following the First Nagorno-Karabakh War, as does the limited satellite imagery coverage for many regions over subsequent years. 

But there is evidence to suggest that most of the major damage to Azerbaijani heritage took place in the decade after the 1994 ceasefire. As Chart 6 shows, of the 39 sites that sustained major damage, as many as 20 had already been impacted by the end of 2005. Most of the remaining 19 heavily damaged sites occurred in places where the earliest publicly available satellite imagery dates to 2009 or 2010, leaving the date of damage an open question. But even though high-resolution satellite imagery for this region was scarce prior to 2005, we can say with certainty that as many as five sites were damaged before 2004, and four sites were definitively damaged by 2002. An additional three sites that were subsequently destroyed (and therefore are not included in the count of 39 sites that sustained “major damage”) had also been damaged by 2002 ( N.5089 ,  N.5077 ,  N.5145 ). In light of the evidence, we deem it likely that most if not all of the major damage took place in the years immediately following the First Nagorno-Karabakh War.

It is extremely difficult to disentangle the convergence of factors that may have made such damage possible in the first decade. To address this problem, we first assessed the relationship between the impacts to cultural heritage and the broader decimation of towns like Aghdam and Fuzuli, and the extent to which Azerbaijani heritage sites might have been specifically targeted in the wide-scale ruination that consumed Karabakh during and after the first war. In the 73 cases in which heritage sites were located either within or near to villages and towns, CHW documented their relative condition compared to the surrounding residential and civic structures.

As Chart 7 shows, we found that in 34 of the 73 cases (47%), the heritage sites suffered the same degree of damage as the surrounding vernacular architecture. That is, the pattern of damage looks the same for the mosques as for other buildings in their vicinity, showing the removal of roofs and other building components, and subsequent erosion, decay, and ruderal plant colonization.

In another 23 of the 73 cases (32%), heritage sites were largely untouched (“unchanged”), or less impacted (“minor damage”) than the surrounding architecture, suggesting that in some cases heritage sites were either disregarded or intentionally left intact. Overall, of the 42 sites in CHW’s database that were assessed as unchanged, 18 of these were located in or near villages and towns that suffered large-scale looting, yet these sites remained largely unscathed.

In 14 cases Azerbaijani heritage sites were more severely impacted than their surroundings. In one of these cases (Pichanis Mosque,  N.4726 ), the heritage site sustained major damage, unlike the surrounding structures. In the remaining 13 cases, heritage sites were destroyed while structures in their vicinity were damaged but not razed (see discussion of destroyed sites below).

The building materials of certain heritage structures appear to have contributed to their preservation. Historic buildings with metal roofing that could be stripped and sold on the scrap metal market or reused elsewhere were more significantly impacted than structures with earth or stone roofing. Nearly all of the mosques assessed as “unchanged” have flat, earthen roofs, such as Chalabilar Mosque ( N.4170 ), Sardarli Mosque ( N.4231 ), Garadaghly Mosque ( N.4237 ), Damirchilar Mosque ( N.4703 ), Dondarly Mosque ( N.4704 ), Yusifbeyli Mosque ( N.4709 ), Khalaj Mosque ( N.4714 ), and Abdal Mosque ( NA.09 ). In several cases, the Soviet-era metal roofs were removed only to expose the traditional domes of a mosque that were left unharmed, as at Aghdam Mosque ( N.202 ), Gochamadli Mosque ( N.4233 ), Haji Abbas Mosque ( N.5073 ), and Chol Gala Mosque ( N.5103 ). In one instance, at the Jijimli Mosque ( NA.03 ), the Soviet-era metal roof was stripped to expose a flat earthen roof without domes that was not damaged. In only two instances did we identify “unchanged” mosques that appear to have metal roofs, namely, the Seyidli Mosque ( N.5151 ) and Mamay Mosque ( N.5166 ), both in Shusha.

Similarly, all of the mausolea located near or within areas that were once settled remained unchanged or only slightly damaged despite looting of nearby buildings, as at Ibrahim Khalil Khan Mausoleum ( N.4028 ), Mehdigulu Khan Mausoleum ( N.4029 ), Skhykhlar Mausoleum ( N.4166 ), Dagh Tumas Mausoleum ( N.4169 ), Yukhary Seyidahmadli Mausoleum ( N.4226 ), and Mir Mehdi Khazani Mausoleum ( NA.04 ). These structures have stone domes and limited metal content.

In sum, the evidence suggests that many of the mosques and mausolea that remained in sound structural condition may have escaped damage to some degree because value could not be extracted from their building materials. And conversely, virtually all of the sites that sustained major damage appear to have been targeted as part of the wider effort to mine fungible economic resources from abandoned property.

We did find a significant correlation between damage to Azerbaijani heritage sites and distance from habitation. Most of the sites that received either major damage (n=23) or minor damage (n=7) since the 1994 ceasefire were located far (> 2km) from any places of settlement. There are undoubtedly a number of factors that explain this complex picture, including the likely timing of damage in the first decade after the first war, the relative lack of Armenian resettlement in the eastern regions of Aghdam and Fuzuli (see below), and the proximity of a number of sites to the militarized line of contact.

The major damage of the post-war years thus cannot be explained as an isolated phenomenon, apart from the wide-scale looting of the abandoned towns and cities. But, while the devastating ruination of Azerbaijani settlements is well-known to those familiar with the conflict due to a proliferation of images in news and social media since the Second Nagorno-Karabakh War, researchers have yet to study it in any depth. The political, legal, and economic conditions that made rampant stripping possible, as well as the actors and networks involved, remain shrouded in darkness. Such questions are beyond the scope of this report, and may not be possible to address in full even with sustained ethnographic research, given the sensitivity of the matter. But the questions at stake are clear.

The first concerns the structure of governance, property relations, and legal jurisdiction in the seven territories in the years after 1994. What was the property status of the abandoned buildings, and what authority administered the territories after Azerbaijani expulsion? It is notable that the de facto National Assembly defined the borders of the new polity to include the wartime conquests only in June 1998 (a political geography later affirmed in the 2006 constitution). ⁵  This implies a void in executive, legislative, or juridical authority. According to Broers’ 2014 interview with Ararat Danielyan, de facto Minister of Justice of the NKR, Armenia’s legislation obtains in NKR, “until and unless the parliament adopts legislation.” ⁶  But it is not clear whether this legal arrangement was established by the mid-1990s, again leaving open the possibility of a juridical void, even within a de facto polity. It is reasonable to suppose some form of military governance in most regions adjacent to the Line of Contact. But the question remains: what authority and institutions held jurisdiction in the seven territories immediately after 1994, and thus the responsibility (even if not the will) to prevent the looting of abandoned properties, including sites of cultural heritage?

A second thorny question concerns the agents of looting and their motivations. Subsistence looting likely played a role, given the dire economic hardships of the 1990s across the former Soviet Union. The economic calamity that followed the collapse of the planned economy was particularly severe in Armenia and NKR due to blockades of road, rail, and pipeline transit by both Azerbaijan and its regional ally, Turkey. Scavenging for abandoned movable property and easily removable metal in this war-torn landscape may have been a strategy to alleviate destitution for some.

Yet research on the post-Soviet ‘ruins economy’ in Armenia has shown that subsistence looting, while relevant, does not alone account for the large-scale asset stripping of the mid-1990s and 2000s. ⁷  In Armenia, the post-Soviet elite (the former nomenklatura) who had benefited from the inequities of the privatization process pursued personal enrichment through the stripping of heavy machinery, building materials, and scrap metal from their newfound industrial assets. As Khatchadourian notes, “This pursuit of wealth accumulation was made possible by an unregulated market, a weak legal structure, and complicity between the private sector and government officials mired in graft. Market demand…was transnational, and it emerged primarily from Iran.” ⁸  Informants in Khatchadourian’s ethnographic research at decommissioned factories across Armenia consistently note that the booming post-Soviet scrap metal economy of the 1990s and early aughts was overwhelmingly directed toward Iran, whose brokers scoured Armenia’s factories, making deals with owners that led to convoys of metal hauled across the Armenian-Iranian border. ⁹  There is every reason to suppose that, in some measure, the massive flows of scrap metal stripped from abandoned properties in NKR belonged to this same system of privilege and extraction that was decimating Armenia’s Soviet factories in those same years.

In short, just as the stripping of Azerbaijani cultural heritage in Karabakh cannot be understood apart from the wide-scale looting of the landscapes within which it is situated, so too must the phenomenon be considered in the context of the wider, contemporaneous post-Soviet ‘ruins economy’, in which powerful, networked, oligarchic interests as well as impoverished individuals played a role. This is especially so given the extreme entanglements of the economic structures and political elites of Armenia and NKR, through both formal institutions and “informal sinews” (Broers 2021, 254). Yet while the processes are interrelated, the property regimes that undergirded them differed. In Armenia, it is possible to speak of “asset stripping” to the extent that much large-scale ruination of decommissioned Soviet factories is generally regarded to have been orchestrated by individuals who owned the properties in question, regardless of the corruption that pervaded and illegitimated the privatization process. In contrast, in NKR the ambiguous and unlawful post-war property conditions render the stripping of fungible materials nothing other than outright looting.

CHW’s analysis of the timing of destruction events shows that only a few Azerbaijani heritage sites were razed in the first decade after they fell under the control of Armenian forces. Chart 8 at right represents the date ranges for the destruction of 14 out of the 16 destroyed sites, excluding those whose destruction we attribute to excessive asset stripping rather than targeted demolition ( N.4239 ,  N.4171 ). In the decade following the 1994 ceasefire, three mausolea were destroyed across a widely dispersed area, suggesting distinct, isolated events ( N.4726 ,  NA.02 ,  NA.28 ).

The rate of destruction remained low well into the early 2000s. Two mosques ( N.5089 ,  N.5092 ) were destroyed sometime between 2004 and 2013. Both were located in the city of Shusha. Site N.5089, the Hajji Yusifli Mosque, initially suffered major damage in the decade after the first war but by 2013 the remains of the mosque’s foundations had been removed and a new structure – likely a residential building – was constructed at the site. The Julfalar Mosque (N.5092), in contrast, survived intact well into the 2000s, with satellite data showing a largely undamaged structure as late as February 2004. By July 2013, the mosque had been removed as construction began on a large building, which from its scale appears to have been a civic or administrative building. In sum, between 1994 and 2013, we have evidence for five destroyed heritage sites: three mausolea in disparate parts of the seven territories and two mosques in Shusha.

Between 2011 and 2020, the deliberate targeting of Azerbaijani heritage sites increased, with nine sites destroyed in the subsequent eight years. Due to the increased availability of commercial satellite data for this decade, it is possible to resolve the dates of several of these destruction episodes rather narrowly. Four sites in close proximity were destroyed between 2011 and 2014: three mausolea in the historic Garaaghach cemetery ( N.4056 ,  N.4057 ,  N.4058 ) appear to have been destroyed, perhaps in a single destruction event. At some point in that same interval, the cemetery ( NA.29 ) as a whole was burned.

We identified another cluster of two destroyed sites in the area of Papravend. Between 2014 and 2019 the Bashsyz Mosque ( N.4045 ) and a mausoleum ( N.4044 ), located across the street from one another, were destroyed. Given their proximity it is possible, but not clearly demonstrable, that these two events were related.

The most recently destroyed sites documented by CHW include two additional mosques in Shusha and a mosque in Fuzuli. In Shusha, the Mardinli Mosque ( N.5077 ) had been renovated into a movie theater during the Soviet period, and hence the sacredness of the site may have already been lost to memory before the first war. Nevertheless, the historic structure suffered major damage prior to the earliest commercial satellite imagery in 2002. The ruins of the structure were still visible in August 2016, but by September 2018 the remaining walls had been completely removed and the site cleared. Shusha’s Kocharli Mosque ( N.5145 ) also suffered major damage in the decade following the first war, but its superstructure was still intact in 2002. The mosque continued to slowly degrade over the following decade and a half, until sometime just prior to June 2019, when it was destroyed and the site graded.

The Hajji Alakbar Mosque in Fuzuli ( N.4208 ) was also impacted during the Soviet era, when its traditional domes were covered by a metal roof. The mosque survived with only minor damage to its roof at least until June 2018. Sometime between that date and August 2019, the structure was completely demolished.

Two aspects of the destruction of Azerbaijani heritage are particularly striking. First, although scattered widely, we do find three locations of heightened impacts: in Shusha, in Papravend along the Aghdere/Martakert highway, and in the large Garaaghach cemetery, west of Aghdam. These three locales represent 71% (n=10) of the 14 episodes CHW identifies as targeted destruction. The remaining four episodes are geographic isolates.

Second, we underline the compressed timing of these episodes of heritage destruction. 5 sites were destroyed sometime between 1992 and 2013 (a cumulative rate of roughly one site every four years). In contrast, nine sites were destroyed in the eight years between 2011 and 2019, a cumulative rate of approximately one site every 11 months. The final years before the second war clearly witnessed a relative acceleration in severe impacts to Azerbaijani heritage sites.

In an effort to explain why only some sites were destroyed, one variable we examined was geographic location. In particular, CHW examined the proximity of destroyed heritage sites to inhabited settlements but found no clear patterning (see Chart 9). Half (n=8) of the destroyed sites were located within inhabited places or in very close proximity to them (within 2km), while the remaining 8 destroyed sites were far from inhabited places (more than 2 km distant). In other words, there was no correlation between proximity to settlement and incidence of destruction.

CHW is not able to determine the specific agents or the proximal causes of the destruction events, but it is notable that the acceleration in destruction after 2011 took place at a time when an expansive vision of NKR’s geography was already firmly entrenched in Armenian geopolitical culture. In what Laurence Broers has called “a kind of inexorable geopolitical mission creep”, by the mid-2000s the seven territories were no longer framed expediently as a ‘buffer zone’ or a ‘security belt’ that would be traded as collateral in return for Armenian demands but, rather affectively, as ‘liberated territories’ belonging to an “augmented Armenia”. ¹⁰  Toponymic and cartographic changes supported this shifting geopolitical vision. It is not overly speculative to suggest that this well-documented transformation in the geopolitical imagination of the region might also have provoked the disparate and disorganized incidents of destruction documented by CHW.

CHW’s investigation revealed relatively few instances of renovation, restoration, and reuse of Azerbaijani heritage sites and hence there is very little data for defining and assessing possible patterns. Renovations and restorations were concentrated entirely in the city of Shusha. The Upper Govhar Agha Mosque underwent a  contentious  restoration in 2019, which framed the structure as a Persian, rather than Azerbaijani, heritage site. The Taza Mahalla Mosque ( N.347 ) was renovated in 2014 and turned into a geology museum. And the Saatli Mosque ( N.355 ) was renovated between 2004 and 2013, but CHW was unable to ascertain the function of the renovated structure.

We documented seven instances of informal agrarian reuse, in which mosques that were either unchanged or had sustained major damage were repurposed as animal pens (i.e.,  N.202 ,  N.4052 ,  N.4212 ,  N.4216 ,  N.4217 ,  N.4709 ,  N.4713 ). In four such cases, satellite imagery makes it possible to determine the approximate date when the mosque was reused as a barn. Giyasly Mosque (N.4052) was turned into a barn between December 2011 and June 2014, during which time the exterior areas to the north were cleared to contain animals. At the Merdinli Mosque (N.4217), a roof was added to a heavily damaged mosque and an exterior area was cleared for livestock between 2012 and 2019. Having sustained major damage, Qajar Mosque (N.4216) was given a new roof by 2013, and by 2018, areas to the north of the mosque were cleared for tending animals. Satellite imagery also shows that an area to the north of the Haji Giyaseddin Mosque (N.4212) was cleared, likely for keeping animals, between 2014 and 2019. Where it can be documented by satellite imagery, agrarian reuse was a relatively late development, occurring after 2011.  

****

Overall, the patterning of the impacts to Azerbaijani heritage between 1994 and 2020 documented by CHW does not indicate a program of cultural erasure implemented and sustained by a state actor. Instead, the patterns provide evidence of an initial phase (1994-2011) of extensive collateral damage caused primarily, but not exclusively, by the mining of building materials. By 2011, the available resources for looting were largely exhausted across the region. But what we found instead was an increase in the rate of targeted destruction concentrated primarily, but not exclusively, in three geographic areas. These patterns set the treatment of Azerbaijani heritage between the wars in a rather unique frame of reference, fundamentally different from the project of state-sponsored cultural erasure that we documented in Azerbaijan’s demolition of Armenian heritage in the autonomous region of Nakhchivan. But impacts to Azerbaijani heritage in Nagorno-Karabakh also cannot be understood as solely a problem of neglect or failed stewardship. In the next section, we seek to better define the patterns we have documented by setting Azerbaijani heritage between the wars in a comparative context.

CHW’s methodology for this investigation consisted of two key components: 1) a database of Azerbaijani heritage sites compiled from official government listings, academic publications, cartographic sources, and public media sources, and 2) high-resolution Cold War-era declassified satellite imagery and publicly available modern satellite imagery that can allow for precise geolocation and condition assessments as well as aid in the approximate dating of destruction events.

It is important to note that in academic parlance, there is a significant difference between Azerbaijani heritage and the heritage of Azerbaijan. Azerbaijani heritage refers specifically to the material and intangible legacy of the Turkic communities of the region. The heritage of Azerbaijan refers to the heritage of all ethnic and religious groups that falls under the stewardship responsibility of the modern Republic of Azerbaijan. 

The baseline source for CHW’s database of Azerbaijani heritage in Nagorno-Karabakh and the surrounding territories is the  supplement No. 1 to the decision No. 132 of the Cabinet of Ministers of the Republic of Azerbaijan  dating to August 2, 2001, which lists all heritage sites under state protection (fig. 1). Based on the supplement, CHW identified 112 mosques, mausolea, and historic cemeteries (pre-dating the 19th century) located in the region administered by the NKR from 1994-2020. In addition to these 112 sites, CHW identified an additional 32 mosques, mausolea, and historic cemeteries using information provided by the  State Committee on Religious Associations of the Republic of Azerbaijan , the  Karabakh Center project , the Ministry of Culture of the unrecognized NKR and various academic publications and cartographic sources. Islamic heritage sites are significantly underrepresented on the official monument lists of the NRK, which was compiled before the 2020 war, but we nevertheless consulted this resource as well. The integration of all these sources resulted in a dataset composed of 146 Azerbaijani cultural heritage sites, including 75 mosques, 55 mausoleums, and 15 cemeteries (see Chart 1).

Geolocation

Once the sample of 146 sites was defined, CHW’s lead researcher for this investigation, Dr. Husik Ghulyan, transformed the inventory into a structured geodatabase. This required geolocating each site with exacting precision to determine its geographic coordinates. These coordinates then allowed us to assess changes in the condition of each heritage site using historical satellite imagery. The only locational information provided in the official state list of heritage sites of the Republic of Azerbaijan is the village name, city name, or, in a few cases, a geographic area. The same is true for the other sources that CHW consulted to create the database, with the exception of Soviet-era topographic maps. In order to geolocate the 146 sites in our database, we used four main sources: photographs and videos, topographic maps, scholarly publications, and satellite images.

Photographs and videos

Since the  official registry of heritage sites of Azerbaijan  does not provide precise locational information, photographs and videos were the most important source for geolocating the monuments and cemeteries in the database. CHW first tried to obtain images of the sites that pre-date the First Nagorno-Karabakh War. The old images were helpful both for geolocation, and to determine the baseline condition of the sites prior to the outbreak of the first war. In some cases, the old images were also helpful for disambiguating monuments that are closely located to one another, since the official registry mentions several sites (e.g., mausolea) in close proximity. But most photographic and video evidence that CHW used for geolocation purposes were those post-dating the 2020 Nagorno-Karabakh War. Numerous images and videos disseminated on social media and news sources about villages that had been transferred to Azerbaijani control after the 2020 war allowed CHW to precisely locate the sites and understand their most recent condition. As a part of geolocation, Dr. Ghulyan examined hundreds of hours of video footage and thousands of photographs captured and taken after the 2020 ceasefire in order to identify the relevant sites and geolocate them on satellite imagery. Photographs were also an important source for identifying additional mosques and mausolea that are absent in the official registries of heritage sites and thereby expanding CHW’s database.

Scholarly publications

CHW used scholarly publications to obtain additional locational information, archival images, sketches, plans, information on site condition, and other important attributes of entities in our database (fig. 2). For example, Dr. Ghulyan was able to disambiguate closely located mausolea by referring to their Soviet-era images and descriptions found in Soviet and Azerbaijani scholarly publications (e.g.  Panah Ali Khan Mausoleum ,  Ibrahim Khalil Khan Mausoleum ,  Ughurlu Bey Mausoleum ).

Topographic sources

In some instances, topographic sources allowed for the approximate geolocation of sites, after which CHW obtained the precise coordinates using historical and modern satellite imagery. Topographic sources also helped us to identify several mausolea that are not on the official registries. CHW’s research drew mostly on 1:25K and 1:50K scale topographic maps produced between the 1950s and the 1980s by the General Staff of the Soviet Army and the Main Department of the Geodesy and Cartography of the Ministers' Council of the USSR.

Satellite imagery 

Satellite imagery was the most important source for establishing high-precision coordinates. CHW made extensive use of the most recent satellite imagery, as well as historical satellite imagery gathered by Maxar Technologies, CNES, and Airbus, and publicly available through the Google Earth platform. Satellite imagery also allowed us to assess the changes in the condition of heritage sites over time and the approximate period of impacts.

For condition assessment purposes, as a baseline source CHW turned to declassified imagery captured by the KH-9 Hexagon satellite, a photographic reconnaissance program launched by the U.S. National Reconnaissance Office between 1971 and 1986. The ground resolution of the Hexagon high-resolution cameras was 2-4 feet (0.6-1.2m). The KH-9 archive was declassified in September 2011, making a vast trove of spatial imagery available to researchers, including a large corpus of materials documenting the territory of the former Soviet Union in its final decades. It is available to the public through the USGS’s EarthExplorer site.

We georeferenced Hexagon photographs in ArcGIS Pro using ESRI’s satellite base layer imagery. The resolution of the Hexagon imagery is more than adequate for discerning large built structures, such as mosques and mausoleums. But cemeteries proved more difficult for our baseline assessments. Since the Hexagon imagery used in our investigation has a resolution of 0.6-1.2m, features less than 1.5m or so wide can sometimes be difficult to discern. Cemeteries thus pose a challenge to satellite investigations since an individual tombstone might provide a lateral extent well under a meter. However, relatively large fields of burial monuments are discernible in satellite imagery, as they produce patterns of light and shadow that result in a stippling effect on the terrain or, in some cases, a jagged effect very distinct from the surrounding ground cover.

In cases where sites could not be confidently located with precision on archival or modern historical satellite resources, that is, when CHW was unable to locate the sites and obtain visual proof of their existence, we designated them as “unlocated”. We assessed 30 of the 146 entities in our dataset (20%) as unlocated. If any of them come to be located following the completion of this report, we will update our results.

CHW’s assessments are based on visual comparisons between the condition of sites as visible in historical baseline imagery, and in recent satellite imagery, photographs, and videos. It is important to emphasize that our assessments refer to the structural condition of heritage sites, as this can be determined from satellite images and the other image sources available to us. Certain forms of damage, such as the elimination of inscriptions or decorative elements, bullet holes, decay, and erosion are beyond the scope of this investigation. We used seven assessment categories, defined as follows:

Unchanged denotes a heritage site that has not changed appreciably since the late Soviet period. To qualify as unchanged, a site must have maintained the same structural form between the First and Second Nagorno-Karabakh wars, even if time and neglect have allowed it to fall into some minor disrepair. In most cases, heritage sites that are unchanged are well-preserved (see Site Inventory). But the assessment category also includes mosques or mausolea that were already damaged in the late Soviet years, such as Yenikend Tomb ( N.393 ), Sharifan Mausoleum and Tomb ( N.394 and N.1824 ), Boyahmadli Mosque ( N.4054 ), Shykhlar Mausoleum ( N.4166 ), Garghabazar Mausoleum ( N.4213 ), Garygyshlag Mosque ( N.4733 ), two mausolea in Shusha ( N.5042 ,  NA.20 ), Mir Mehdi Khazani Mausoluem ( NA.04 ), and a Tomb in Aghdam ( NA.30 ). These sites are unchanged as they do not appear to have suffered any further damage while under the administration of the unrecognized NKR.

Indications of Minor Damage include impacts to exterior facades or decorative elements but the structural integrity of the building or site remains preserved (e.g., Aghdam Mosque,  N.2020 , Gurjulu Maosoleum,  N.307 , Gochahmadli Mosque,  N.4233 , Chol Gala Mosque,  N.5103 ). Other forms of minor damage include a small hole in a roof (Kar Gunbaz Mausoleum,  N.312 , Abdal Mosque,  NA.09 ), damage to a wall (Mammadbeyli Mausoleum,  N.391 ), or limited encroachment on the territory of a cemetery ( N.5751 ).

An assessment of Major Damage indicates that a heritage site has sustained significant impacts since the late Soviet years that have compromised its architectural integrity. In the case of mosques, major damage typically entailed the removal of roofs, which resulted in significant impacts on the structure and its interior (e.g., Dashkasan Mosque,  N.4172 , Gejagozlu Mosque,  N.4223 , Qajar Mosque,  N.4216 ). In the case of mausolea, major damage denotes significant impacts to the dome, roof, and sometimes the walls of the structure (e.g., Melik Ajdar Mausoleum,  N.311 , Panah Ali Khan Mausoleum,  N.4027 , Khanazur Mausoleum,  NA.15 ). Major damage to historic cemeteries denotes the visible disturbance of tombstones or excavations in the territory of the site that clearly impact buried remains (e.g., Papravand Cemetery,  N.5750 , Karkhulu Cemetery,  N.5791 , Imarat Cemetery,  NA.25 ).

CHW assesses locations as Destroyed when virtually nothing remains of a heritage site that was still standing in the late Soviet years. It has either been stripped to its foundations (e.g., Papy Mosque,  N.4171 , Pirahmadli Mosque,  N.4239 ); completely erased (Mardinli Mosque,  N.5077 ; Haji Yusifli Mosque,  N.5089 ; Julfalar Mosque,  N.5092 ; mausoleum in Aghdam,  NA.26 ); burned (Garaaghach Cemetery,  NA.27 ); or reduced to rubble (such as three mausolea in Aghdam,  N.4056 ,  N.4057 ,  N.4058 , the Haji Alakbar Mosque  N.4208 , Sary Masyg Mausoleum,  N.4726 , Kocharli Mosque,  N.5145 , Rzaqulu Mausoleum,  NA.02 ). 

A limited number of sites were Renovated to serve a different purpose, as was the case at the Taza Mahalla Mosque ( N.347 ) and the Saatli Mosque ( N.355 ).

Only a single site in CHW’s database was Restored such that the building was repaired while retaining its original function: the Upper Govhar Agha Mosque ( N.344 ).

In some cases, CHW was unable to make a determination concerning the changing condition of the heritage site with the sources of evidence available to us. These sites were classified as Indeterminate.

 ¹  Weizman, Eyal. 2017. Forensic Architecture: Violence at the Threshold of Detectability. Brooklyn, NY: Zone Books, 128.

 ²  Broers, Laurence. 2021. Armenia and Azerbaijan: Anatomy of a Rivalry. Edinburgh: Edinburgh University Press, 122.

 ³  ibid.

 ⁴  ibid, 123.

5 ibid, fn. 29, 336.

 ⁶  ibid, 253.

 ⁷  Khatchadourian, Lori. 2022. “Life Extempore: Trials of Ruination in the Twilight Zone of Soviet Industry.” Cultural Anthropology, 37(2): 317-348.

 ⁸  ibid, 324.

 ⁹  ibid, 333.

 ¹⁰  Broers 2021, op. cit., 98.