The loss, degradation and fragmentation of habitats represent major threats to terrestrial mammal diversity around the globe (Rands et al. 2010; Crooks et al. 2017; Kuipers et al. 2021). In recent decades, the impacts of roads on ecosystems have received concerted attention (Coffin 2007) and roads have been recognised as a main driving force behind the global alteration of natural habitats (Forman and Alexander 1998; Spellerberg 1998; Trombulak and Frissell 2000; Rhodes et al. 2014). Roads can affect many components of ecosystems and their associated edge effects can manifest at local and landscape levels (Coffin 2007). Animal species with wide ranging movements, large home ranges and long dispersal distances are especially vulnerable to roads (Rytwinski and Fahrig 2012). In particular, possible adverse effects of roads have been documented for large carnivores, including canids (Jȩdrzejewski et al. 2004; Riley et al. 2006), felids (Palma et al. 1999; Kerley et al. 2002; Niedziałkowska et al. 2006; Colchero et al. 2011; Litvaitis et al. 2015) and ursids (Proctor et al. 2018; Morales-González et al. 2020).

The relationship between roads and brown bears (Ursus arctos) is complex, because road effects can be area- and/or sex-specific, may vary by time of day and season and can be influenced by traffic volume (Penteriani et al. 2018). Roads facilitate access of people to bear habitats, increasing the chances of human-bear encounters and bear mortality risk (Benn and Herrero 2002; Ciarniello et al. 2009; McLellan 2015). In some areas, collisions with vehicles represent a major cause of documented bear mortalities (Huber et al. 1998; Kaczensky et al. 2003; Gunther et al. 2004). Certain components of roadside vegetation, especially during spring and early summer (Nielsen et al. 2004a; Roever et al. 2008a), as well as other food sources associated with human presence on roads, such as waste (Huber et al. 1998), can lure bears close or onto roads. While seasonally attractive roadside vegetation can potentially improve female body condition and reproductive success, the benefits of roadsides are countered by high mortality (Boulanger et al. 2013). Road placement, for example, in areas of low ruggedness, can combine with previously enumerated factors to further increase the attraction of roads (Roever et al. 2008a).

Some bears decrease their use of areas near roads or avoid these altogether, suggesting that roads can cause effective habitat loss at varying scales (McLellan and Shackleton 1988; Kasworm and Manley 1990; Mace et al. 1996; Waller and Servheen 2005). Displacement from habitats near roads reduces habitat extent and might affect body condition, reproductive rates and, ultimately, population density of bears (McLellan and Shackleton 1988; Mace et al. 1996). Adult males generally avoid roadsides (McLellan and Shackleton 1988; Boulanger and Stenhouse 2014). Females with cubs (McLellan and Shackleton 1988; Graham et al. 2010) and subadults (Mueller et al. 2004; Graham et al. 2010) tend to exploit the vicinity of roads more often, due to preferred forage availability (e.g. herbaceous vegetation layer for grazing) and/or as an avoidance mechanism against potentially aggressive/infanticidal adult males.

Traffic volumes are negatively correlated with road permeability for bears (Gibeau 2000; Waller and Servheen 2005; Northrup et al. 2012). High traffic levels (e.g. along highways) can create home range boundaries for resident animals (Kaczensky et al. 2003; Find’o et al. 2019). Once traffic volumes exceed a threshold of 5,000 vehicles/24 hrs, roads may become absolute barriers to bear movements (Skuban et al. 2017). Roads can also offset the social structure of bear populations, because females are less likely to cross busy roads than males (Gibeau and Heuer 1996; Waller and Servheen 2005) and stop crossing roads altogether at a lower traffic threshold than males (4,000 vs. 5,000 vehicles/24 hrs; Skuban et al. 2017).

North American studies advocate for limiting road access (Mace et al. 1996; Wielgus et al. 2002; Graves et al. 2006; Roever et al. 2008a, 2008b) and reduction of road density in bear habitat by targeted road closure and removal (Nielsen et al. 2006; Ciarniello et al. 2007; Nielsen et al. 2008; Switalski and Nelson 2011). In Europe, on the other hand, bears mostly have to contend with crowded, highly fragmented, multi-use landscapes, with little wilderness areas left (Swenson et al. 2000; van Maanen et al. 2006; Linnell et al. 2008), where they are frequently exposed to roads (Torres et al. 2016; Psaralexi et al. 2017).

Romania is an important stronghold for brown bears in Europe, hosting approximately 6,000 individuals (Swenson et al. 2000; van Maanen et al. 2006; Linnell et al. 2008; Kaczensky et al. 2013), although this number might be overestimated (Salvatori et al. 2002; Popescu et al. 2016). As a European Union (EU) Member State since 2007, Romania has plans to extend and modernise its transport infrastructure to meet EU standards, with the aid of both national and dedicated EU funding (Romanian Ministry of Transport 2008). The goal is to cope with steadily increasing traffic levels: in the period 2007–2019, the number of vehicles has almost doubled, reaching more than 8 million in 2019 (Eurostat 2021). In 2020, the total length of Romanian highways was 904 km, with a highway density of 3.8 km/1,000 km². Major transport infrastructure developments are envisioned to enlarge this network to a total of 2,416 km of highways and 1,784 km of express roads (Papp et al. 2022). The country’s best bear habitats are in the Carpathian Mountains and their foothills (Swenson et al. 2000; van Maanen et al. 2006; Cristescu et al. 2019), with many of the planned highways intersecting bear habitat. As a result, there is a potential risk that, without proper road mitigation measures in place, the Romanian bear population and its habitats will become severely fragmented.

The planned A8 (Tîrgu Mureș–Iași–Ungheni) highway, linking the city of Tîrgu Mureș in the west to the national border between Romania and the Republic of Moldova in the east, has been identified as a major threat to brown bear habitat connectivity (Fedorca et al. 2019). In particular, its westernmost section will intersect both important bear denning habitats (Faure et al. 2020) and critical movement corridors linking denning habitats to seasonal feeding grounds, as indicated by telemetry data from bears fitted with GPS collars (Domokos, unpublished data). The goals of this study were to evaluate brown bear occurrence, habitat use and genetic connectivity along a central section of the planned A8 highway. We aimed to identify locations on the landscape that are conducive to 1) bear occurrence, 2) bear movement and 3) to estimate the minimum number and sex ratio of bears using the planned highway route prior to the highway’s construction. Our overarching hypothesis was that the distribution of the brown bear population would be ubiquitous and relatively homogeneous within a landscape that maintains some permeability despite human settlements, given that highway construction had not started at the time of the study. We anticipated that specific landscape characteristics might influence local patterns of bear occurrence.

Overall, 68 hair-trap stations were active for 17 consecutive days across all sessions. Sampling effort was 1,156 trapping days in 2014 and 2017, respectively 1,207 trapping days in 2020. Three additional hair-trap stations active only in 2020 were excluded from modelling bear occurrence, but included in all other analyses. Three other locations that had been planned for sampling were excluded due to the presence of shepherd camps or livestock water troughs in all survey years.

Using non-invasive repeat survey methodology, we assessed the distribution and documented the minimum local population size of brown bears along a planned highway route in Romania, as part of an effort to collect information before highway construction. We detected bears at 21 sampling stations along the planned highway route, but with a more restricted distribution than expected and a concentration of presence in the western part of the study area. Our study did not succeed in producing direct evidence of bears crossing the planned highway route (e.g. same individual detected by hair traps on both sides of the future highway). Nevertheless, we provide genetic evidence that the population uses both sides of the planned development, including the detection of related animals on both sides of the highway route.

We found a positive association between bear occurrence and terrain ruggedness. When confronted with human disturbance, such as in human-dominated landscapes of Europe, bears may select rugged terrain (Martin et al. 2010; Dorresteijn et al. 2014; Roellig et al. 2014). Rugged terrain limits human access and provides secure habitat, minimising the risk of human-bear encounters and of human-induced bear mortalities (Nielsen et al. 2004b). With decreasing distance to human settlements, the use of increasingly rugged terrain has also been documented in the case of denning bears (Sahlén et al. 2011).

Most hair traps that detected bears were close (< 1 km) to human settlements (mean ± SD distance of all hair traps to human settlements was 1.35 ± 1.33 km). Mountainous villages in Romania commonly comprise solitary houses or small groups of homesteads, which are often not recorded as part of settlements in Corine Land Cover (i.e. discontinuous urban fabric). Thus, some hair traps that registered bears were even closer to buildings than revealed by the land-cover layer. Our results are in accordance with previous studies reporting that in Romania bears regularly use human-dominated landscapes and in general habitats in the proximity of human settlements (Dorresteijn et al. 2014; Roellig et al. 2014; Borka-Vitális et al. 2017), without necessarily coming into conflict with humans. After investigating the mechanisms underlying the occurrence of bears near settlements, Elfström et al. (2014) concluded that bears approaching settlements display a natural behaviour, best explained through avoidance of intraspecific aggression and/or interference competition. This adaptive behaviour is shaped by the despotic distribution of conspecifics more than by naivety, food conditioning or human habituation. In despotic distribution, dominant individuals exploit high quality habitats more often than subordinate conspecifics, whereas subordinate bears seem to fear dominant conspecifics more than they fear people. We can confirm that during the extended periods of fieldwork in the area, despite frequent interactions with locals, we have never heard complaints about habituated/nuisance bears, although this can be an issue elsewhere (Cristescu et al. 2016). When close to human settlements or human activity, bears may adjust their behaviour to avoid encounters with people (Ordiz et al. 2011), being most active at crepuscular or nocturnal hours to avoid overlap with human diel activity patterns (Kaczensky et al. 2006; Martin et al. 2010; Schwartz et al. 2010; Ordiz et al. 2014; Oberosler et al. 2017). Habitat selection may also vary with time of day and season according to risks associated with people, with bears near settlements selecting steep slopes and highly concealed resting sites during daylight hours (Martin et al. 2010; Ordiz et al. 2011; Cristescu et al. 2013; Skuban et al. 2018).

Longitude of the hair trap location was a good predictor of bear presence, with westernmost hair traps more successful. One possible explanation for this pattern is habitat fragmentation of the region in the west–east direction by Lake Bicaz due to its large size, as well as numerous contiguous settlements around it. However, poaching with firearms is also an issue of concern around the Lake and in the region east of it (Anonymous, Harghita County Police Inspectorate, Miercurea Ciuc, Romania, personal communication 2015, 2016). As these are some of the best bear habitats in Romania (Pop et al. 2018; Cristescu et al. 2019), widespread poaching can transform them into ecological traps for bears (e.g. attractive habitats with high mortality risk; Schlaepfer et al. 2002), also affecting bears originating from other source areas (Robertson and Hutto 2006; Lamb et al. 2017). While it is possible that some bear individuals may avoid areas with high poaching risk as an evolutionarily adaptive response to fear of humans (Ordiz et al. 2013), in general, poaching is an activity that may be difficult to predict and adapt to and could impact bear populations substantially (Kaczensky et al. 2011).

The habitat types in which the hair traps were mounted did not influence the success rate of collecting bear hair samples. In Romania, during summer when our surveys were conducted, female bears typically select mixed forests, whereas males select all three forest types: deciduous, mixed and conifer (Pop et al. 2018). Pastures are important feeding grounds for brown bears during the same period mainly because of the availability of ants (Dorresteijn et al. 2014), an important food source for the species (Swenson et al. 1999; Große et al. 2003; Roellig et al. 2014).

Although we identified 24 distinct bear individuals in the three surveys, we expected to detect a larger number of individuals. A possible reason is the close proximity of sampling stations to human settlements, which can act as a filter for the bear population, selecting for subordinate individuals or demographics of age or reproductive classes that are more tolerant towards human presence and/or actively avoid larger/more aggressive conspecifics. Nellemann et al. (2007) found that 52% of bears in the wider surroundings of settlements in Sweden were subadults of both sexes, with only 8% of adult males present in the < 10 km radius of larger settlements and resorts. While the techniques used in this study did not allow us to differentiate between age classes of bears that we sampled, we know that at least some of the detected bears were adults, as confirmed by four documented PO relationships.

Proctor et al. (2012) demonstrated that mortality associated with settlements has been a major force impacting bear populations and connectivity in western Canada, the northern United States and southeast Alaska. In our study system, we documented gene flow through PO, FS and HS relationships, both on the north–south and west–east axis and across the existing network of settlements. On the west–east axis (e.g. along the planned highway route), with the exception of some human settlement barriers near Lake Bicaz, bear movements are mostly unobstructed by human habitation, at least parallel with the planned highway route. However, because of often contiguous settlements spread along valleys stretching from west to east, bear movements on the north–south axis (e.g. across the planned highway route) are likely already limited and possibly restricted to the remaining undeveloped areas. Tunnels, viaducts and bridges that are planned for highway development could help maintain some of the remaining functional connectivity for bears and other wildlife, especially because these structures will be relatively close to each other (mean distance between structures 326.2 m (range 20–2,072 m); Silvia Borlea, EPC Environmental Consulting, Bucharest, Romania, personal communication 2023). The mean length of the planned tunnels is 188 m (range 16–940 m), while the mean Openness Index (width × height / length of the structure) of the selected viaducts and bridges is 120.6 (range 4.4–853.9). The most significant linkage areas on the north–south axis are situated between villages, such as Ditrău and Hagota (12 km), Hagota and Recea (4 km) and Petru Vodă and Pluton (4.8 km). The latter area, however, is situated east of Lake Bicaz and our study did not document bear presence in its surroundings.

Widespread, cryptic poaching could have contributed to relatively low bear detection rates in our study. Due to low densities and slow reproductive rates, large carnivores are especially vulnerable to poaching and previous studies have documented substantial effects of illegal killings on large carnivore demography (Kruckenhauser et al. 2009; Liberg et al. 2012; Persson et al. 2015; Červený et al. 2019; Benson et al. 2023). Additionally, the timing of our surveys might have also influenced bear detection rates, with part of the bear population moving during the summer to richer feeding grounds situated at lower altitudes, either to the west or to the east from the surveyed area. This pattern of significant seasonal movements of at least part of the population has been observed in another area of the Romanian Eastern Carpathians (Domokos, unpublished data) and other regions (Cozzi et al. 2016; De Angelis et al. 2021). The timing of our surveys might have influenced bear detection rates in other ways too. While our surveys took place during summer, bears are more likely to respond to scent lures in spring (Gervasi et al. 2008). Lamb et al. (2016) found that starting hair trapping at lure-scented sites towards the end of the mating season (which corresponds to early June in Romania) maximises female detections, while starting early in the mating season (late April - early May in Romania) maximises male detections. Another potential limitation of our survey design could have been the fact that scent lures do not offer a reward to the visiting bear, which might thus become trap-wise and lose interest in revisiting the site or visiting other hair-trap locations.

Even if we were unable to determine the sex of three of the 24 individual bears we identified, the data are indicative of a large population segment of females (1:1.3 [male:female]). This is comparable to the 1:1.6 sex ratio estimated for the Romanian Southern Carpathians (Skrbinšek et al. 2019) or to the 1:1.5 documented in Slovenia and 1:1.4 in Croatia (Skrbinšek et al. 2017). Prior to a ban introduced in October 2016, bear trophy hunting was a common, decades-old practice in Romania (Salvatori et al. 2002; Popescu et al. 2019). With male-biased hunting, a sex ratio skewed in favour of females is to be expected.

We showed that domestic dogs are present throughout the region, at least during the summer. The frequent detection of dogs at the hair traps is likely due to the presence of large numbers of guardian dogs accompanying livestock, stray dogs or dogs associated with human settlements. A similar finding was recorded in a study of wolf diet which revealed the importance of dogs in the diet of wolves in the south-eastern Carpathian Mountains (Sin et al. 2019). Although their benefits for protecting livestock from carnivore attacks have been demonstrated (Smith et al. 2000; van Eeden et al. 2018), domestic dogs can have negative impacts on wildlife when not under human supervision (Potgieter et al. 2016; Wierzbowska et al. 2016; Drouilly et al. 2020). If dogs as a depredation mitigation strategy are adequately applied, they can be an effective strategy for livestock protection (van Eeden et al. 2018), but may enable disease transmission at the wildlife-domestic animal interface (Borka-Vitális et al. 2017). Bears are susceptible to a number of pathogens of domestic dogs, such as canine distemper virus (CDV, the etiological agent of distemper; Di Francesco et al. 2015; Vitásková et al. 2019; Balseiro et al. 2024), canine parvovirus type 2 (CPV-2; Di Francesco et al. 2015; Vitásková et al. 2019) and canine adenovirus type-1 (CAdV-1, the etiological agent of infectious canine hepatitis; García Marín et al. 2018; Balseiro et al. 2024). Measures to decrease the risk of disease transmission between domestic dogs and bears and between livestock and wildlife in general should be incorporated into decision-making programmes for livestock husbandry under free-ranging conditions.

We also genetically confirmed the presence of unidentified Canis sp. (either dogs or wolves) in several locations. There is a possibility that at least some of these samples originated from wolves, in particular, the ones identified as haplotype w4, which occasionally occurs in dogs, but is commonly found in Romanian wolves (Jarausch et al. 2023). We did document wolf scats and a partially consumed livestock guardian dog in the area in spring 2015. However, our evidence indicate that most unidentified Canis sp. samples originated from dogs. Firstly, we confirmed the presence of dogs (either through visual inspection or genetic analysis of hair samples) at the majority of locations where unidentified Canis sp. samples were collected. Secondly, we collected samples later confirmed as originating from unidentified Canis sp. due to their resemblance to bear hair. These samples consisted of long, dark-coloured, undulating, soft guard hairs, which are also characteristic of dark-coloured, large bodied, mixed breed livestock guardian dogs. Dark-coloured dogs are traditionally used in the area, although less commonly than light-coloured animals, as shepherds appear to gradually replace mixed breed dogs with purpose-bred shepherd breeds such as Caucasian, Central Asian or Anatolian.

Connectivity on the north-south axis is relatively limited in the study area due to existing human settlements. Completion of the A8 highway could potentially further impede bear movements in important bear habitats centrally located in the Romanian Eastern Carpathians and their foothills. This area provides a vital link to other national-level populations located further North, including Ukraine, Slovakia and Poland (Straka et al. 2012). Together, these national populations form the vast majority of the transboundary Carpathian bear population, completed by a small population in eastern Serbia (Kaczensky et al. 2013; Chapron et al. 2014). Preserving and enhancing functional connectivity within the Carpathian bear population (Matosiuk et al. 2019; Papp et al. 2022), including maintaining permeability of Romania’s Eastern Carpathians is of crucial importance (Fedorca et al. 2019). In this respect, dedicated wildlife crossing structures could have been planned by the responsible authorities during the pre-construction phase of the highway, based on the best available information concerning bear presence and movement in the area. Given the decisions already made through the environmental permit, we recommend permeability studies post-completion of the A8 highway section, with a particular focus in the area of the potential crossing structures associated with highway development as imposed by topography; and after bears have had the time to explore and start using them. If the structural features of the highway meant to bypass topographical challenges prove insufficient for wildlife connectivity, dedicated wildlife crossing structures (e.g. wide overpasses, Ford et al. 2017) might be required, even though their construction costs would be much higher at that stage. To maximise their effectiveness, these should be located in the vicinity of the still undeveloped areas identified in this study. Permanent development should be limited as movement is somewhat constricted already by the high density of human settlements, although for the time being i.e. before highway construction bear population connectivity is not yet fully curtailed. In parallel with maintaining habitat connectivity, the issues of poaching and dogs in the wild should be addressed by wildlife managers and law enforcement authorities.