Introduction

Sustainability, environmental protection, and biodiversity conservation are key objectives for European policymakers1 as they pursue design-based actions for climate change mitigation2, Green Deal3, and sustainable development4. The signing of the Bern Convention on the Conservation of European Wildlife and Natural Habitats (Bern Convention) in 19795 was widely lauded as a key starting point for substantial progress in achieving ecological sustainability in Europe. Among other achievements, the Bern Convention enabled joint scientific efforts among member countries to help underlie sustainable development and biodiversity conservation policies focused on wildlife habitat preservation6, bird conservation7, and invasive alien species management8 in European ecosystems. The Bern Convention also enabled state-level legislative action on the same sustainable development and biodiversity conservation goals9. In large part because of the Bern Convention, the European Union (EU) and neighboring countries created a transboundary conservation network called the Emerald Network, also called Nature 2000 for EU countries9. Beyond protecting endangered flora and fauna across a broad spatial area, the Emerald Network promoted sustainable human-nature interactions9,10. The Emerald Network required signatories to create special sites for biodiversity and endangered species conservation, implementation of EU environmental-protection legislation, and policies harmonized across the global network10. Nominating a site to be included in the Emerald Network involves four steps: (1) a national environmental commission identifies national areas of significant environmental value, (2) formal documentation is prepared and submitted, detailing the habitat characteristics of rare flora and fauna of the sites, (3) nominating the sites for including, and (4) integrating the new site into the international network through the conclusion of a biogeographical evaluation seminar11. By implementing policies to decrease soil degradation and erosion and improve climate regulation, water purification, and food production, for example, the Emerald Network has helped advance sustainable development and democratic decision-making concerning the environment. Positive impacts from the Emerald Network can be observed in two decades of reforestation trends in Europe12 as well as long-term positive forecasts for future European forest extent13 compared to global forest loss estimates of 2.3 million square kilometers during approximately the first decade of the 21st century14. Developing adaptive agri-environmental systems15 and conservation practices16 has helped to increase the ability of the Emerald Network signatories to maintain healthy agro-ecosystems while continuing to restore natural vegetation17 and protect species population growth rates18. Many other examples of conservation success due to the Emerald Network exist, for example, the stabilization of the International Union for Conservation of Nature (IUCN) Red-List species population dynamics in different countries such as Armenia19 and Ukraine20.

Ukraine joined the Bern Convention in 1996 and started working towards the implementation of associated policies, environmental-protection legislations, and creation of Emerald Network sites in 2000 with the adoption of the national state program “National Program for the Formation of the National Ecological Network of Ukraine for 2000–2015”21. Today these sites serve as the southwesternmost boundary of the seven-country Emerald Network22. Ukraine’s commitments to the Emerald Network were originally slated to be fully implemented by 2020, as outlined in the 2015 revised schedule for Emerald Network Implementation23. Ukraine has successfully included 377 sites in the Emerald Network, and currently, 160 of these sites are undergoing the verification process.24,25. The Russian Federations’ launch of a military offensive in at least eight oblasts (regions) in Ukraine in early 2022, including the eastern oblasts of Donetsk and Luhansk which were already affected by occupation starting in 2014, substantially distributed the Emerald Network implementation process. Parts of Ukraine’s Emerald Network sites were occupied via militarization by the Russian Federation. This conflict has not only provoked an economic and humanitarian crisis in Ukraine26,27,28, bringing wide-spread destruction to industrial and civilian infrastructure and damaging socio-ecological systems28,29,30, but also separated territories from conservation legislation, policies, and agreements31. Among the myriad of sustainable development and biodiversity conservation consequences stemming from this conflict are physical damage to protected areas and the absence of management and actions needed for adaptive management and evidence-based decision making31,32. New complex and dynamic risks to sustainable land use and conservation have emerged, in particular, new hotspots of high-level land cover/and use changes28 and reduced socio-economic activities33. Ukraine has consistently demonstrated a commitment to restoration according to official statistics and remote sensing observations34. Such trends can be observed in territories that have been impacted by the war. For example, the “Kreidova Flora”, an Emerald Network site located in the northern part of Donetsk oblast increased in pine tree cover area from 198.9 ha in 2009 to 316.9 ha in 201535.

The Luhansk region, characterized by flat terrain and a temperate climate, encompasses three distinct types of terrain: forests, steppe, and sand dunes36. Presently, deciduous forests cover 8% of the land area, predominantly along the banks of the Aydar and Siverskyi Donets rivers. These forested areas consist mainly of floodplain forests and bairak forests (i.e., broad-leaved forests situated on the upper reaches and slopes of streams), classified under the 2nd and 1st protection categories. The regulations associated with these categories prohibit “main purpose forest use” logging in protected forests, and the 2nd category permits limited forest use, allowing only tree felling for forest development and enhancement, such as maintenance and sanitary logging. Additionally, the Luhansk region features artificial forests predominantly comprised of coniferous trees, which have been planted on the plains. Fifteen Emerald Network sites that cover over 326 thousand hectares (th. ha) of area in this region were created to protect at least 307 animal and 139 plant species36 listed as endangered or under threat of extinction in the Ukrainian and European Red Books and Bern Convention Resolutions. These Emerald Network sites also cover seven extremely valuable natural forest habitat types protected at the pan-European level and listed in the Resolution 4 of the Bern Convention37. War in the Luhansk region is now associated with a range of ecological impacts related to militarization, abandonment of land, and specifics of industrialization that increased vulnerabilities of ecosystems32. The largest land cover, land use, and land management changes on both parts of the Luhansk region divided by the front line are related to militarization that damaged infrastructure and natural lands. For example, large wildfires on both parts of the Emerald Network, divided by a front line, were recorded in July 202038 and after the February 24th, 2022 large-scale invasion of Ukraine. Flooding of abandoned or damaged coal mines disturbed the water balance and decreased water quality due to consequent pollution of the region39. For instance, a disruption in the electricity supply at the “Zolote” coal mine resulted in its flooding, posing a substantial risk of pollution to the Siversky Donets river39,40. Since 2014, recurring mine flooding incidents have caused a notable increase in the concentration of pollutants such as lead, nickel, cadmium, and other persistent organic substances in the water bodies within the Luhansk region41. Extensive damage to Ukraine’s energy infrastructure has led to interruptions or inefficient operation of water treatment facilities, resulting in a substantial decline in water quality standards due to the presence of polyaromatic hydrocarbons, volatile organic compounds, octylphenol, heavy metals, and mercury42. One of the most hazardous areas in eastern Ukraine is the “Yunkom” mine, situated in the Donetsk oblast near the administrative border of the Luhansk region. This mine was the site of a Soviet government nuclear explosion in 197943. Since then, the mine has implemented special exploitation restrictions, including underground water pumping, to prevent radiation contamination of neighboring rivers and the Azov Sea. Since 2017, exploitation restrictions have been halted by the Russia-controlled local administrations; the current state of the nuclear waste storage facilities deep in the mine is unknown.

Modern warfare in Ukraine has dramatically increased risks to both the regional and global environment44,45. In the Luhansk region, both active and decommissioned industrial facilities house chemicals; damage from artillery shelling, fire, and electricity cuts release chemicals and cause substantial soil, water, and air pollution. For instance, in March and April of 2022, explosions on the grounds of the former “Barvnyk” factory in Rubizhne released unhealthy amounts of aromatic hydrocarbons and their derivatives into the atmosphere44. In April 2022, two powerful explosions at the “Zorya” chemical factory released a massive cloud of nitrogen oxide46. In March 2022, an explosion at the railway station in Rubizhne city resulted in the release of chemical products from the “Zarya” and “Azot” plants, leading to pollution of nearby areas, including water reservoirs46. Beyond explosions from missiles and carpet bombs at industrial chemical facilities, air pollution has been exacerbated by hard-to-extinguish fires at oil refining and storage facilities47,48. Otherwhere examples of environmental impacts from the war include the cessation of insect control activities, sanitary logging, and forest “fuel treatments” techniques, all of which are essential forest management practices implemented by Ukrainians but halted by Russia-controlled administrations. Forest management innovations such as the implementation of electronic wood accounting49 and an electronic open register of forest tickets (i.e., an informational portal granting access to all issued logging approvals)50 have helped reduce illegal logging and associated trade. These systems facilitate wood chipping51 and electronic monitoring of its condition, purpose, transportation, and export52, as well as the oversight of legal logging activities. The lack of monitoring has left the logging situation in the separated territories unregulated and unsupervised.

Modern warfare has also had profound impacts on the well-being of communities, particularly those directly relying on natural resources for their livelihoods (e.g., agriculture). Unsurprisingly, international laws emphasize the need for restoration of and compensation for war-damaged ecosystem services53. The destruction wrought by war goes beyond direct damage to ecosystems through shelling; it extends to the indirect effects of military infrastructure development, weapons testing, and the disposal of military equipment and ammunition once peace is secured54. The use of artillery and landmines not only inflicts direct harm on people but also triggers a cascade of secondary impacts from soil acidification55, disruptions in soil formation due to soil material displacement56, and contamination with metal and plastic fragments. Moving military equipment such as heavy tanks can lead to altered soil hydraulic properties, erosion, liquefaction, and mud formation57,58. Restoring damaged soils is a costly and time-consuming process59. Furthermore, leaking diesel fuel and other chemicals into the soil and underground water can take decades to remediate. The presence of unexploded shells and land-mine fragments in the soil heightens the risks and complexities of cleanup efforts, as their magnetic properties can hinder metal detectors, making the process extremely dangerous for remediation. Landmines threaten local populations and wildlife species, as explosives may go undetected60. These adverse effects underscore the critical importance of addressing the impact of warfare on biodiversity and wildlife populations, which is of paramount concern for the preservation of nature itself. The study conducted by Daskin et al. in 201861 sheds light on the profound impact of conflicts on Africa’s natural reserves from 1946 to 2010. This research highlights the pivotal role of conflict frequency as an essential variable directly affecting wildlife populations within these reserves, showing a strong correlation with significant wildlife population declines. On a global scale, military conflicts emerge as potent catalysts for species extinction, posing a grave threat to biodiversity. Furthermore, the ongoing trends in conjunction with other drivers of climate change, such as greenhouse gas emissions and deforestation, paint a dire picture for the future. If these patterns persist, it is projected that by 2060–2080, we may witness the extinction of as much as 10–15% of animal species62. However, the situation becomes even more alarming when we consider the potential consequences of a nuclear conflict, where this staggering figure could skyrocket to a devastating 20–50%62. The war in Ukraine can have catastrophic consequences for European ecological security and biodiversity. According to the Land-Mine Monitor reports in July 2022 over 280,000 km2 were exposed to the conflict, 21,000 km2 of which were exposed in 201863. The environmental devastation, coupled with its adverse effects on Ukraine’s ability to fulfill its climate change commitments and SDGs64, as well as the reliability of the Russian Federation as a partner for climate change initiatives65, will have far-reaching and detrimental consequences on the long-term prospects of sustainable development and climate resilience in the region.

In summary, the war has inflicted a range of harms on the socio-ecological system which are very likely to have lasting socio-ecological impacts. Assessing the environmental consequences of these events is nearly impossible due to the lack of physical access to these areas. To these ends, we aimed to explore the impact of the militarized occupation of natural protected areas and the subsequent interruption of conservation efforts on ecosystem sustainability. Such research can be conducted only using long-term observations before the military conflict and after the hot phase of the conflict and the establishment of a stable demarcation line. Due to the active warfare in 2022 and 2023, it is premature to quantify ecosystem conservation consequences. Thus, we focused on the progress of the Emerald Network establishment in the Luhansk region (Fig. 1) in terms of land-cover changes in the environmental protection zones from 1996 to 2020.

Fig. 1: Emerald Network sites in Luhansk region separated by demarcation line.
figure 1

Each unique ID of the Emerald Network site is listed in the European Union documentation77 and linked to the centroid of the polygons represented on the map. The gray hashed area represents territory under Russian control after 2014 with an established demarcation line in 2015. The hashed area represents Emerald Network sites established on the blanks of Siverskyi Donets river.

Full size image

Quantification and evaluation of policies and processes underlying the establishment of Emerald Network in the Luhansk region was conducted by analyzing forest cover changes trends in the territories on both sides of the conflict demarcation line established after the “Minsk 2” and “Misnk-3” agreements in 2014 and 2015. Utilizing the principles of deforestation pressure-based management regime comparison66, we evaluated the effectiveness of the Bern Convention’s conservation policies and examined the impact of the territory’s separation on the environment by comparing forest area changes in territories under Ukrainian and Russian control. Gathering ground-referenced data in active conflict zones is impossible because of the high risks to security. Consequently, we employed remote sensing-based approaches, commonly used for assessing areas affected by warfare, to analyze changes in land cover and land use before, during, and after the conflict. A similar approach was employed in a study by Aung67, which investigated the environmental impact of armed conflict in Rakhine. We chose 1996 to 2020 as a research time period for the analysis because during these years Ukraine joined The Bern Convention, and the planned final year for Emerald Network establishment23 (excluding time after the catastrophic wildfire38). This time period is further divided into three parts. The first (from 1996 to 2000) shows the trend before the creation of the Emerald Network. The second (from 2000 to 2013) shows the progress of the Emerald Network establishment before the conflict. The third (from 2013 to 2020) shows the progress of conservation after the beginning of the conflict. After the end of the Russian–Ukrainian war, the establishment of a stable demarcation line, and ecosystem’s restoration process beginning, the methodology presented in this article can be extended to other regions of Ukraine for the post-war ecosystem conservation damage assessment.

Results

We carried out a land-cover change assessment with the use of generated maps of the Luhansk region for 1996, 2000, 2013, and 2020 (Fig. 2). The accuracy values of tree cover maps in terms of F1 score (which is the harmonic mean of the user’s accuracy and producer’s accuracy68) were 0.9, 0.9, 0.84, and 0.88, correspondingly (Supplementary Table 1).

Fig. 2: Land-cover classification maps for Luhansk region from 1996 to 2020.
figure 2

Land-cover classification at 30 m spatial resolution was conducted based on the compositing of Landsat-5 data for 1996, Landsat-5 and 7 for 2000, Landsat-8 for 2013, and Sentinel-2 for 2020. The dashed line represents the demarcation line established after 2015.

Full size image

Between 1996 and 2000, the annual deforestation rate on the territories under Ukrainian control was −0.86 ± 0.22 th. ha per year or overall −3.42 ± 0.86 th. ha (Fig. 3). A similar rate was observed in territories taken under Russian control after 2014 at an annual rate of −0.28 ± 0.1 th. ha per year or overall −1.1 ± 0.4 th. ha. Since 2000 Ukrainian government considered the creation of Emerald Network sites as a priority for the short- and long-term environmental-protection strategies69,70. After work began on establishing Emerald Network sites and implementing EU sustainable development policies, changes in trends on both parts of the Emerald Network are evident. Between 2000 and 2013, annual reforestation rates in territories under Ukraine control were +0.67 ± 0.09 th. ha per year with +8.7 ± 1.2 th. ha total forest area growth. At the same time, territories that were to be under Russian control had an annual rate of +0.19 ± 0.06 th. ha per year and total +2.48 ± 0.79 th. ha growth. The military conflict in 2014, and the subsequent segmentation of territory on the occupied and non-occupied by demarcation line, changed trend’s patterns during 2013–2020. During this time period, we found that territories that remained under Ukraine’s control kept in place reforestation and conservation processes with +1.19 ± 0.18 th. ha per year annual rate and total area in an increase of +8.3 ± 1.25 th. ha. However, territories that were taken by the Russian control experienced rapid deforestation of −1.23 ± 0.15 th. ha per year annual rate and −8.6 ± 1 th. ha total forest area loss. Results (Supplementary Tables 2 and  3) indicate that territories remained under Ukraine’s control even under the conditions of military conflict with increased vulnerability and consequent ecological problems31,32,36 in the region continued progression of conservation while territories under Russian control lost 20 years of sustainable development progress with 25% of forest loss (compared the 2013 estimates).

Fig. 3: Tree cover area changes for Emerald Network sites in the Luhansk region.
figure 3

The top plot shows the total areas for territories under Ukraine (blue) and Russia’s control (red). The bottom plots demonstrate net gain changes and net gain rates for respective time periods. Error bars represent standard error.

Full size image

The majority of forest area in the Luhansk region is concentrated in the floodplains of the Siverskyi Donets River site (Fig. 4) and was divided into two parts by a demarcation line. Both parts have the same ecological communities of flora and fauna36 and are equally vulnerable to ecological problems due to post-military action damage. Before the conflict in 2014, this area was entirely under Ukrainian government control and not segmented; this is reflected by the uniform and consistent land-cover change trends before the war. However, after the partial separation of the region, we observed severe deforestation. Between 2013 and 2020, Ukraine-controlled territories gained 18% of forest area, while Russia controlled lost 31%.

Fig. 4: Emerald Network sites established on the blanks of Siverskyi Donets river in Luhansk region.
figure 4

The map represents Emerald Network sites with IDs UA0000069, UA0000315, UA0000209, UA0000078, and the southern portion of UA000031377. The left images show Landsat images for 1996 to 2013 and Sentinel-2 images for 2020 in false color with short wave infrared 1 (~1.6 μm), near-infrared (~0.8 μm), and blue (~0.4 μm) band combinations. The right images show the stable, gained, and lost tree cover for respective years and time periods. Areas and changes represent tree cover changes and net gains for territories under Ukraine and Russia’s control, respectively.

Full size image

Our analysis indicates that Ukraine achieved a total reforestation area of 11.17 ± 1.45 th. ha before the beginning of the conflict and 17 ± 1.74 th. ha from 2000 to 2020 on the Ukraine-controlled territories. At the same time, deforestation rates of territories under Russian control with similar bio-physical characteristics36 and the same war-related vulnerability factors indicate that the separation of ecosystems from environmental-protection institutions and policies through the occupation of territory led to dramatic degradation of the environment and loss of ecosystem sustainability.

The shift from deforestation trends during 1996–2000 to reforestation between 2000 and 2020 (in Ukraine-controlled territories) can be attributed to amendments and ongoing enhancements in national legislation related to forest management. Ukraine’s primary legal framework governing the forestry sector is the “Forestry Codex of Ukraine,” which was adopted in 199471. Although Ukraine joined the Bern Convention in 1996, the country initiated legislative amendments to align with international environmental standards and EU policies only in 2000. From 2000 to 2013, there was a consistent evolution of national legislation addressing the legal and financial aspects of forest use in Ukraine. During this period, the “Ukrainian Forests” state program72 was introduced, outlining Ukraine’s forestry development strategy and the establishment of the Emerald Network as part of the “National Program for the Formation of the National Ecological Network of Ukraine for 2000–2015”21. The new forestry strategy in Ukraine encompasses various objectives, including reforestation, enhancing forest productivity, quality, environmental conservation capabilities, biodiversity preservation, sustainability, climate resilience, the implementation of sound forest management practices, forest protection, and the enhancement of state oversight72. These principles were reflected in the further changes of forestry legislation71:

  • Implementation of forest tickets for logging control (article 69 with changes in 2009 and 2012);

  • Conservation of endangered tree species listed in the Red Book (article 70 from 2012), seeding trees and trees used for Red Book species nesting (article 70, from 2017);

  • Compulsory reforestation of forest areas damaged by logging, fires, storms, or drought within one year of the incidents (article 80 after 2012);

  • Mandate for forest owners and users to implement fire prevention and other measures for the protection and preservation of forests (article 86 from 2012);

  • Regulation of forest protection, restoration, utilization, and the role of aerial operations (article 86 and article 94 from 2012);

  • Halt of forest-hazardous operations by enterprises (article 109 from 2012);

  • Establishing allowable limits for timber harvesting (article 74 from 2012);

  • Rules and limitations for altering land use in forested areas (articles 57 and 59 from 2014);

  • Definition of natural forests (article 1 from 2017);

  • Supervision of the environmental impact caused by enterprises (article 29 from 2017);

  • Discontinuation of a forest user’s operations due to environmental impact (article 292 from 2017);

  • Ban on all forms of logging in natural forests. (article 70, 79, 84 from 2017);

  • Identification, protection, and integration of natural forests into the ecological network have become integral aspects of forest management (article 46 from 2017).

Between 2000 and 2020, the “Forestry Codex of Ukraine” underwent a total of 110 amendments, with 55 of them being enacted before 2014. The reforestation of regions within the Emerald Network was notably influenced by the implementation of these new amendments and forest policies, along with adherence to forest legislation. In the Ukraine-controlled areas, most of the restored forest was the result of deliberate planting as part of site restoration and development plans. Conversely, in areas under Russian control, there was a lack of forest management activities, and there was discouragement in the implementation of chapters 14, 15, and 16 of the “Forestry Codex of Ukraine,”71 which are crucial for forest protection and restoration. This contravenes Ukrainian forestry legislation, which mandates that forest users or responsible administrations must restore lost forest areas with the same tree types within one year of deforestation events. This discrepancy helps to explain the considerable deforestation in areas of the Emerald Network under Russian control.

Methods

Land-cover classification

Land-cover mapping was conducted in the Google Earth Engine platform using Landsat and Sentinel-1/2 satellite images and a supervised classification approach73. For this purpose, we generated collections of cloud-free month composites based on four data collections (Supplementary Table 4) for the study area and ground reference samples that covered main land-cover classes: tree cover, grassland, cropland, water bodies, artificial objects (impervious surfaces), and bare land. Taking into account the geospatial variations in the imagery used for compositing and the constraints posed by GEE, we created separate training data collections comprising 5000 points for each year. It was not possible to acquire ground-based reference data for the study region because of warfare activities and/or the absence of historical data. Therefore, the reference data were derived through photointerpretation by visual analysis of satellite data from Sentinel-2 and Landsat missions as well as historical very high spatial resolution imagery available through the Google Earth platform. For labeling, we used all available images for each target year which allowed us to accurately identify each land-cover class. Due to the differences in the numbers of available images and properties of data collection, to reach optimal accuracy and considering limitations of GEE in terms of resources and the model’s complexity to classify each pixel of the map represented as a vector of multi-spectral features in the time-series of composites of satellite images we used a random forest algorithm74 with 10 decision trees for 1996 and 2000, 100 for 2013, and 200 for 2020. The resulting maps were delivered in the Albers Equal Area projection at 30 m spatial resolution. For validation purposes, we independently generated and labeled 1,200 random points over the territory of the interest that was never used for model training.

Area estimation

Area estimation of land-cover changes directly from maps (pixel counting) is a biased estimator due to the problems related to the pixel’s resolution limitations, mixed-pixels effect, pixel shifts caused by the image misregistration problems, and classification errors. Thus, we used a stratified random sampling approach68, where samples were represented by a 30 × 30 m pixel. The number of samples Ni for strata i of stratification set S that targets overall accuracy can be estimated with the use of precision pi obtained by validation with the use of generated random points and weight of strata (proportion of strata area and total area) wi by formula68,75:

$${N}_{i}={w}_{i}{\left(\frac{\mathop{\sum}\limits_{j\in S}{w}_{j}\sqrt{{p}_{j}(1-{p}_{j})}}{0.01}\right)}^{2}$$
(1)

Our stratification was designed based on changes in the four maps and represented three types of temporal behaviors for the forest: stable forest, forest with changes (at least one per 4 years), and stable non-forest. We used the strata buffering technique to mitigate the emission error effect on the land-cover change area76. Area estimations required 876 samples for territory of interest, however considering the high level of segmentation of forest in the region we decided to generate individual sample sets for each Emerald Network site to reduce uncertainties. In total, we generated 2327 points for territories under Russian control and 2665 for territories under Ukraine control (Supplementary Table 5). Based on this stratification we estimated areas A of each change type k of C (tree cover gain, loss, stable, or stable not forest) for each time period by formula68,75:

$${A}_{k}={A}_{{{{{{\rm{t}}}}}}}* \mathop{\sum}\limits_{i\in S}\frac{{n}_{i,k}* {w}_{i}}{\mathop{\sum}\limits_{j\in C}{n}_{i,j}}$$
(2)

where ni,k and ni,j represent the number of samples of each i strata class that correspond to k or j on the map and reference. The error Ek was estimated based on the total area of the study area (At)68,75:

$${E}_{k}={A}_{{{{{{\rm{t}}}}}}}* \sqrt{\mathop{\sum}\limits_{i\in S}{w}_{i}^{2}* \left(\frac{{n}_{i,k}* {w}_{i}}{\mathop{\sum}\limits_{j\in C}{n}_{i,j}}\right)* \left(1-\frac{{n}_{i,k}* {w}_{i}}{\mathop{\sum}\limits_{j\in C}{n}_{i,j}}\right)}$$
(3)