Grassland Alterations Do Not Affect Breeding Success, but Can Explain Dietary Shifts of a Generalist Raptor Species

Abstract

Habitat alteration is a widespread threat severely affecting large raptors because of their low density and the huge area they inhabit. In this study, we assessed whether human-driven habitat alterations mediated dietary shifts of apex predators, focusing on the Eastern imperial eagle (Aquila heliaca). Following a bottom-up conception (before–after), we evaluated the effect of grassland change on the eagle’s dietary shift and breeding success. Land use patterns underwent a significant transformation over the study period, creating a large decrease in grasslands. The territories lost an average of 25.79% of their grasslands. Habitat alteration mediated dietary shifts, but had no reproductive consequences for eagles. Eagles became 1.90 times more likely to predate on northern white-breasted hedgehog and 1.62 times more likely to forage on white stork in the period after grassland alteration. The frequency of tortoises also increased, and they were 4.04 times more likely to be predated on in the years after transformation. Conversely, brown hare was 0.51 times less likely to be consumed in the grassland loss period, while this likelihood was 0.54 times lower for rodents and 0.64 times lower for the European souslik. Doves, meanwhile, were 2.73 times more likely to be predated on in the years following grassland destruction. We found that the presence and biomass of songbirds correlated negatively with the breeding success of eagles, and biomass supply from European souslik was negatively associated with breeding success, while the white stork’s presence and biomass resulted positively in more progeny. Diet diversity did not have an effect on the eagle’s reproduction. The responses of these eagles may vary across territories, depending on how they rank their prey, as the territory effect was a powerful factor shaping dietary shifts for this top predator. Our results offer new evidence of the link between habitat alteration, dietary shifts, and reproductive success, contributing to our understanding of the enigmatic mechanism through which an apex predator successfully adapts to large-scale land use pattern transformation by increasing dietary specialization. We recommend restoration of habitat complexity, including preservation of field margins, grassland patches with scattered small shrub formations, and grassland margins between medium-sized arable lands, promotion of measures for traditional grassland management through gradual grazing, and a ban on the use of shredders.

1. Introduction

Land use and climate change and their potential synergistic effect are the major reasons for biodiversity loss globally. In the late 20th century, a period characterized by agricultural intensification, most farmland bird populations collapsed [1,2,3]. Agricultural intensification is mainly expressed in the conversion of grassland habitats into arable lands, changes in the land use pattern and agricultural practices, and increased use of fertilizers and pesticides, leading to continuous habitat loss and degradation [1]. Habitat loss is a widespread threat severely affecting large raptors because of their low density and the huge area they inhabit. Top predators are considered good indicators of biodiversity and ecosystem health due to their high position at the trophic level and their sensitivity to a changing environment [4,5]. Thus, their abundance and behavior can be used to monitor the biotic effects of environmental pollution, landscape alteration, and biodiversity, in general [6]. Although agricultural intensification often leads to the loss of ecological heterogeneity [7,8,9], it can also provide resource opportunities that can be beneficial in the same circumstances [10,11,12]. However, sedentary territorial raptors are extremely vulnerable to habitat deterioration because they are typically dependent on productive, year-round territories with a stable supply of predominant prey [13]. Several open-land raptors, including the Eastern imperial eagle (hereafter EIE), have been strongly influenced by agricultural intensification [14,15]. Therefore, understanding the mechanism through which apex predators respond to habitat transformation is crucial for their conservation.

Habitat alteration via agricultural intensification can degrade or destroy preferred habitats and thus deplete prey for raptor species [16,17,18]. Changes in landscape features affect the habitats of prey species, turning them into areas of unfavorable conditions, thus forcing the animals to avoid these territories and leading to reduction in the dietary resources for predators. However, raptors can adapt their behavior by shifting their dietary and foraging patterns, which affects their reproduction or nestling survivorship [12,19].

Grasslands represent a complex mixture of various herbaceous species, legumes, and herbs. They are important habitats for different taxa, so changes in their condition and exploitation affect the functioning of ecosystems [20]. Globally, grasslands are listed as the second most commonly used habitat type for raptors, offering foraging and nesting places for different species [21]. As a result of the contradiction between the European Agricultural Policy and nature conservation [22,23], changes in land use patterns have caused a dramatic decline of different raptor species closely related with grasslands [16,24,25].

In this study, we assessed whether anthropogenic habitat alteration mediated dietary shifts of apex predators, focusing on a territorial sedentary species, the EIE, which inhabits open grounds and prefers foraging in grassland habitats [26]. In Bulgaria, the EIE population has been recovering since 2001 after a severe decline in the 20th century [27,28]. The conservation efforts invested to increase the breeding success and the survival of adults and floaters have included nest guarding, supplementary feeding, and insulation of hazardous electricity poles, as well as different campaigns among local communities and authorities [29]. The population has increased in lowland and hilly areas, but not in high mountains, where the last known territory has been abandoned since 2016 [28]. In addition to common threats, such as electrocution, persecution, disturbance, and poison baits [30], a new limiting factor has emerged in the past decade, that is, large-scale transformation of the main foraging habitats of the species, leading to a decreased territory occupancy rate and reduced habitat quality [15]. It was found that the abundance of profitable prey, such as the European souslik (Spermophilus citellus), has severely declined in most of the EIE territories in the past decade [28]. Previous studies considered regional and seasonal diet differences [31] and temporal and spatial diet alteration [28], but the enigmatic mechanisms through which the species responds to landscape changes, including trophic interaction, have remained unexplored.

We hypothesized that land use pattern alteration of the main foraging habitat has mediated EIE dietary shifts. The classic Optimal Foraging Theory (OFT) predicts that individuals will select the most favorable habitats for foraging to minimize energy expenditure and maximize fitness [32,33]. In addition, according to the Landscape Heterogeneity Hypothesis (LHH), more diverse landscapes harbor higher number of species, which means a wider trophic niche for an animal [34,35]. EIEs consume a diverse array of prey from different taxa, but grassland-associated species, such as sousliks (Spermophilus sp.), leporids (Leporidae), and hedgehogs (Erinaceus sp.), are predominant in the diet [31,36,37]. Moreover, sousliks [26] and leporids [38] are favored and probable energetically profitable prey when present (i.e., higher-quality). Given their habitat associations, it is possible that anthropogenic changes in grasslands have altered the distribution and abundance of the main prey in modified territories. Thus, because prey availability and abundance may vary among landscapes and influence breeding success of raptors [39], which presents a valuable opportunity to investigate whether anthropogenic habitat changes mediate diet alteration and the reproductive output of apex predators.

We integrated changes in land use patterns, reproductive output, and dietary data to test four predictions related to our central hypothesis. First, at the local (territory) scale, we predicted that the degree of grassland transformation would affect the diet of eagles, so in more transformed territories, there would be a shift from the prevalence of grassland-associated prey in the diet toward other substitute prey. Our expectation was that the eagle’s dietary response would depend on the particular territory. Second, we predicted that at the population scale, diet alteration of the EIE in both frequency of prey and biomass contribution would be mediated by large-scale land use pattern change. Third, we hypothesized that habitat alteration would decrease (i) habitat diversity and (ii) diet diversity. Fourth, we predicted that if favored higher-quality grassland-associated prey species decreased, then breeding success would also drop. By testing these predictions across landscapes with different ecological conditions and histories, we have produced an insight into the role of habitat alteration as a mechanism that shapes the response of apex predators to anthropogenic landscape modification.

2. Materials and Methods

Study design First, we evaluated bottom-up consequences of grassland changes in 24 breeding territories occupied by EIE pairs in Bulgaria. Detailed data about landscape alteration and the general effect on the eagle’s demography were obtained from the literature [15]. Since the EIE is strongly associated with different types of grasslands [15,26], we analyzed whether there were any effects of the alteration of this main foraging habitat on the eagle’s diet, and if so, whether the breeding success was influenced. For the purpose of our study, we assessed the effect of grassland alteration by applying a bottom-up concept (before–after). Following Bulgaria’s accession to the EU (2007), in 2011–2012, a massive transformation of the land use patterns through ploughing of natural and semi-natural grasslands and conversion into arable lands was recorded [15]. Therefore, the first period we studied (before) included the condition of habitats prior to their transformation, between 2001 and 2010. The second period included the years after the transformation (2012–2021). In the habitat analysis, we followed the procedures previously described [15].

2.1. Habitat Associations

Habitat associations were analyzed through Arc GIS 10.4.1 [40]. The analysis was based on a Corine Land Cover layer of Bulgaria valid for 2006, as well as an orthophoto layer valid for that same year (http://212.122.182.101/MRRB/ accsessed on 3 March 2021). A buffer of a 5 km radius around the nesting centroids of each EIE pair was used. This corresponded to the width of the territory occupied by an EIE pair in Bulgaria [15]. Remote verification of the grassland habitats was carried out during the first period of the study through orthophotography covering the country’s territory. Changes in land use patterns were verified through visits to all polygons in 2012 to detect any differences compared to the baseline. For each territory, changes that occurred during the survey periods were recorded. The habitat categorization was based on the Corine Land Cover classes. For each of the surveyed territories, we calculated the diversity index (H) for the habitats using the original land cover types over the different periods of the study [41].

2.2. Dietary Data Analysis

We assessed the bottom-up relationship (before–after habitat changes) of dietary data in the same EIE territories. General dietary data had already been published, but in a different context [28,31]. Food remains, bones, feathers, and pellets were collected from inside and under nests [31,42]. The following types of remains were not included in the data in order to reduce the bias of indirect sampling, even if they were found under the nest sites: (1) single feathers, which could have been shed by live birds; (2) full carcasses of large animals, which could not have been brought there by the eagles; (3) old or deteriorated samples, which could have remained from previous years [37]. The material was identified through the comparative osteological collections of the National Museum of Natural History at the Bulgarian Academy of Sciences. Whenever possible, the minimum number of individuals (MNI) in each pellet or sample of prey remains was estimated based on the number of skeletal or keratinized body parts [42]. The MNI was determined by taking into account the age (juvenis, subadultus, adultus), sex, and size differences between individuals.

Prey was grouped into main categories following the previously published methodology [28,31]: lizards and snakes (Squamata), tortoises (Testudines), water birds (Anatidae, Ardeidae), poultry (Gallus gallus f. domestica, Anser anser f. domestica, Meleagris gallopavo f. domestica), phasianids (Phasianidae), gulls (Laridae), doves (Columbidae), songbirds (non-Corvidae passerines), corvids (Corvidae), stork (Ciconia ciconia), raptors and owls (Accipitridae, Falconidae, Strigidae, Tytonidae), hedgehog (Erinaceus roumanicus), hare (Lepus europaeus), souslik (Spermophilus citellus), rodents (Rodentia excl. European souslik), carnivores (Carnivora), carrion (Artiodactyla, Perissodactyla), and other animal food components (including other vertebrate taxa). The identified minimum number of individuals (MNI) was used in the analysis [28,31,42]. We quantified EIE diet diversity by calculating Levin’s food niche breadth (FNB) [43] for each territory in each period. Larger values of Levin’s index indicated higher dietary diversity. In order to avoid bias from disproportionate samples, territories represented by less than 25 prey specimens for a given period were excluded from the analysis [44]. Then, the final habitat and dietary data for 15 territories were fitted (Figure 1).

We analyzed the variation of main prey categories using data from about 3735 prey specimens. We compared the frequency, biomass contribution of the different prey categories, and Levin’s index of diet diversity during the two periods (before–after grassland alteration) in order to investigate if there were any evident long-term and large-scale dietary shifts.

We considered pairs that laid eggs and started incubating as breeding pairs [36,45]. The breeding success of eagles was evaluated as the number of fledglings per incubating pair in a given territory [29]. We examined near-equal samples of breeding attempts before (n = 120) and after (n = 122) the landscape transformation of our 15 focal territories. However, 21 breeding attempts were excluded from the analysis due to nest robbery or human disturbance.

Lastly, we used a before–after paired design to study whether EIE breeding success changed following the land use pattern alteration in Bulgaria, Southeast Europe.

2.3. Statistical Procedures

Firstly, we applied a set of over-parameterized linear models (GLMs) with Type III error distribution. We ran the following models: (1) for the proportion of grasslands; (2) for the frequency of prey categories; (3) for the biomass contribution of prey categories; (4) for habitat diversity; (5) for diet diversity; and (6) for the breeding success, including the period (before–after habitat change) as a response factor (fixed effect) and grassland abundance, each prey category, Shannon’s diversity index of habitats, Levin’s diversity index of diet, and breeding success as explanatory factors. To control for spatial variation in our explanatory factors, we included a random factor “eagle territory” to account for data pooled within each territory. The eagles nesting in a given breeding territory can change over the years [46], and the design we used could not account for whether the same individuals predated on all prey or reproduced during the whole study period. Hence, the factor “individual eagle” could not be included in our model. Nonetheless, we considered our aim to detect long-term and large-scale changes in the eagles’ diet and breeding success achieved because of the large applicable and representative data sampling.

Secondly, to describe the dietary shifts, we built simple mixed models (GLMMs) with a binomial error distribution and logit link function including the period as a binary factor (before; after) and prey categories that had demonstrated a powerful significant trend (adjusted R² > 0.64) as explanatory factors. “Eagle territory” was included in the models as a random factor. We ran two GLMMs: one for the prey frequency, and one for the prey biomass. The design we used could not account for the fact that samples collected in the same nest or nearby trees could have been predated in different years by the same individuals (see above) [28,37]. So, the factor “individual eagle” could not be included in our models. We used the Akaike information criterion corrected for small sample sizes (AICc) for model selection and chose models with the lowest AICc value from the set of candidate models. All models with an AICc value < 2 from the model with the lowest AICc (AICcmin) were considered as among the best models (_∆AICc = AICi − AICcmin) [47]. The relative importance of each model was estimated through the weight of AICc (w_i), so that all the weights of the models added up to 1.

Explanatory parameter estimates (β²) with lower (95%) and upper CL (95%) and a probability value (p) of the explanatory factors were evaluated. Results with p ≤ 0.05 were considered significant. Values were provided as means ± standard error (SE).

The data calculated as percentages (grassland frequency, prey frequency, and biomass frequency) were converted into proportions and then Arcsin transformed, to achieve a close-to-normal distribution [48]. To evaluate the results of the regression models, we used the adjusted R² value as a correction factor. We checked for the presence of heteroscedasticity of residuals of our dataset using a fitted value vs. residual plot diagram [49]. We fitted the regression line to a set of data, and then created a scatterplot that showed the fitted values of the model vs. the residuals of those fitted values. After that, variables that showed heteroscedasticity were excluded from the analysis. we excluded the categories lizards, snakes, and gulls from the prey frequency analyses, and gulls, rodents, and other animals from the biomass analyses.

The relationship between degree of grassland alteration, proportion of dietary shifts, diet diversity, breeding success, and prey frequency or biomass supply was estimated through Spearman’s rank correlation (r_s) and Pearson’s (r) correlation coefficients.

All data were analyzed using Statistica for Windows, release 12 [50], R v.2.15.2 [51], and Past Version 4.08 [52].

3. Results

3.1. Grassland Alteration

Land use patterns in EIE were subject to significant transformation, creating a large decrease in grasslands, both for individual territories and as a whole (adjusted R² = 0.84, β² = −0.34, CL: −0.18/−0.50) (

Table 1. Results of the over-parameterized linear model (GLM) carried out to analyze the trends for the following categories: grassland abundance, diversity of habitats (Shannon’s H), food niche breadth (Levin’s FNB), and breeding success. Territory was included as a “random” factor; period (before: 2001–2010; after: 2012–2021) was included as a “fixed” factor. Significant values are given in bold.

Category	Factor	Effect	df	F	p
Grassland abundance	Territory	Random	14	10.56	<0.001
Grassland abundance	Period	Fixed	1	20.69	<0.001
H	Territory	Random	14	43.88	<0.001
H	Period	Fixed	1	4.88	0.04
FNB	Territory	Random	14	0.50	0.90
FNB	Period	Fixed	1	1.61	0.23
Breeding success	Territory	Random	14	0.96	0.53
Breeding success	Period	Fixed	1	7.73	0.01

). While before landscape transformation, the mean percentage of grasslands in our focal territories was 42.42 ± 21.57, after the plowing, their mean share dropped to 29.1 ± 13.99 (Figure 2). The territories lost an average of 25.79% of their grassland habitats. The most transformed territory had its grassland habitat reduced by 51.55%, and one-third of the territories had lost an average of 41.78% of their pastures. Out of all 15 studied territories, only one recorded an increase in the area of grassland in the second period, by 3.24%.

3.2. Dietary Shifts

We identified 2062 and 1673 individual prey items in the pre-change (2001–2010) and post-change (2012–2021) periods, respectively. The eagles’ diet in the two compared periods differed both in composition and biomass supply (Figure 3). Eagles were 1.90 times more likely to forage on northern white-breasted hedgehog (Erinaceus roumanicus) (adjusted R² = 0.66, β² = 0.53, CL: 0.76/0.29) and 1.62 times more likely to forage on white stork (Ciconia ciconia) (adjusted R² = 0.66, β² = 0.30, CL: 0.53/0.07) in the period after grassland alteration. The frequency of tortoises also increased, and eagles were 4.04 times more likely to predate them in the years after landscape transformation (adjusted R² = 0.52, β² = 0.35, CL: 0.63/0.08). Conversely, brown hare (Lepus europaeus) was half (0.51) as likely to be consumed in the grassland loss period (adjusted R² = 0.53, β² = −0.48, CL: −0.20/−0.75), while this likelihood was 0.54 times lower for rodents (adjusted R² = 0.53, β² = −0.33, CL: −0.06/−0.61) and 0.64 times lower for the European souslik (adjusted R² = 0.71, β² = −0.25, CL: −0.04/−0.46). Furthermore, water birds also had a reduced share (0.24 less likely) in the second period (adjusted R² = 0.48, β² = −0.41, CL: −0.13/−0.70). A marginal increase in corvids frequency in the EIE diet was observed only in some territories (

Table 2. Results of the over-parameterized linear model (GLM) constructed to analyze the trends of the different prey categories (frequency and biomass contribution). Territory was included as a “random” factor; period (before change: 2001–2010; after change: 2012–2021) was included as a “fixed” factor. Due to heteroscedasticity of residuals, the categories lizards, snakes, and gulls were excluded from prey frequency analyses, while the categories gulls, rodents, and other animals were dropped from biomass analyses. Significant values are given in bold.

Categories	Frequency of Prey Contribution					Biomass of Prey Contribution
	Factor	Effect	df	F	p	Factor	Effect	df	F	p
Tortoises	Territory	Random	14	2.81	0.03	Territory	Random	14	2.73	0.04
Tortoises	Period	Fixed	1	7.53	0.02	Period	Fixed	1	5.43	0.04
Water birds	Territory	Random	14	2.33	0.06	Territory	Random	14	3.24	0.02
Water birds	Period	Fixed	1	9.53	0.008	Period	Fixed	1	16.02	0.001
Poultry	Territory	Random	14	0.43	0.94	Territory	Random	14	0.51	0.89
Poultry	Period	Fixed	1	3.22	0.09	Period	Fixed	1	2.48	0.14
Phasianids	Territory	Random	14	1.70	0.17	Territory	Random	14	2.65	0.04
Phasianids	Period	Fixed	1	6.93	0.02	Period	Fixed	1	8.51	0.01
Doves	Territory	Random	14	1.85	0.13	Territory	Random	14	3.75	0.009
Doves	Period	Fixed	1	6.39	0.02	Period	Fixed	1	5.62	0.03
Songbirds	Territory	Random	14	1.37	0.28	Territory	Random	14	1.85	0.13
Songbirds	Period	Fixed	1	6.52	0.02	Period	Fixed	1	4.37	0.055
Corvids	Territory	Random	14	2.71	0.04	Territory	Random	14	2.16	0.08
Corvids	Period	Fixed	1	0.002	0.97	Period	Fixed	1	0.007	0.94
Stork	Territory	Random	14	4.62	0.004	Territory	Random	14	4.14	0.006
Stork	Period	Fixed	1	7.79	0.01	Period	Fixed	1	9.61	0.008
Raptors and owls	Territory	Random	14	1.56	0.21	Territory	Random	14	2.58	0.04
Raptors and owls	Period	Fixed	1	1.48	0.24	Period	Fixed	1	0.50	0.49
Hedgehog	Territory	Random	14	3.35	0.02	Territory	Random	14	2.93	0.03
Hedgehog	Period	Fixed	1	23.48	<0.001	Period	Fixed	1	22.25	<0.001
Hare	Territory	Random	14	2.44	0.054	Territory	Random	14	4.88	0.003
Hare	Period	Fixed	1	14.09	0.002	Period	Fixed	1	43.58	<0.001
Souslik	Territory	Random	14	5.76	0.001	Territory	Random	14	4.64	0.003
Souslik	Period	Fixed	1	6.39	0.02	Period	Fixed	1	7.31	0.02
Carnivores	Territory	Random	14	0.69	0.75	Territory	Random	14	0.47	0.91
Carnivores	Period	Fixed	1	0.04	0.84	Period	Fixed	1	0.16	0.70
Carrion	Territory	Random	14	1.11	0.42	Not evaluated
Carrion	Period	Fixed	1	0.27	0.61	Not evaluated
Rodents	Territory	Random	14	2.92	0.03	Excluded due the heteroscedasticity
Rodents	Period	Fixed	1	6.87	0.02	Excluded due the heteroscedasticity
Other animals	Territory	Random	14	1.27	0.33	Excluded due the heteroscedasticity
Other animals	Period	Fixed	1	2.41	0.14	Excluded due the heteroscedasticity
Lizards and snakes	Excluded due the heteroscedasticity					Territory	Random	14	3.76	0.009
Lizards and snakes	Excluded due the heteroscedasticity					Period	Fixed	1	2.18	0.16
Gulls	Excluded due the heteroscedasticity					Excluded due the heteroscedasticity
Gulls	Excluded due the heteroscedasticity					Excluded due the heteroscedasticity

). Meanwhile, a significant decline in the post-alteration period was observed for the presence of phasianids (adjusted R² = 0.35, β² = −0.40, CL: −0.07/−0.72) and songbirds (adjusted R² = 0.27, β² = −0.41, CL: −0.06/−0.75) (Table 2). The frequency of consumed doves clearly increased after habitat transformation (adjusted R² = 0.37, β² = 0.37, CL: 0.69/0.06), with eagles 2.73 times more likely to predate them in the years that followed grassland destruction.

Shifts in biomass between time periods were more or less similar to shifts in frequency (Figure 3). The proportion of biomass from hedgehogs (adjusted R² = 0.62, β² = 0.54, CL: 0.78/0.29), storks (adjusted R² = 0.64, β² = 0.34, CL: 0.58/0.11), and tortoises (adjusted R² = 0.50, β² = 0.31, CL: 0.59/0.02) increased in the post-transformation period. Biomass provided by brown hare (adjusted R² = 0.77, β² = −0.59, CL: −0.40/−0.78) and European souslik (adjusted R² = 0.66, β² = −0.29, CL: -0.06/-0.52) largely decreased after grassland transformation. The supply from water birds (adjusted R² = 0.62, β² = −0.46, CL: −0.21/−0.71) and phasianids (adjusted R² = 0.51, β² = −0.38, CL: −0.1/−0.66) was lower in the post-transformation period, while doves’ biomass contribution rose (adjusted R² = 0.60, β² = 0.28, CL: 0.53/0.03), though, in general, all these categories were of negligible importance for biomass provision (less than 2% on average) (Figure 3). Finally, in some eagle territories, lizards, snakes, raptors, and owls contributed to the delivery of more biomass (Table 2).

3.3. Modeling General Effect of Dietary Alteration

The top model for prey frequency (_ΔAIC = 0.00, w_i = 0.99) clearly demonstrated that white stork, northern white-breasted hedgehog, and European souslik shaped dietary shifts of the EIE with a random effect of territory. In terms of biomass change, the first candidate model included “random” territory and brown hare (_ΔAIC = 0.00, w_i = 0.46), while the second model determined only brown hare as an important factor (_ΔAIC = 0.42, w_i = 0.37) and the third candidate model for biomass shift instead contained white stork and brown hare (_ΔAIC = 2.07, w_i = 0.17).

3.4. Threshold Effect of Grassland Change in Individual Territory

We did not find a significant relationship between the proportion of transformed grassland cover in a given territory and the proportion of dietary shifts of different prey categories. However, a negative, but not statistically significant (r_s = −0.42, p = 0.12) association was found of the degree of grassland alteration and the increasing proportion of hedgehogs. A marginal negative relationship was also observed between the proportion of altered grassland and the changed frequency of rodents (r_s = −0.36, p = 0.18). Meanwhile, a positive, yet not clearly expressed association was noted between the change in European sousliks consumed and the degree of altered grassland habitats in a given territory (r_s = 0.28, p = 0.31).

3.5. Habitat Diversity and Diet Diversity

While habitat diversity was statistically lower in the post-alteration period (2012–2021) compared to the pre-alteration period (2001–2010) (adjusted R² = 0.95, β² = −0.09, CL: −0.003/−0.17), diet diversity did not show a significant trend (adjusted R² = 0.28, β² = −0.27, CL: 0.18/−0.72) (Table 1, Figure 4). The mean habitat diversity was 1.78 ± 0.33 and the mean FNB was 5.67 ± 2.08 before landscape alteration, followed by 1.72 ± 0.33 for the habitat index and 4.64 ± 1.78 for the diet index in the post-alteration period. However, FNB significantly positively correlated with brown hare (Pearson’s r = 0.62, p < 0.001), poultry (Pearson’s r = 0.53, p = 0.003), phasianids (Pearson’s r = 0.52, p = 0.003), carnivores (Pearson’s r = 0.44, p = 0.02), and water birds (Pearson’s r = 0.42, p = 0.02), while it showed a negative correlation with northern white-breasted hedgehog (Pearson’s r = −0.40, p = 0.03).

3.6. Breeding Success and Effect of Grassland Transformation and Dietary Shifts

In general, the breeding success significantly increased in the period after grassland alteration (adjusted R² = 0.18, β² = 0.47, CL: 0.83/0.11), without a “territory effect” affecting this process (Table 1) (Figure 5). The mean success of reproduction in the years before habitat transformation was 0.97 ± 0.60 fledglings per incubating pair. In the period after grassland alteration, the mean breeding success of eagles grew to 1.45 ± 0.37.

We found that the presence (Pearson’s r = −0.60, p < 0.001) and biomass (Pearson’s r = −0.73, p < 0.001) of songbirds statistically negatively correlated with the breeding success of eagles, while the presence (Pearson’s r = 0.41, p = 0.02) and biomass (Pearson’s r = 0.39, p = 0.03) of white stork significantly positively correlated with more progeny. Meanwhile, biomass supply from European souslik was negatively associated with breeding success (Pearson’s r = −0.43, p = 0.02), while the influence of their frequency had only a marginal impact (Pearson’s r = −0.35, p = 0.058). It should be also noted that the frequency of northern white-breasted hedgehog in the diet marginally positively affected the breeding success of the eagles (Pearson’s r = 0.34, p = 0.06). However, FNB did not have any effect on the eagles’ reproduction in our study (Pearson’s r = 0.05, p = 0.79).

4. Discussion

Following Bulgaria’s accession to the EU and the launch of subsidies in agriculture related to the implementation of the Common Agriculture Policy (CAP), significant changes have occurred in the land use patterns over vast territories [15,25,53]. Large raptors are strongly dependent on extensive stockbreeding and pastures as they find abundant prey in these habitats. For instance, grasslands are rich in various small- and medium-sized mammal species, birds, and reptiles. A positive relationship between the amount of grassland and the occupancy rate of territories was proven for the EIE [15]. We found that long-term alteration of grassland habitats may imply loss of up to half of the favorable foraging habitat for some eagle territories. However, different raptor species show different response to habitat alteration. Some are positively influenced and increase in abundance [10,54], while others disappear locally [24,25].

In this study, we found that anthropogenic-driven grassland transformation mediated dietary shifts of the EIE, a top predator. Confirming the LHH, after large-scale and widespread grassland decreases, habitat diversity also decreased, the EIE diet became less diverse, and its composition shifted. Eagles predated fewer grassland-associated prey, such as European souslik, brown hare, small rodents (Microtinae, Muridae, Spalacidae), and partridges, and more northern white-breasted hedgehog, white stork, tortoises, and doves in the years after habitat transformation. Water birds and songbirds were also less frequently detected in the diet. In some territories, the EIE foraged more for corvids, lizards, and snakes or increased intraguild predation. In the years before grassland alteration (2001–2010), hedgehogs were the primary caught prey in six of our fifteen focal territories, but in only one of these did it dominate the eagles’ diet (over 50%). In this period, the diet was more diverse and, of the remaining nine territories, sousliks were the primary prey in four, small rodents in two, and brown hare, gulls, and squamates in one territory each. After habitat transformation (2012–2021), hedgehogs became the dominant prey in twelve of the fifteen studied territories, and in six of them their frequency was more than 50% of the predated animals. In the years after habitat change, European souslik was the favorite prey in only two eagle territories, and white stork became the primary prey in one of the EIE’s territories. The near-doubled frequency and biomass contribution of hedgehogs and storks after the grassland loss was a novel finding and suggested a substantial change compared to the historical diet pattern before the main foraging habitat transformation, with it observed that the brown hare and European souslik populations crashed. Vast landscape pattern shifts, expressed in severe grassland habitat reduction, were followed by a significant decline of European souslik and brown hare abundance [28,55]. Thus, the eagles faced a severe constraint and had to shift their main food source.

The differences in the food niche breadth between pre-alteration and post-alteration periods were not significant, corresponding to a low level of food stress [56]. A positive association of FNB with brown hare, poultry, phasianids, carnivores, and water birds indicated that the amounts of these prey types were likely not sufficient for the eagles in their home ranges, meaning they had to include a variety of prey in their diet, thus widening the niche breadth. Conversely, the significant negative correlation of FNB and hedgehogs meant that when hedgehogs were abundant and available in the eagles’ territory, they specialized in the predation of this plentiful and predictable prey and had low diet diversity.

Our suggestion that the proportions of altered grasslands at the EIE local territory scale would determine the dietary shifts of grassland-associated prey was not clearly proven. The median proportion of grassland vegetation within an eagle breeding territory declined from 98% to 48%. However, in one EIE territory, the proportion of pastures inversely demonstrated a negligible increase (see above). As a whole, the decline of grassland had a negative effect on the changed proportion of consumed hedgehogs and rodents and pushed eagles to predate more tortoises and white storks. Unfortunately, we had not estimated the abundance of such prey species before habitat alteration, which prevented us from comparing their densities in the pre-alteration and post-alteration periods. One finding we did make was that there was a continuous decline in the European souslik abundance in our study area, which offers support for a previous finding in the literature of such a decline [28].

However, the proportion of sousliks eaten demonstrated a contradictory effect in terms of changed grassland proportion. In some eagle territories, this prey disappeared from the diet even if only a small share (less than 2%) of grassland cover was destroyed, while in the others, the loss of half of the pastures in the breeding home range did not significantly affect the frequency at which sousliks were eaten. Since sousliks are a patchily distributed prey, the particular patch of grassland fragments affected by alteration is of crucial importance. Even if a negligible part of the eagle territory is plowed, if habitat alteration affects foraging fragments important to birds, such as souslik colonies, the eagles’ main food resource will be compromised and they will switch to another subsequent prey if available. For example, in four of our fifteen focal EIE territories, European souslik was the primary prey caught before grassland transformation. In the years following grassland alteration, in two of these four territories, which had lost grassland vegetation at between 28% and 35%, respectively, sousliks were still the most frequently captured prey. In the third one, the grassland loss was less than 3%, but sousliks were no longer the main prey species in the years following habitat transformation. In the opposite case, in the latter territory, sousliks were displaced from the first place in the eagles’ diet, although the area of grassland had increased by 3.24% (see above). In contrast, in another case, the eagles were 4.66 times more likely to eat sousliks, regardless of the fact that in this territory, the pastures had been halved. As mentioned before, it is highly significant which part of the territory is altered. For example, an eagle territory might be largely destroyed but the souslik colony may remain unharmed, meaning its population increases.

The severe decline of another grassland-associated species, the brown hare, in the EIE’s diet was not dependent on the proportion of altered grassland. Hares can have a variety of habitats including different types of grassland habitats, cultivated areas, forests, etc. [55]. They prefer grasslands with scattered bushes, where they find shelter and protection from different predators. The conversion of grassland vegetation into arable land generally reduces their preferred habitats. In parallel, the process of widespread removal of scrub vegetation from grasslands, which began in the past decade, further degrades their habitats and leads to avoidance by hares (author’s data). In response to habitat change, eagles shift from the increasingly declining hares toward a substitute prey in their diet. For example, in four of our study focal territories, brown hare was substituted as the main biomass-providing source, with hedgehogs or white storks taking over in the period after transformation of grasslands. Of note, the stress on the brown hare in our study area was rising due to its increasing role in the eagle owl’s (Bubo bubo) diet [57]. Further investigation of the intraguild trophic relationships among top predators will be of critical importance to understanding this effect.

We also speculated that eagles had possibly shifted their foraging behavior and captured more hedgehogs and tortoises as a result of a change in the overall management pattern of grassland habitats, not just their alteration. In the years following the collapse of the planned socialist economy, the country’s agriculture was in decline and much of the grasslands were left to overgrow, making prey, such as hedgehogs and tortoises, hard to catch by eagles. Conversely, the intensification of agriculture following Bulgaria’s accession to the EU in 2007 led to the maintenance (most often mechanized) of low grass cover in pastures, as well as the complete removal of bushes from such areas. This benefited the eagles’ access to prey, such as hedgehogs and tortoises, and they became more frequently predated. However, this issue needs further clarification.

Our expectation that dietary shifts of the EIE would be revealed to have depended on the particular territory found full confirmation. When modeling the effect of dietary alteration, we found that the random territory effect was a powerful factor shaping the dietary response of the EIE to habitat changes. This suggests that the response of individual eagles may vary across territories, depending on how they rank their prey, resulting in different resource use patterns, giving different importance to their substitute prey [58,59]. This finding follows the competitive refuge model, which is in accordance with the optimal foraging theory (OFT) [32,33].

Among the most important factors influencing the reproduction of raptors are the quantity and quality of food [60,61,62]. Our prediction that the decrease in profitable and high-quality prey, such as European souslik and brown hare, would decrease the breeding success of eagles was rebutted. The European souslik and brown hare frequency and biomass supply in the EIE diet significantly declined, while, at the same time, the breeding success of eagles increased. Data on sousliks and brown hare indicated a high reproductive output of the EIE in a large part of the species distribution range [38,63,64], and they were removed as profitable and high-quality prey. So, why was there such a contradiction? We propose several independent explanations related to this system. Firstly, the breeding success of raptors was affected by other factors, such as weather and human disturbance [65,66,67,68]. As another factor, it is well known that nest guarding significantly improves the breeding success of the EIE [45]. However, the proportion of our focal nests that were guarded in the period before habitat alteration (n = 43) was almost the same as that in the years after transformation (n = 46) (author’s data). So, it seems that, in our study, this factor did not influence the differences in productivity. Looking elsewhere, the age of breeding birds is another important factor driving the reproduction of large raptor species [69,70,71]. It was documented that pairs of adult EIEs had significantly better breeding success than immature birds involved in reproduction [68]. In our study, 21.07% of the breeding eagles were in immature plumage in the pre-alteration period, compared to only 6.58% immature breeders in the years after habitat transformation in the same focal territories. Adult eagles gain more experience and successfully raise more progeny [68,71], so this change may have played a role.

Secondly, in light of OFT, individuals specializing in one prey type will be more efficient than generalist foragers [72]. New insights in OFT have identified a number of widespread patterns, such as localized searching after prey detection (“area-restricted search”) and enhanced detection after gaining experience with a particular prey type (“search image”) [73]. A case study on the EIE demonstrated that dietary specialization led to more efficient foraging and higher productivity [42]. One reasonable explanation is that adult birds with more specialized diets may provide a more consistent food supply to their nests. Then, the EIE responds to habitat changes via specialization in foraging for hedgehogs and storks, and thus significantly increases its productivity. Our finding of a significantly positive correlation between breeding success and the presence of storks and hedgehogs supported this hypothesis. Furthermore, the presence and biomass of sousliks correlated negatively with eagle breeding success. A reasonable hypothesis to explain this finding is the long-term and large-scale decline of sousliks in our study area [28]. In this scenario, the lack of sufficient food in the territories of eagles relying primarily on sousliks for food most likely resulted in fewer deliveries to the nest and fewer successfully reared fledglings. Similarly, it has been reported that the reproductive cycles characteristic of rodents, such as the European souslik [28], lead to an unpredictable food source, and in souslik-poor years, eagles have low breeding success. However, this issue needs further research. The lack of another sufficient abundant food source for the eagles to exploit, and hence successfully adapt to the sharply deteriorating conditions, could also be a reason to subsequently stop breeding and thus lower the territory occupancy rate [15]. Songbirds also negatively correlated with the productivity of eagles, mainly because of the small amount of biomass they provided. Thus, when included as subsequent prey when the primary prey was not abundant, they failed to provide sufficient resources for the offspring, and the eagles had lower fecundity rates, as a result.

The availability and biomass of hedgehogs and white storks may make them viable alternative prey for the EIE because, in terms of calory provision, they may be as profitable as hares and sousliks. The average weight of adult white storks is 3500 g and that of hedgehogs is 800 g, compared to 3000 g for the brown hare and 300 g for the European souslik [74,75,76]. As such, the profitability of alternative prey for the EIE, such as hedgehogs and white stork, is clear and should support increased nesting success. Further clarification on this issue is required.

Conservation Consequences

Our findings indicate several conservation consequences for the EIE. First, landscape pattern alteration can create an ecological trap. The conversion of grasslands into arable land and the removal of bushes from existing grasslands make it easier for eagles to access previously hard-to-reach prey, such as hedgehogs and tortoises. The extreme specialization of the EIE in hunting hedgehogs while ignoring other available prey, on the one hand, also modifies their foraging behavior. On the other hand, the assemblage of top predators such as the eagle owl, golden eagle (Aquila chrysaetos), white-tailed eagle (Haliaeetus albicilla), and lesser spotted eagle (Clanga pomarina), nesting in the same or neighboring areas, raises the pressure on hedgehogs for feeding ([56,57], author’s data). For the eagle owl, hedgehogs provide most of the food biomass, contributing 22.9%, [57], and the golden eagle has also adapted by shifting its diet to consist predominantly of hedgehogs [56]. Specialization in a particular major food source will unleash intra-specific and intra-guild competition. It is not clear how long the abundance of hedgehogs will be able to support this entire complex trophic interaction system before it collapses. The stress on hedgehogs is also increased by their direct extermination when shredders or mulchers are used to maintain grasslands. There have already been signs of a decrease in their abundance in our studied area (author’s data). Thus, the rapid and sharp deterioration of the main food source of both the EIE and other raptors may have consequences for the entire ecosystem that are difficult to predict.

Second, the consumption of more doves (mostly feral pigeon—Columba livia f. domestica) is likely to raise the toxins or pathogens in the EIE, such as Trichomonas gallinae, the protozoan that causes avian trichomoniasis and that can be particularly harmful [19,77,78]. Increased exposure to diseases or toxins resulting from dietary shifts may act as a threat multiplier together with the landscape change, leading to a population-level effect. Future research in this regard is warranted.

Last but not least, the increasing frequency of feral pigeon in the EIE’s diet can raise conflicts with pigeon fanciers, which, in turn, could result in persecution incidents. In our study area, there has already been evidence of such negative practices causing mortality of different raptor species, including the EIE [28].

5. Conclusions

In this study, we examined widespread land use pattern change, as expressed in the conversion of grassland-dominated landscape into large arable blocks sown with monocultures. We found that this land use pattern alteration affected the main foraging habitats and mediated dietary shifts, but did not have reproductive consequences for the EIE, an apex predator. The response of eagles may vary across territories, depending on how they rank their prey, and we found that the territory effect was a powerful factor shaping the dietary shifts of this top predator. Our results offer new evidence of the link between habitat alteration, dietary shifts, and reproductive success, as well as improve our understanding of the enigmatic mechanism through which an apex predator successfully adapts to large-scale land use pattern transformation by increasing dietary specialization. Given the increasing role of the northern white-breasted hedgehog in the trophic systems, and the widespread decline of sousliks and leporids, future work on the intraguild trophic interaction and consequences for raptor population (apex and meso-predator) trajectories and the entire ecosystem is warranted and will be useful.

One of our findings was that it is especially challenging to determine the threshold of transformation tolerance above which large-scale favorite-habitat alteration will severely affect the viability of the populations of prey. Our results also suggested that eagles may be resilient to some bottom-up effects of large-scale and long-term habitat alteration, but threats related to the dietary exposure to disease, the possibility of becoming ecologically trapped by increasingly narrow specialization, and intraguild trophic interaction or human persecution may increase the vulnerability of the EIE in the Thrace Ecoregion, Southeast Europe.

We recommend the re-establishment and preservation of field margins, grassland patches, and other landscape features that ensure its structural functioning. Restoration of habitat complexity, which includes the establishment of mosaic habitat structures, consisting of complexes of permanent grasslands with scattered small shrub formations and grassland margins between medium-sized arable patches, will enhance the connectivity of preferred grassland habitats and thus benefit the conservation of the different prey species in these already human-altered and fragmented landscapes. Promotion of the EU’s CAP and environmental policy should encourage measures for traditional management of grasslands through gradual grazing and a ban on the use of shredders.