The Influence of Infrastructure on the Breeding Distribution of a Threatened Top Predator

Abstract

The eastern imperial eagle (Aquila heliaca) has shown a marked population increase in the past decades in Hungary. The breeding range is expanding towards homogeneous agricultural habitats of the Hungarian Plain, where the already existing and recently growing infrastructural network is thought to be one of the main factors limiting distribution. We used data from 508 breeding attempts between 1989 and 2008 to assess the effects of infrastructural networks on breeding distribution. We constructed a single cumulative infrastructure effect (CIE) variable based on the avoidance of different infrastructure types by eagles in the past 20 years. Conditional autoregressive models were built in a Bayesian framework to quantify the effects of infrastructures on the spatial breeding pattern in a pre-defined core study area. Both multivariate and CIE models were able to classify the presence of breeding attempts with high accuracy. The CIE variable was used to build a predictive distribution model for the Hungarian Plain. The results suggest that infrastructure has a significant local effect but does not necessarily hinder the future range expansion of imperial eagles, as two-thirds of the prediction area seems to be suitable for the species. The methods and results described enable conservation managers and policy makers to assess the trade-off between infrastructural development and nature conservation priorities.

1. Introduction

The expansion of settlements and infrastructures has an overall direct negative impact on biodiversity through the destruction, degradation, and fragmentation of natural habitats. Indirectly, the impact is through increased human presence, causing disturbance or persecution of wildlife. The growth of built-in areas has been identified as a key threat to many vulnerable bird species in most parts of the world, including highly populated areas in Europe [1,2]. More than a quarter of the European Union’s (EU) territory has now been directly affected by urban land use, and further intensive developments are foreseen, especially in the most recently accessed member states of Central and Eastern Europe [3,4]. Besides urban sprawl, the development of the Trans-European Transport (TEN-T) and Energy (TEN-E) Networks is also fragmenting the natural habitats of Europe [5]. The EU funds causing socio-economic and land use changes in the most recently joined member states could have a strong impact on Europe’s biodiversity, since the natural and semi-natural habitats of these countries still hold significant populations of several threatened species, which have already disappeared or become rare in the Western European part of their former range [6]. To compromise nature conservation and economic claims, decision-makers need precise site- and species-specific empirical information, but this information is rarely available comprehensively. For the conservation of the most threatened species, studies should locate important habitat patches and find out what density of or distance from infrastructures significantly affects their distribution, survival, and reproduction.

Habitat selection models of bird populations frequently incorporate infrastructural factors (e.g., settlements, roads, railways, and power line networks) to investigate their effects on species distribution [7,8,9,10]. The number and quality of such habitat selection studies have increased, mainly due to the evolution of geographical information systems (GISs) and powerful statistical methods [11,12,13]. Such studies typically analyze habitat variables measured in occupied (i.e., breeding or foraging habitats) and unoccupied sites [14]. Unoccupied sample sites might be selected at random, which contain a representative proportion of unoccupied areas but are theoretically suitable parts of the study area [15,16]. Alternatively, sample sites can be occupied and unoccupied quadrats of a grid system, and the methodology enables habitat models to analyze the distribution of species and habitat variables for the entire study area [17,18].

Raptor species are widely used in habitat selection studies because (1) they usually breed at low densities; therefore, large spatial scale habitat preferences can be studied within one population [19,20]; (2) a large proportion of raptor species are classified under an unfavorable conservation status; therefore, exact information on habitat requirements is needed for planning effective conservation studies [21,22]; and (3) raptors, as spectacular top predators, are frequently used flagship or umbrella species in conservation policy [23], and their presence locally could potentially detect areas with relatively high levels of biodiversity [24].

The eastern imperial eagle (Aquila heliaca) is a large-sized raptor, distributed along the Eurasian forest-steppe habitats from Central Europe to Dauria, Russia. Despite its wide range, the breeding areas are scattered, and the world population is less than 10,000 breeding pairs [25], with the great majority (>90%) located in Russia and Kazakhstan [26]. The species’ conservation status is treated as “vulnerable” globally [25], and it has disappeared from large parts of its former distribution area in Europe during the 19th and 20th centuries. Recently, only two small isolated populations have been found in the territory of the EU, namely, in Bulgaria (40–45 pairs) and in the Pannonian ecoregion in Hungary, Slovakia, Austria, the Czech Republic, Romania, and Serbia (356–381 pairs) [26,27,28].

The largest long-term surveyed European population of the eastern imperial eagle inhabits the Hungarian Plain and the surrounding hilly areas [29,30,31,32]. Intensive conservation efforts significantly decreased the level of raptor persecution (especially poisoning and shooting) in Hungary; therefore, the national population of imperial eagles has continuously increased since the 1980s [30]. Long-term population monitoring has assumed that this large-sized raptor species is highly sensitive to human disturbance and habitat alteration and that most infrastructural factors could directly or indirectly (by increasing disturbance and persecution) affect the mortality and/or breeding success of imperial eagles. Population monitoring has shown that human-related mortality represented 86% of the known causes of death in the Hungarian population (n = 72, 2001–2009), including electrocution (28%), persecution (poisoning, 40%; shooting, 6%), and collision with vehicles (11%) and electric wires (1%) [31]. Human disturbance (26%) and persecution (8%) were also among the most significant identified causes of recorded breeding failures (n = 64, 2003–2009) [31]. Poisoning, which poses a major conservation threat to raptors across Europe [33,34], particularly in Hungary [35,36], is only indirectly influenced by infrastructure, primarily through increased accessibility to remote habitats. In contrast, the risk of electrocution is directly associated with the distribution of medium-voltage power lines and remains one of the leading causes of mortality for imperial eagles and several other threatened bird species in Hungary [37].

In this study, our objective was to assess the impact of settlement, road, railway, and power line networks (hereafter collectively referred to as “infrastructure”) on the breeding distribution of imperial eagles in the Hungarian Plain. We hypothesized that different types of infrastructure influence the spatial distribution of breeding sites in this Central European population. Given the recent south-eastward expansion of this population, we also developed a predictive model for other parts of the Hungarian Plain to identify areas where existing infrastructure may impede colonization and potential undisturbed habitats where more intensive field surveys are needed to confirm the presence or absence of the species.

2. Materials and Methods

2.1. Study Area

The effects of infrastructure on nest site selection were modeled in two study areas in two consecutive steps. First, we developed the model using data from the “core study area”, where the breeding distribution of imperial eagles became coherent and reached the highest density in Hungary between 1989 and 2008 [31]. In the second step, we applied this model to a larger region, the “model prediction area”, to estimate potentially suitable nest sites. This broader area has been colonized since 2001 and remains incompletely saturated as of 2024, with a more fragmented distribution [32].

The core study area encompasses a 2800 km² northwestern section of the Hungarian Plain, specifically the Jászság and Heves-Borsod Plains (47°23′–47°50′ N, 19°46′–20°54′ E) (Figure 1). This lowland region (elevation 90–120 m a.s.l.) is bounded by the Tisza and Zagyva rivers to the south and the foothills of the Mátra and Bükk mountains to the north. The landscape is dominated by intensively cultivated agricultural fields interspersed with shelterbelts composed of poplar (Populus sp.) and black locust (Robinia pseudoacacia) plantations, along with occasional remnants of semi-natural grassland and oak (Quercus robur) forest fragments. In this area, imperial eagles nest almost exclusively in trees, with only a single documented exception of a nest built on a high-voltage transmission pylon [38].

The model prediction area comprises the eastern part of the Hungarian Plain, specifically the lowland areas (78–150 m a.s.l.) of the Tisza River drainage basin. Covering 34,400 km² (46°07′–48°32′ N, 19°30′–22°54′ E), it is bordered by foothills to the north, the national border to the east and south, and the Kiskunság Sand Ridge to the west. Forested areas in South-Nyírség were excluded, resulting in a large coherent open region dominated by agricultural fields, similar to the core study area. Suitable nesting trees (i.e., Populus, Robinia, or Quercus trees taller than 8 meters) and three primary prey species (Lepus europaeus, Cricetus cricetus, and Phasianus colchicus) are widely distributed throughout the region [39,40,41,42]. Therefore, for the purposes of this large-scale spatial analysis, we considered the entire model prediction area as a potential breeding habitat for the imperial eagle.

2.2. Population Survey

The entire core study area was systematically surveyed for active imperial eagle nests by members of the Hungarian Imperial Eagle Working Group at the start of each breeding season (February–March) between 1989 and 2008 [31]. The sedentary behavior of the species, frequent display flights and vocalizations, and typically conspicuous nest sites greatly facilitated the detection of breeding pairs. As a result, we consider the distribution of breeding imperial eagles within the core study area to be reliably documented.

The breeding population increased steadily over the 20-year period, from just two breeding pairs in 1989 to 37 by 2008. By the end of the study period, this area supported the densest known subpopulation in Central Europe, accounting for 42% of Hungary’s and 25% of the European Union’s eastern imperial eagle population.

Nest locations were initially recorded on 1:25,000-scale topographic maps (Hungarian Army Cartographic Service) and later georeferenced via GPS. For breeding attempts prior to 2002, GPS coordinates were retroactively assigned based on the original map data. In total, 319 nesting attempts across 43 different territories within the core study area were recorded in a GIS database. Additionally, 189 breeding attempts in 38 territories located in more sparsely populated areas of the Hungarian Plain were monitored during the same period. These supplementary data were used to evaluate the predictive accuracy of the model. By 2008, the overall study area sustained 64 breeding pairs, representing 73% of the national population.

2.3. Selection of Spatial Scales and Measurement of Habitat Variables

The average nearest neighbor distance (NND) between active imperial eagle nests in the core study area was 4.512 ± 1.405 km (n = 37) in 2008. This metric is commonly used in raptor habitat selection studies to define the appropriate spatial scale for territory-level analyses [10,43]. Based on this, we selected a spatial resolution smaller than the average territory size, yet large enough to capture broad-scale habitat selection patterns while minimizing the influence of local microhabitat variation. A 2.5 km × 2.5 km UTM grid system, which is widely applied in national bird monitoring and census programs in Hungary [44], was deemed appropriate for analyzing the relationship between infrastructure density and the presence or absence of eagles.

For each grid cell, we calculated the density of various types of infrastructure, including the proportion of land covered by settlements (km²/km²) and the length of linear features such as roads, railways, and power lines (km/km²). In multivariate analyses, a key challenge is balancing model complexity (i.e., ensuring that the model captures essential patterns) against the need for interpretability and the risk of overparameterization [45]. To address this, we developed a single composite explanatory variable: the “cumulative infrastructural effect” (CIE). This variable represents the percentage of each grid cell covered by areas lying closer to a given type of infrastructure than the minimum distance recorded between that infrastructure type and any occupied nest site within the core study area (1989–2008). Because the observed minimum avoidance distances varied widely among infrastructure types (0–700 m, see

Table 1. Effects of different infrastructures on mortality, disturbance, and nesting locations of imperial eagles and their observed density per km² (mean ± SD) in the occupied and unoccupied UTM cells (2.5 × 2.5 km² scale) of the core study area.

Type (Unit)	Mortality Risk	Disturbance Risk	Minimal Nesting Distance *	Buffer Coverage **	Density in Occupied (n = 121)	Density in Unoccupied (n = 327)
Settlement (km2)	high	high	700 m	20.74%	0.018 ± 0.050	0.071 ± 0.126
Isolated building (number)	medium	medium	0 m	-	0.284 ± 0.528	0.226 ± 0.354
Major paved road (km)	low	medium	100 m	0.75%	0.063 ± 0.147	0.084 ± 0.177
Minor paved road (km)	medium	high	250 m	8.39%	0.124 ± 0.190	0.205 ± 0.211
Railway (km)	medium	medium	0 m	-	0.055 ± 0.129	0.084 ± 0.186
Medium-voltage power line (km)	high	low	100 m	6.25%	0.219 ± 0.294	0.427 ± 0.425
High-voltage power line (km)	low	low	100 m	2.71%	0.137 ± 0.224	0.181 ± 0.315
Cumulative infrastructural effect (km2)	high	high	-	33.05%	0.197 ± 0.191	0.380 ± 0.274

* Values were rounded to 50 m precision, and a single nest, which was built on a high-voltage electric pylon, was excluded. ** The minimal nesting distance was applied for creating buffers around the infrastructures.

), the CIE variable not only reflects the overall density of infrastructure but also incorporates weighted avoidance distances, providing a biologically meaningful estimate of disturbance impact based on empirical observations.

Most infrastructure data, such as settlements, major and minor roads, and railways, were obtained from the National Geographical Information Database (OTAB©, GraphIT Ltd., Budapest, Hungary, version 2003). Detailed GIS layers for medium- and high-voltage power lines were provided in 2008 by local electric companies (ÉMÁSZ Nyrt., E.ON TITÁSZ Zrt., and MAVIR Zrt.). The spatial configuration of the listed infrastructural network remained largely unchanged between 1989 and 2008, with the exception of major roads. To account for this, we manually incorporated 125 km of new highway segments constructed between 2003 and 2008. As a result, the final GIS dataset represented infrastructure as of 2008. These developments affected only a limited number of grid cells and had no significant impact on the overall model outcomes. However, it should be noted that subsequent infrastructure development may exert localized effects on habitat suitability that were not captured in this study. All spatial analyses were conducted using ArcMap© software (ESRI Inc., Redlands, CA, USA, version 9.0) with Hawth’s Analysis Tools extension [46].

2.4. Statistical Analyses

We applied intrinsic conditional autoregressive (CAR) models within a Bayesian framework [47,48] to analyze the relationship between infrastructure variables and the presence or absence of imperial eagles in the core study area. Based on the sample dataset, prior assumptions regarding data distribution, the constructed likelihood model, and posterior distributions of the parameters were estimated using Markov Chain Monte Carlo methods. The resulting posterior distributions were summarized using descriptive statistics (mean, median, and quantiles) and 95% credible interval, which serve as the Bayesian analog of frequentist 95% confidence intervals.

To account for spatial autocorrelation, a global random spatial effect term (ρ_i) was incorporated into the linear predictor of the model:

$$\log \left({\displaystyle \frac{p_i}{1-p_i}}\right)={W_i}^{\prime }\beta +{\rho }_i+{\epsilon }_i$$

where p_i is the probability of the presence of imperial eagle nests in ith cell, ${W_i}^{\prime }$ is the matrix of descriptive environmental variables, $\beta$ is the matrix of associated coefficients, and ${\epsilon }_i$ and ${\rho }_i$ are the spatially uncorrelated and correlated spatial heterogeneity, respectively. The left side of the equation is the logit link function. Formally, the correlated spatial random term $\rho$ is defined by its conditional distribution given the values at neighboring locations as follows:

$${\rho }_i|{\rho }_j\approx N\left({\displaystyle \frac{{\displaystyle \sum _{j\in \delta i}}{w}_{ij}{\rho }_j}{w_{i+}}},{\displaystyle \frac{{\sigma }_{\rho }^2}{w_{i+}}}\right)$$

where $w_{i+}$ denotes the pre-defined total number of neighboring cells of the cell (i), and ${\sigma }_{\rho }^2$ is the conditional variance. Neighbors of any cell were defined as the eight adjacent cells. Uninformative prior settings were applied to all model parameters. Posterior distributions were sampled using three initial Markov Chains with 100,000 iterations and a burn-in value of 10,000.

We developed two separate models to assess the influence of infrastructure-related variables on the distribution of imperial eagles within the core study area. In the multivariate model, we included all seven individual infrastructure variables to estimate their relative importance. In the CIE model, we used only the previously defined CIE variable. A third model, the predictive model, was constructed to assess the effects of infrastructure on the potential distribution of imperial eagles across the eastern Hungarian Plain. This model applies the mean effect size of the CIE parameter, derived from the posterior distribution in the CIE model, to the CIE values calculated for each grid cell in the broader prediction area. Probabilities were obtained by calculating the inverse logit transformation for each cell:

$$p(y)={\displaystyle \frac{\exp \left(LP\right)}{1+\exp \left(LP\right)}}$$

where LP is the linear predictor [49].

Model fitting and the predictive power of both models were tested by calculating Cohen’s κ statistics [50] and the receiver operating characteristic (ROC) curve [51].

R 2.9.0 software [52] and Winbugs 1.4.3 were used for data analysis and modeling [53].

3. Results

3.1. Minimal Observed Distance of Nest Sites from Infrastructures

Our analysis revealed that imperial eagles consistently nest within defined buffer zones surrounding most types of infrastructure. These buffers, represented by the CIE variable, covered approximately one-third of both the core study area (33.04%) and the broader eastern Hungarian Plain (32.89%). However, nest sites were rarely located within these zones, accounting for less than 1% of all recorded nests. Because both the observed minimal distances and the spatial density of different infrastructure types varied substantially, each contributed differently to the overall CIE variable (Table 1). Among these, settlements and their associated buffers accounted for the largest proportion of coverage. Minor roads and both medium- and high-voltage power lines also affected substantial areas, whereas the sparser major road network had a more localized impact. In contrast, railways were excluded from the CIE variable because nests were found in seven instances located very close to their tracks (i.e., within 10 m), indicating no consistent avoidance pattern.

3.2. Infrastructural Effects on Present Distribution Within the Core Study Area

The posterior distributions of the multivariate and CIE models’ coefficients are presented in

Table 2. Distribution of coefficients of the two models built for the core study area.

Model/Random Node	Mean	SD	MC Error	2.5%	Median	97.5%
Multivariate model
Settlement	−8.694	2.724	0.08263	−14.5	−8.523	−3.786
Major road	−7.04 × 10−4	0.001104	1.95 × 10−5	−0.00293	−6.82 × 10−4	0.001414
Minor road	−0.00217	9.53 × 10−4	2.85 × 10−5	−0.00417	−0.00213	−4.27 × 10−4
Railway	−0.00142	0.001043	2.25 × 10−5	−0.00355	−0.00138	4.89 × 10−4
Medium-voltage power line	−8.66 × 10−4	6.91 × 10−4	1.45 × 10−5	−0.00231	−8.37 × 10−4	4.09 × 10−4
High-voltage power line	−0.00141	5.87 × 10−4	1.16 × 10−5	−0.00262	−0.00139	−3.14 × 10−4
Deviance	365.17	36.9	2.348	288.5	368.4	431.3
CIE model
CIE	−5.196	1.022	0.04896	−7.494	−5.113	−3.431
Deviance	371.9	34.37	2.158	296.1	375.2	429.8

.

Settlements exerted the strongest negative effect on eagle presence in the multivariate model, with the 95% credible intervals of the coefficient distribution falling entirely below zero. Minor roads and high-voltage power lines also had significantly negative, albeit smaller, effects. The median coefficients for major roads, railways, and medium-voltage power lines were negative as well; however, their credible intervals overlapped with zero, indicating fewer certain effects. In the CIE model, the univariate coefficient distribution was clear, suggesting a substantial adverse impact on eagle distribution.

The models estimated the probability of occurrence for each UTM cell as follows: The highest proportion of correctly classified cells was achieved using a probability threshold of 0.40 in the multivariate model and 0.37 in the CIE model (Figure 2). Using these thresholds to separate predicted from unpredicted cells, the classification accuracy reached 92% for the multivariate model and 89% for the CIE model (

Table 3. Comparison of the multivariate and CIE models’ prediction results with the probability values that gave the maximum percentage of cells classified correctly in the core study area.

	Multivariate Model		CIE Model
	Predicted	Unpredicted	Predicted	Unpredicted
Probability	>0.4	<0.4	>0.37	<0.37
Occupied	95	26	94	27
Unoccupied	11	316	22	305

). The cells with probability values below 0.10 accounted for approximately one-third of the core study area (30% in the multivariate model and 32% in the CIE model). Although these low-probability cells included only two (<2%) occupied cells in both models, we used this threshold to delineate areas most likely unsuitable for eagles.

The AUC and Cohen’s $\kappa$ values of the two competing models were similar (

Table 4. Comparison of the multivariate and CIE models’ AUC and Cohen’s $\kappa$ statistics.

Model	AUC	AUC SD	Cohen’s $\mathit{\kappa }$	Lower CI	Upper CI
Multivariate	0.935	0.013	0.622	0.54	0.73
CIE	0.932	0.014	0.578	0.5	0.65

), and the goodness-of fit plots showed no differences in pattern or bin distribution (Figure 3). Given that the simpler CIE model performed equally well in describing the presence of imperial eagles within the study site, we selected this model to estimate the probability of occurrence across the eastern Hungarian Plain.

3.3. Infrastructural Effects on the Potential Distribution in the Eastern Hungarian Plain

We divided the UTM cells of the eastern Hungarian Plain into three probability categories based on the values predicted by the prediction model: low, medium, and high. A threshold of 0.37, which maximized the percentage of correctly classified cells in the CIE model, was used as the lower limit for high-probability cells, while a threshold of 0.10 was defined as the upper limit for low-probability cells (Figure 1).

The prediction map indicated that a substantial portion (32.06%) of the Hungarian Plain is likely suboptimal for imperial eagle nesting due to the existing infrastructure network. However, the model also suggested that approximately two-thirds of the region still provides potentially suitable breeding habitats. Within this area, 24.37% of the plain was identified as high-priority habitats, where the absence of infrastructure may facilitate future colonization by the species. The observed distribution of imperial eagles within the model prediction area during the period 1989–2008 supported the model’s outcomes: only one occupied cell (1.41%) fell within the low-probability category, while 43 occupied cells (60.56%) were located in areas classified as high probability.

4. Discussion

Over the past two decades, eastern imperial eagles have shifted their primary breeding habitats from relatively undisturbed mountain forests to densely populated open agricultural landscapes in the Pannonian ecoregion [30,31]. Their growing population appears to be well adapted to human presence and, in some cases, can tolerate proximity to infrastructure [54]. Nevertheless, our habitat-selection models demonstrate that the network of settlements and linear infrastructure continues to significantly influence the breeding distribution within this densest Central European population of the species.

Similar findings have been reported in other European studies, in which the presence of infrastructure and human disturbance were among the primary factors shaping habitat selection in several rare raptor species [15,17], including the Spanish imperial eagle (Aquila adalberti) [55,56]. The Spanish imperial eagle, the closest relative of the eastern species, is confined to the Iberian Peninsula and is considered one of the rarest raptors in the world [57,58]. In Spain, significant negative impacts of human disturbance and infrastructure on habitat selection and behavior have been demonstrated at various spatial scales [55,56,59,60]. However, until now, such data have been lacking for the eastern imperial eagle, which inhabits markedly different landscapes with various infrastructure densities across Central and Eastern Europe.

Habitat selection models are typically designed to identify the most influential environmental variables but do not explicitly quantify the habitat-reducing effects of infrastructure. In this study, we specifically focused on different types of infrastructure under the assumption that other environmental and ecological conditions (i.e., prey abundance, nesting and foraging habitat structure, climate, etc.) are generally suitable within the species’ local distribution range.

Our results indicate that the existing infrastructure network significantly reduces the availability of suitable nesting habitat by preventing settlement in approximately one-third of both the core study area and the broader eastern Hungarian Plain. However, the current gaps between settlements, roads, and power lines still offer sufficient suitable habitat to support further range expansion of the species.

Habitats with low infrastructural impact were occupied by imperial eagles three times more frequently than expected based on their relative availability. Several large contiguous areas of suitable habitats (such as the Hortobágy, Bihar, and Dévaványa Plains) are designated nationally protected areas and classified as Special Protection Areas (SPAs) within the EU Natura 2000 network. The legal obligations associated with these protected areas, including population monitoring and habitat preservation, are expected to support both the immediate detection and long-term conservation of breeding pairs colonizing these regions.

However, as observed for other wide-ranging European raptors [17,61], a substantial proportion of high-probability habitats for imperial eagles lie outside both SPAs (67.63%) and nationally protected areas (84.07%). In these unprotected regions, conservation efforts are often hindered by lower survey intensity and limited capacity to advocate against conflicting land use interests, such as infrastructure development. Importantly, these findings also support other globally threatened bird species, such as the saker falcon (Falco cherrug), red-footed falcon (Falco vespertinus), great bustard (Otis tarda), and European roller (Coracias garrulus) [6]. The infrastructure projects affecting these unprotected priority habitats should consider the presence and ecological sensitivity of these species, highlighting the urgent need for both single- and multi-species habitat selection studies in the region.

Species’ tolerance to infrastructure varies considerably depending on the species [62], habitat [63], individual behavior [64], and type of infrastructure. Interestingly, the imperial eagles in our study area appeared more tolerant of major paved roads than minor ones, contrary to general expectations. This may be due to their sensitivity to irregular and unpredictable human activities, such as pedestrians, cyclists, and parked vehicles, which are more common along minor roads. In contrast, the traffic on major roads and railways tends to be continuous and predictable, which may elicit less disturbance [54].

These findings are consistent with studies on Spanish imperial eagles, which reported minimal reactions to moving vehicles but heightened sensitivity to stationary cars, pedestrians, and especially to hunters and ecotourists (who could pay direct attention to the birds) within a specific buffer distance around the nest [60]. To mitigate such disturbances, conservation practices often designate buffer zones around nesting sites where specific human activities are temporarily restricted [65]. For example, conservation measures for Spanish imperial eagles include an inner buffer zone (500 m radius) with complete restrictions on human activity during the breeding season and an outer buffer zone (500–800 m) where limited activities, such as vehicle transit, may be permitted [60].

In Hungary, similar buffer zones ranging from 100 to 500 meters were applied around eastern imperial eagle nests from the 1980s to the 2010s. However, these were typically implemented only when a specific human activity (e.g., forestry, hunting, beekeeping, or ecotourism) was deemed likely to cause breeding failure [30,66]. In our study, imperial eagles successfully nested at distances well within 500 meters of all types of linear infrastructure, including paved public roads and railways. In such contexts, broad bans on human activity are neither feasible nor justifiable.

Moreover, similar restrictions around Spanish imperial eagle nests have been criticized, as the species can habituate to certain types of human activity. Overly cautious restrictions may exacerbate human–wildlife conflict and even increase persecution risks [67]. We suggest that highly territorial and long-lived species such as the imperial eagle are capable of assessing human disturbance levels during nest site selection and can often choose alternative locations if consistent disturbance is present. Therefore, we do not advocate for general restrictions on human activity around nesting sites. Nevertheless, conservational authorities may consider targeted protective measures in such cases where specific human activities or the large-scale development of infrastructure networks pose a significant threat to small or vulnerable populations.

5. Conclusions

Based on our findings and the implications discussed above, we propose the following conservation recommendations for the operation or development of infrastructure within active or potential imperial eagle breeding habitats:

(1)

Minimize further habitat fragmentation: Plans for new linear infrastructure (e.g., roads, railways, and power lines) should aim to minimize the further fragmentation of priority habitats of imperial eagles and other globally threatened species. Negative impacts can be reduced if new infrastructure is routed close to existing infrastructure (of the same or different types) and avoids bisecting previously continuous, undisturbed habitat patches.

(2)

Mitigate Electrocution Risks: Electrocution remains the most significant infrastructure-related cause of mortality and is likely the only one that can be effectively reduced through targeted mitigation measures [68,69,70]. According to the 2009 amendment of the Hungarian Nature Conservation Law (53/1996), all newly constructed or fully renovated power line sections must be designed to minimize risk to wild bird populations. In addition to enforcing this regulation, retrofitting approximately 700,000 existing hazardous power poles across the Hungarian Plain would provide a long-term solution [37]. In the medium term. modifying the roughly 14,000 most dangerous pylons, those located in high-probability imperial eagle habitats, as identified by our model, would substantially reduce electrocution-related mortality.

(3)

Prioritize Monitoring in High-Probability Areas: Intensive field surveys and monitoring should be focused on high-probability habitats, particularly in regions with limited coverage by SPAs or nationally protected areas (Nagykunság, Hajdúhát, and Békés Plain). Our habitat suitability model can assist in locating new breeding territories more efficiently, which is essential for implementing timely conservation measures and addressing threats such as habitat degradation or direct persecution.

(4)

Develop Further Predictive Models: Further research should aim to refine habitat occupancy and population dynamics models to better forecast changes in the distribution and breeding behavior of the Pannonian imperial eagle population. These models should incorporate additional variables, such as nesting tree availability, prey abundance, microhabitat structure, breeding success, mortality rates, density dependence, and long-term behavioral adaptations (i.e., habituation to human activity).