Habitat Fragmentation on Bee (Hymenoptera: Apoidea) Diversity, Food Niches, and Bee–Plant Interaction Networks

Abstract

Habitat fragmentation poses a serious threat to bee communities, which are essential for pollination and biodiversity conservation. This study evaluated how habitat fragmentation in an oak forest in Zapopan, Mexico affected bee diversity, abundance, feeding niches, and bee–plant interaction networks. We compared a protected natural area with a nearby forest fragment that has been isolated from the main forest by urbanization for the past 10 years. Wild bee abundance and species richness in the fragmented area decreased by 74% and 70%, respectively, compared to the natural area, accompanied by a significant reduction in diversity. Community composition shifted mainly due to species loss; furthermore, there was persistence of generalist species such as Apis mellifera, which became more abundant in the forest fragment. Feeding niches in the fragmented area were narrower according to the Levin index, reflecting more restricted diets and increased interspecific competition. Interaction networks were simplified, showing fewer interactions, loss of specialist bees, and decreased equitability, although network specialization remained stable, and an almost-total turnover in interactions indicated a reconfiguration of pollination patterns. These findings suggest that fragmentation negatively affects bee community structures and their interactions with plants, potentially compromising pollination and ecosystem services. Conservation of protected areas and restoration of disturbed sites with native plants are recommended to support the recovery and stability of bee communities and their ecological interactions.

1. Introduction

Bees are one of the most diverse groups of pollinators in natural and agricultural ecosystems [1,2,3,4]. Furthermore, they participate in the reproductive cycle of angiosperms, so they are important for fruit and seed formation, ensuring the generational continuity of plant species, biodiversity, and food security [5]. Despite their importance, a decrease of more than 25% in the number of species has been recorded when comparing recent records to those of the past [6].

The observed decline in bee species diversity can be attributed primarily to environmental and agricultural factors, including habitat loss, pesticide use, increased pathogens, climate change, and the expansion of monocultures. These factors limit the resources necessary for their survival and negatively affect bee populations [7,8,9].

Habitat loss caused by urbanization and forest fragmentation are causes of bee species loss [10]; in Mexico, for instance, deforestation has increased in recent decades, reaching estimates of up to 746,000 ha per year for jungles and forests [11]. Therefore, the effects of habitat alterations are greater in smaller, isolated fragments, and are magnified over time [12], compared to large fragments which support greater diversity of bee species and can be used for the conservation of these pollinators [13,14,15].

Habitat loss caused by land use change can result in the formation of isolated areas that retain their original vegetation. Although these areas have not been significantly affected by human activities, they show significant changes in their structure and composition. In fragments, it is common to observe a decrease in both species’ richness and abundance [16].

This phenomenon affects pollinators, as the loss of plant species reduces the amount of food and nesting resources. In addition, fragments negatively affect connectivity between areas that still have resources, which causes a decline in habitat quality [17]. On the other hand, the areas bordering the fragments cause changes in the microenvironmental conditions of these areas. This phenomenon is known as the edge effect. Consequently, alterations in the environment can negatively impact species that depend on more stable and less modified conditions [18].

In particular, in highly disturbed habitats, pollinator–plant networks undergo several changes, including a decline in the frequency of interactions between both groups [19] and significant loss of unique pollinator–plant interactions, resulting in smaller, simpler, and less specialized networks with a predominance of generalist species [20,21]. Furthermore, it has been observed that bees are highly susceptible to extinction in fragmented habitats, and the species that remain are compelled to modify their relationships with plants. This finding suggests that pollinator–plant networks are highly vulnerable, as documented by the observation that, in areas of fragmentation, the percentage of original interactions declines. The network’s structural characteristics have been demonstrated to impact the resilience of communities and the functioning of the ecosystem in the context of environmental change. Consequently, the preservation of pollinators and their interactions is imperative for the maintenance of biodiversity within a particular ecosystem [22].

The study was conducted in an oak forest that had been divided into sections. The larger segment of the site is now designated as a protected natural area, while the smaller segment has been isolated by urbanization. Initially, the floristic composition exhibited a high degree of similarity, and most plant species remained shared between the two regions.

This scenario offers a valuable opportunity to assess the effects of habitat fragmentation on bee communities and their interactions with plants, thereby facilitating a more profound understanding of the impact of this disturbance on the ecological dynamics between bees and plants. The findings of this research could be important to the formulation of management and conservation strategies for fragmented habitats in urban areas, with the potential to mitigate adverse effects on bee communities.

The objectives of this study were as follows:

(1)

Evaluate the effects of habitat fragmentation on bee abundance, richness, and α- and β-diversity in a protected and a fragmented area.

(2)

Compare the structure of plant–bee interactions between both conditions.

(3)

Determine bee species responses to fragmentation based on their feeding niche breadth and overlap.

2. Materials and Methods

2.1. Study Area

The Nixticuil–San Esteban–El Diente Natural Protected Area and adjacent urban zone are situated in the municipality of Zapopan, Jalisco, Mexico. The site’s geographic coordinates are 20°46′ N and 103°24′ W, and its area encompasses 1590 ha. The study area is in the physiographic region of the Transmexican Volcanic Belt, in proximity to the border with the Sierra Madre Occidental, and has an average altitude of 1600 m above sea level [23,24].

The area’s plant cover is characterized by a variety of vegetation types, including tropical deciduous forest, oak forest, pine forest, gallery forest, and aquatic vegetation. At present, the study area is adjacent to the Metropolitan Area of Guadalajara (Figure 1), in which approximately 5,200,000 inhabitants live [25]. This proximity has resulted in the loss and fragmentation of nearby forest areas due to urban growth [24].

2.2. Data Collection

Two distinct sample areas were established for the purposes of this study. One of these areas was a conserved area of oak forest located within El Nixticuil−San Esteban−×El Diente Natural Protected Area (NA). The second was an area of fragmented oak forest (FF) with a similar floral composition. In fact, this area was in the past connected to NA, but is now approximately 1.5 km away, separated from the natural area by urban housing development. The FF is estimated to be approximately 30 ha. Despite its degraded state, the site exhibits distinctive elements of an oak forest ecosystem, including Quercus resinosa trees and a diverse array of herbaceous flora.

Bee sampling was conducted over a period of six months, started in July and finished in December 2022. The collection of samples in the study areas was conducted during the rainy season, which coincided both with the bloom of numerous plant species and the period of greatest adult bee emergence. A sampling day was realized monthly, dedicating one sampling day to each condition (NA and FF). This ensured that all samplings were carried out under conditions that were comparable (sunny days) to minimize variations attributable to weather parameters.

For each designated study area, three plots measuring 10 × 10 m² were randomly selected and separated by a distance of at least 200 m. Each plot was used during the entire study period. Afterward, the different bee species, their abundance, and the frequency of plant species visited by the bees were recorded. The data collection process was executed monthly, with a duration of one hour per plot. Sampling was carried out on consecutive days. The insects were collected between 10:00 and 15:00 h due to this corresponding to the period of maximum activity for insects. The bees were captured using an aerial entomological net while they were visiting the flowers. In each plot, two individuals participated in the collection of specimens. Bee specimens were processed in accordance with the standard methodology for preservation [26] and identified to the lowest possible taxonomic level using specialized data from the literature. Individuals that could not be identified to a species were grouped into morphospecies.

The species of plants were identified in their natural habitat. Only plants that could not be accurately identified were collected and determined by a specialist.

2.3. Data Analysis

The records for the three plots were combined by condition for data analysis. Bee abundance among the different conditions was compared with a Mann–Whitney test, with two approaches considered: (1) including all species and (2) excluding Apis mellifera to exclusively evaluate native bee species. Conversely, the total abundances of the various bee families were analyzed using the chi-square goodness-of-fit test for proportions, both implemented with R software version 4.4.3 [27].

To compare the richness of bees by condition (q0) and diversity based on the effective number of common (q1, exponential of Shannon index) and dominant (q2, inverse of Gini-Simpson index) species of bees, rarefaction/extrapolation curves were used with bee abundances based on samples with equal completeness with 95% confidence intervals according to the bootstrap method [28], using R software and the iNEXT package version 3.0.2 [29].

This procedure was used to calculate the completeness of plant pollinator interactions for each condition to analyze the parameters of mutualistic network.

The β-diversity of bees and plants between the two different conditions was analyzed using incidence data [30], where Sorensen’s dissimilarity index (ꞵSor) was decomposed into species turnover (ꞵSim) and nesting components (ꞵSne) using the vegan software vesion 2.6-10 [31].

Data collected from sampling, a quantitative bee–plant interaction matrix was constructed with the number of times each species pair was observed interacting and “0” in cases where no interaction was recorded. The metrics of the bipartite network and its graph were elaborated with the R-bipartite package version 2.21 [32].

The following indices were estimated for the bee–plant interaction network: (i) size of the network (the richness of bee and plant species); (ii) number of interactions (total number of interactions recorded) [33]; (iii) network asymmetry (the balance between floral visitors and plants from the deviation of the 1:1 ratio), where negative values close to −1 indicate more plant species than animals, and positive values close to 1 indicate more animals than plants [34]; (iv) connectivity (proportion of interactions observed with respect to the total number of possible combinations) [35]; and (v) nestedness, whose metric is defined by the formula N = 100 − T/100, where T represents the degree of disorder in the matrix, also referred to as “Temperature”. This temperature is a comparative measure calculated with respect to a perfectly nested model, as described in previous studies [36,37]. Values of N close to 1 indicate strong nesting patterns within the matrix, suggesting a highly ordered structure. In contrast, intermediate values of N or values approaching 0 imply that the matrix’s organization may be more random than nested [38].

The Shannon–Weiner index for interactions was also estimated. H2′, which is a measure of network specialization, is based on the deviation of the number of interactions recorded for a species and expected in relation to the total number of species [34]; the value varies between 0, with no specialization in interactions, and 1, with full specialization [39].

The calculation of the β-diversity of interactions (β_WN) and their components, (β_ST) and (β_OS), was performed using the index of dissimilarity of Sorensen [40,41].

For niche overlap, which is calculated as the mean similarity between interactions at the same trophic level based on the Horn–Morisita index, values close to 0 indicate different niche use and values close to 1 indicate food niche overlap [39]. The chi-square goodness-of-fit test for proportions was used to determine if there were significant differences in bee–plant interaction network metrics between the two conditions.

To compare the overlap of the metrics of the interaction network between conditions, we used the specieslevel and networklevel functions of the bipartite package in R. Because the interaction numbers differed between networks, we applied a nonparametric bootstrap procedure following the approach of [42]. The interactions of each network were resampled with replacement until they matched the total number of interactions of the site with the lowest number of interactions (Nmin). For each replicate, the interaction matrix was reconstructed, and the indices were calculated. This procedure was repeated 100 times per condition to obtain the bootstrap distribution, the mean, and the 95% confidence interval. Differences between sites were evaluated by comparing confidence intervals.

To evaluate the use of food resources by bee species shared between the NA and the FF, Levins’ niche breadth (B) was calculated. Maximum values of B are reached when the abundance of individuals is equal among the resources used, while minimum values occur when a species feeds on only one resource. Therefore, the range of B varies from 1 to n, representing the number of resources used. In addition, the overlap of feeding niches was estimated using the Horn–Morisita index under the described conditions to analyze possible changes in the diets of species shared between both localities [43].

3. Results

3.1. Abundance

The study revealed no statistically significant differences between the natural area (NA), comprising 695 individuals (sample median = 25.5), and the fragmented forest (FF) with 859 specimens of wild bees and Apis mellifera Linnaeus, 1758 (sample median = 19.5) (W = 148.5, df = 1, p = 0.68). However, a significant decline in the population of wild bees excluding A. mellifera was observed (W = 45, df = 1, p = 0.0002), representing a 74% decrease from 454 individuals (sample median = 22) in the NA to 118 in the FF (sample median = 4.5) (Figure 2).

It was observed that there was a significant difference in abundance among bee families when comparing the NA and FF. The Apidae family exhibited a higher number of specimens in the FF. However, this relationship was inverted if A. mellifera was left out of the analysis. Apidae specimens were significantly more abundant ($x^2$ = 114.63, df = 1, p = <0.05) in the NA (202) vs. FF (75). Moreover, the abundance decreased significantly in the remaining bee families or even disappeared completely, such as in the case of Colletidae (

Table 1. Abundance of bees recorded in both conditions, natural area (NA) and fragmented forest (FF), degrees of freedom (df), proportional binomial test statistic of chi-squared proportions $\left(x^2\right)$ and probability values (p).

Bee Family	NA	FF	df	${\mathit{x}}^2$	p
Apidae	439	806	1	215.19	<0.05
Apidae *	202	75	1	114.63	<0.05
Megachilidae	97	31	1	66.01	<0.05
Halictidae	56	4	1	86.7	<0.05
Andrenidae	85	18	1	84.58	<0.05
Colletidae	18	0	1	32.1	<0.05

* Only native bee species are included, excluding Apis mellifera.

). In addition, the species with the highest abundance in the NA were A. mellifera, Melitoma marginella (Cresson, 1872), Melissodes tepaneca Cresson, 1878, Andrena sp. 4, and Megachile flavihirsuta Mitchell, 1930. In the FF, the following were recorded: A. mellifera, Centris nitida Smith, 1874, and Thygater montezuma.

3.2. Diversity α y β

In the NA, a total of 94 bee species from 36 genera and 5 families were registered. Conversely, within the FF, 28 bee species across 18 genera and 4 families were documented, indicating a reduction of 70% in richness. The interpolation/extrapolation curves demonstrated marked discrepancies in species numbers between NA, with an estimated richness of 134 species, and FF, with an estimated richness of 37 species (Figure 3).

The richness per family in NA was as follows: Apidae (33 species), Megachilidae (22 species), Halictidae (20 species), Andrenidae (15 species), and Colletidae (4 species). On the other hand, in the FF, bee families had reduced richness: Apidae (12 species), Megachilidae (8 species), Halictidae (4 species), Andrenidae (4 species), and especially the Colletidae species, which disappeared in this condition.

The diversity of bees, measured by species with average abundance (q1), exhibited statistically significant differences across varying conditions. Furthermore, the NA had eight species and the FF only three species, as evidenced by the non-overlapping confidence intervals. A similar result was observed in the diversity of species, which was measured with high abundance (q2), which decreased from three species in the NA to two in the FF. Consequently, there was a substantial decline in bee diversity in the fragmented habitat relative to the protected natural area (Figure 3).

Significantly, in this study, 66 bee species were exclusive to the NA and only 1 bee species (Melitoma segmentaria (Fabricius, 1804) was exclusive to the FF. It is important to point out that 27 bee species were found in both conditions (shared species) (Figure 4). The beta diversity of Sorensen dissimilarity was (βSor = 0.55). The differences in beta diversity were mainly due to species assembly nesting (βSne = 0.52), which was attributed to the loss of bee species due to habitat fragmentation in relation to the natural area. On the other hand, there was very low species turnover (βSim = 0.03).

Alternatively, the plant species that attracted the most bees to their flowers in the NA were Cosmos sulphureus (181 specimens), Cosmos bipinnatus (131 specimens), and Ipomoea orizabesis (96 specimens) (Figure 5). In contrast, in the FF, the plants with the highest abundance of bees were Cosmos sulphureus (376 individuals), Salvia veronicifolia (154 individuals), and Tithonia tubaeformis (124 individuals). (Figure 6). The plants with the highest number of bee species as floral visitors were Cosmos sulphureus (34 species), Cosmos bipinnatus (21 species), and Bidens pilosa (20 species) (Figure 5). In FF, the plants with the greatest bee richness were Cosmos sulphureus (nine species), Ipomoea tyrianthina (eight species), and Salvia veronicifolia (six species) (Figure 6). In addition, the bees that visited the greatest number of plant species in the NA were Megachile albitarsis Cresson, 1872 (10 species) and Melissodes tepaneca and Exomalopsis moesta Cresson, 1875 (9 species each) (Figure 5). In contrast, in the FF, the top three were Apis mellifera (11 species), Centris nitida (4 species), and Melissodes tepaneca (3 species) (Figure 6).

The analysis of completeness in the fragmented area, employing rarefaction/extrapolation curves of the bee–plant interaction network, produced a high completeness of 0.98, while in the natural area, the sampling efficiency was lower, with a completeness of 0.80.

The availability of food resources (i.e., plant species) for bees was reduced by 46% in the FF, and the number of bee–plant interactions recorded was lower in relation to the total possible number. However, an increase in connectance was observed in the FF. Asymmetry was also observed with statistical differences in the network between the NA and FF conditions, with a greater variety of bee species than plant species. In the NA, 41% more bee species were recorded than plant species, whereas in the FF, this ratio was halved, reflecting a greater loss of bee species than plant species due to habitat fragmentation (

Table 2. Values of the bee–plant interaction network indices under different conditions: natural area (NA) and fragmented forest (FF).

	NA	FF
Observed bee richness	94	28
Observed plant richness	39	18
Observed connectance	6.02	9.92
Estimated connectance	5.9 (5.3–6.4)	10 (9.1–11.6)
Observed web asymmetry	0.41	0.21
Estimated web asymmetry	0.38 (0.33–0.42)	0.16 (0.10–0.22)
Observed nestedness	3.79	10.08
Estimated nestedness	4.47 (3.68–5.66)	8.78 (7.02–10.69)
Observed interaction Shannon diversity	4.22 (68)	2.27 (10)
Estimated interaction Shannon diversity	4.04 (3.9–4.19)	2.24 (2.13–2.38)
Observed interaction evenness	0.51	0.36
Estimated interaction evenness	0.51 (0.49–0.52)	0.37 (0.35–0.39)
Observed interaction strength asymmetry	0.10	0.07
Estimated interaction strength asymmetry	0.17 (0.13–0.21)	0.07 (−0.04–0.17)
Observed specialization asymmetry	−0.05	0.36
Estimated specialization asymmetry	−0.06 (−0.09–−0.03)	0.36 (0.24–0.45)
Observed H2′	0.57	0.49
Estimated H2′	0.63 (0.61–0.66)	0.49(0.42–0.56)
Observed bee niche overlap	0.13	0.14
Estimated bee niche overlap	0.12 (0.09–0.13)	0.15 (0.12–0.19)
Observed plant niche overlap	0.05	0.33
Estimated plant niche overlap	0.05 (0.04–0.06)	0.36 (0.30–0.45)

).

Two interaction networks were nested, although a difference in nesting was observed where nesting in the FF decreased. Furthermore, interaction diversity was significantly greater in the NA, where 68 common bee–plant interactions were recorded ($e^{4.22}=68$), compared to only 10 common interactions observed in the FF ($e^{2.27}=10$). This decrease is supported by interaction evenness, where nearly 50% of the maximum diversity of interactions was recorded in the NA, in contrast to a significantly lower percentage in the FF (Table 2).

Apart from that, there were differences in interaction strength due to habitat fragmentation. The asymmetry of specialization indicates that, in the NA, bees were more specialized than plants, whereas in the FF, plants were more specialized than bees.

Moreover, the average niche overlap for bees was low in both conditions, indicating that they tended to distribute the available floral resources similarly in both areas. In contrast to this, the niche overlap of plants was very low in the NA; however, this increased significantly in the FF (Table 2).

The dissimilarity of interactions (β_WN) between the two conditions was estimated at 0.89; the difference explained by species composition (β_ST) was 0.63; and the dissimilarity due to the reconnection of shared species (β_OS) was 0.26. Therefore, drastic changes in species, as well as an almost-total replacement of interactions between bees and plants, were observed. These changes were primarily caused by the loss of species, as well as by changes in the diets of some bees in the FF.

On the other hand, the H2′ index reveals that both conditions showed significant differences on levels of specialization in their networks. When the network graphs were compared between area conditions, species that tended to disappear in the FF were more specialized and were in the middle or lower part of the interaction network (see Figure 5 and Figure 6).

Furthermore, there were significant differences in Levins’ niche breadth among bee species shared between the NA and the FF area (W = 145, df = 1, p = 0.0007). The median niche breadth was significantly higher in the NA (median = 2.27 plant species) than in the FF (median = 1 plant species) (Figure 7). This indicates a significant reduction in the bees’ diet in the FF, which could be related to increased competition and forced specialization due to reductions in food resources caused by habitat disturbance (

Table 3. Values of Levin’s niche breadth index (B) and Morisita–Horn’s niche overlap index for bee species found in different conditions: protected natural area (NA) and fragmented oak forest (FF).

Species	B NA	B FF	Morisita–Horn
Thygater montezuma	1	1	0
Peponapis azteca *	1	1	0
Pseudoaugochlora gramminea	3	1	0
Augochlora aurifera	4.76	1	0
Augochlora quiringuensis	1	1	0
Protandrena sp. 1	1.75	1	0.1157
Lasioglossum (D.) sp. 7	6.23	1	0.1914
Anthidium maculifrons	3.77	1	0.2258
Megachile albitarsis	9	1.47	0.2528
Megachile (L.) sp. 2	4.23	1	0.2696
Exomalopsis arida	1.8	2	0.3157
Apis mellifera	2.81	3.49	0.3174
Centris atripes	5	1	0.3333
Heriades sp. 1	1.8	1	0.4285
Centris nitida	1.92	3.12	0.4908
Anthophora squammulosa	3	1	0.5
Melitoma marginella *	1.07	2.8	0.5244
Melissodes sp. 1	3.11	1.8	0.6338
Megachile petulans	2	1	0.6666
Calliopsis hondurasica	2.28	2	0.6666
Heriades sp. 2	1	2	0.67
Melissodes tepaneca	5.53	2.57	0.6829
Protandrena sp. 2	2.27	1	0.8333
Megachile flavihirsuta	1.39	1.38	0.9748
Ashmeadiella sp. 1	1	1	1
Pseudopanurgus sp. 1	1	1	1
Peponapis utahensis *	1	1	1

* Indicates bees with oligolectic habits and + indicates higher niche breadth values.

).

The overlap in feeding niches tended to be medium to low between the two conditions. For example, of the shared species, 22 showed values between 0.68 and 0, which indicates a high degree of flexibility in changing food resources. Also, five species had food niches with high overlaps between the two locations, in addition to one species of oligolectic bee present in the FF, Peponapis. The presence of these species is due to the fact that certain plant species on which they strictly feed prevail in this area. However, species with low diet overlap, which can use different resources in the fragmented area, are predominant, which could be a response to changes in competition or food resource availability following habitat disturbance (Table 3).

4. Discussion

Our study revealed a substantial decline of 74% in abundance and 70% in richness of native bee species in the FF. Also, only six species increased in abundance in the FF, and diversity decreased significantly too. Dissimilarity between communities was associated with species loss in the FF. These patterns show the usual result of reduced habitat quality in fragmented landscapes, where declines in structural complexity, degradation of soil conditions, and simplification of floral communities limit the capacity of ecosystems to support pollinator diversity [44,45].

Changes in the structure of the bee–plant interaction network were observed, including variations in size, asymmetry, interaction diversity, specialization, and niche overlap in plants, as well as almost-total replacement of interactions. Similarly, a reduction in the food niches and changes in the diets of the bee species present in the FF were observed compared to the NA. This suggests possible effects of habitat fragmentation on pollination dynamics.

4.1. Bee Community Structure

These results are consistent with previous studies detailing the negative effects of habitat fragmentation on the abundance, richness, and diversity of bees by reducing the size and connectivity of fragments [46,47,48]. However, in some cases, no decrease in bee abundance was observed, with certain species even increasing in number in these habitats [49].

The decline in bee abundance in fragmented habitats has been linked to limited floral resources for foraging and reduced availability of suitable nesting sites [49]. This is consistent with the significant decline in plant species diversity and number recorded in the fragmented area of our study [50]. Hence, this suggests that the main factors determining bees’ responses to fragmentation are their feeding traits and diet breadth.

In the present study, A. mellifera increased in abundance in the FF, while families such as Colletidae, Andrenidae, Halictidae, and Megachilidae showed reductions or disappeared locally. This pattern can be explained by these families’ specialization in narrow feeding niches or nesting requirements [51,52].

In addition, these families also face competition for resources, particularly with A. mellifera, which can monopolize nectar and pollen, thereby limiting availability for native species [53]. This pressure, coupled with reduced floral diversity, affects community richness and abundance, as well as the stability of pollination networks.

In fact, Colletidae, Andrenidae, and Halictidae are short-tongued bees that are closely associated with small, open-corolla flowers, from which they obtain their food. The reduction in floral resources in fragmented areas could particularly affect these groups [51]. Furthermore, Colletidae and Andrenidae build their nests in the ground, and previous studies have indicated that these bees are especially sensitive to reductions in fragment size [21].

The Megachilidae family has specific requirements for nest construction, which are different depending on the species. While most species build nests in pre-existing cavities, a few build their nests in the ground. However, the construction of their nests depends on resources, such as leaves, petals, trichomes, fibers, and plant resins, as well as water mixed with mud [52]; as a result, reducing the availability of these materials in fragmented areas could negatively affect their reproductive cycle.

High-quality habitats usually support a wider variety of flowers and places for bees to build their nests. This allows different types of bees to live together. On the other hand, pieces of land that are in poor condition and have a lot of plants along the edges tend to support bees that can handle disturbances. These bees also eliminate species that need stable, small habitats inside the land [54]. The growth of A. mellifera is made easier by the fact that it can adapt to many different flowers and is good at using resources efficiently. This makes it hard for native species to survive when there are not many different resources available.

Fragmentation not only affects niche breadth but also the spatial scale of habitat use. The foraging range of bees is primarily determined by body size and sociability, which in turn influence their ability to detect and exploit resources. Highly eusocial species such as A. mellifera have been shown to possess complex communication strategies and form colonies comprising thousands of individuals. These species are known to forage over considerable distances, ranging from 800 m to 2 km. In contrast, solitary species have more restricted ranges of 100 to 300 m [55], as is the case for most species in this study. In fragmented habitats where species and food resources are declining, such foraging limitations could affect reproductive success and larval provisioning, thereby contributing to the observed decline in populations.

Ecosystem fragmentation isolates populations and simplifies landscape structure, leading to reductions in population size and leading to changes at the community level, such as decreases in pollinating insect richness [22]. This is because fragmented habitats cannot support high pollinator diversity due to high local extinction rates caused by smaller populations, limited colonization due to a lack of habitat connectivity, and stochastic demographic processes [56]. The results from this study support the idea that there is a significant decline in bee diversity due to a reduction in dominant species with medium abundances and the disappearance of a large number of rare species.

The effects of fragmentation on bee beta diversity are limited. While the conversion of natural habitats for agricultural or urban use has been shown to decrease alpha diversity, a review of 157 global bee studies found no significant differences in beta diversity between natural and urban or agricultural sites [57]. However, the authors note that the impact of fragmentation on beta diversity varies, emphasizing the importance of assessing these effects in different land use contexts.

In our study, beta diversity revealed changes because most of the species that disappeared from the fragmented area were not replaced by new ones. Thus, this suggests that the bee community became impoverished, with only the most abundant and disturbance-tolerant species remaining, while rare or sensitive species disappeared. This pattern has also been documented in fragments of different sizes [13] and in urban parks near the study area [58], where species loss has not been offset by recolonization processes.

The lack of replacement may be related to a reduction in floral resources and nesting sites, as well as increased competition with opportunistic species. Furthermore, the simplification of plant communities could limit the persistence of plant–pollinator interactions; therefore, changes in plant composition and distribution affect the behavior and diversity of pollinators [59,60].

Agricultural areas with greater plant diversity have been observed to be home to more diverse bee communities [61], whereas urbanization tends to reduce both plant and associated bee richness [58]. In our study, fragmentation was associated with a 46% reduction in the richness of plants visited, which supports the hypothesis that the loss of key resources can trigger local extinctions in pollination networks.

4.2. Network Metrics

The observed connectivity was low in both locations, which is typical of mutualistic networks due to morphological and phenological constraints, as well as network size [62,63,64]. However, an increase in connectance was recorded in the FF; this was statistically significant and could be attributed to the loss of rare species and their interactions [14]. Furthermore, habitat fragmentation favors more generalist species, which interact with many plants, increasing the number of links within the networks [65,66].

The increase in connectance in the fragmented area probably arose as a consequence of the loss of floral resources, the reduction in bee richness, and the specialization of interactions [67], resulting in a bee community composed mainly of generalist species, which are highly connected in the network, leading to a high dependence on a small group of species and vulnerability of the pollination system [68].

Asymmetry was significantly reduced in the fragmented area. This suggests a greater loss of bees, particularly generalist species, relative to plants. This may reflect the greater vulnerability of bees to fragmentation due to their dependence on floral resources and limited foraging capacity in small areas or the local disappearance of species with very specific ecological requirements [59,60]. Subsequently, this could compromise the long-term stability and functioning of the ecosystem [63], although network asymmetry properties, such as functional redundancy, act as a buffer against the loss of bees resulting from habitat fragmentation [60].

Nesting patterns remained the same in both locations, suggesting structural stability in the network, possibly due to the persistence of generalist or polylectic species sustaining multiple interactions. However, the loss of specialists or species with few interactions in the study area drastically reduced the total number of common interactions by 85%, making the network more homogeneous and simplified. In fact, this trend has also been reported in other studies [65,69].

This result coincides with reports that habitat loss causes a reduction in floral resources and bee species richness but favors generalist species which increase nesting by interacting with diverse plant species [65,67]. The increase in nesting in the fragmented area may be a consequence of the simplification of the network of interactions and a high dominance of generalist interactions [68].

Interaction evenness was significantly reduced in the FF, reflecting the fact that most interactions were concentrated among a few species, thereby increasing the system’s functional dependence on a small set of organisms. This phenomenon is associated with the simplification of network structures in the face of disturbances [37].

The asymmetry in interaction strength was significantly greater in the natural area than in the fragmented area. This was probably because the natural area had more diverse plant and pollinator communities, which facilitated differences in specialization and in the intensity of visits and dependence on interactions, so the natural area had more heterogeneously distributed interactions and specialized pollinators with strong links to some plants [70,71].

However, asymmetry in specialization showed changes: in the NA, bees were more specialized than plants, whereas in the FF, the opposite occurred. This indicates a functional reorganization resulting from the loss of diversity. This difference could be explained by the drastic reduction in bee diversity in the FF, since plants depend on a small number of generalist bees for pollen dispersal [72].

Network specialization (H2′) was significantly higher in the natural area than in the fragmented area, probably due to greater plant and bee richness, which encourages greater food niche partitioning and the presence of species with varying degrees of specialization [73,74]. On the other hand, the reduction in H2′ in the fragmented area suggests a reduction in network specialization because habitat loss causes a decrease in food resources for bees, which leads to the disappearance of specialist or oligolectic bee species [65].

The average niche overlap between bees from both locations was low and not significant, suggesting a distribution of floral resources that avoided direct competition. However, pollinator overlap in plants was greater in the FF, where plants competed more intensely for pollinators. This was related to the reduction in bee diversity, as fragmentation has been reported to limit pollinator availability and affect their relationships with plants [74].

Dissimilarity analysis of interactions showed an almost-total change in network composition and interactions (β_WN = 0.89), indicating reconfiguration of the network. This was mainly explained by loss of species (β_ST = 0.63) and to a lesser extent by rearrangement of shared species (β_OS = 0.26), possibly due to reduced food resources.

4.3. Niche Breadth and Overlap

The analysis of Levins’ niche breadth of the bee species found at both sites revealed significant differences, with the natural area exhibiting greater breadth. This suggests that fragmentation reduced the variety of floral resources available to bees, which could have increased interspecific competition [60,75].

In high-quality habitats, floral diversity supports broader niche breadths, facilitating resource partitioning and coexistence. In contrast, edge-dominated habitats experience temporal and spatial bottlenecks in floral availability, forcing bees to converge on a reduced set of species and thereby increasing the probability of competitive exclusion [46]. These resource constraints can also impair reproductive success and larval provisioning, further contributing to population declines.

The overlap in the feeding niches of species present at both sites showed high variability, ranging from 0 to 0.68 for most species, with a few having a value close to 1. This indicates that, while some species maintained similar diets in both locations, most modified their feeding habits. This suggests that species can adjust their diet to survive in disturbed environments [73,74].

Therefore, to conserve bee species in the area, it is recommended that conservation strategies prioritizing the connectivity of forested areas are established and implemented. Additionally, the conservation of the protected natural area is crucial, as it supports a high diversity and abundance of bee species and food resources. Restoration of disturbed habitats and the creation of biological corridors with herbaceous plants visited by bees are also essential to promote the conservation of these insects and the ecosystem services they provide.