The associations between floral traits and pollinator groups in two Mediterranean mountainous plant communities in the Middle Atlas of Morocco

Abstract

Background

The relationships between floral traits and pollinators have been extensively studied over the last few decades. The concept of pollination syndrome suggests that plants pollinated by the same group of pollinators tend to develop similar combinations of floral traits. However, several studies have demonstrated the low predictability of these trait combinations and found high levels of pollination generalization within plant communities. In this context, on the basis of direct field observations, our study aimed to explore the relationships between pollinator functional groups and floral traits in two Mediterranean mountainous plant communities in the Middle Atlas of Morocco.

Results

Our research included 83 plant species belonging to two plant communities and 10 insect groups. Pollination generalization (PG) levels varied significantly between the two plant communities. Among the thirteen floral traits included in the present study, PG levels were significantly associated with only flower clustering and reward type. Clustered flowers and those producing both nectar and pollen were more generalized than solitary flowers and those producing only pollen. Bumblebees frequently visit closed zygomorphic flowers with hidden anthers in a vertical orientation, whereas flies, hoverflies, and ants prefer open actinomorphic flowers with exposed anthers in a horizontal orientation. Butterflies and beeflies were strongly associated with tubular flowers with long corollas and pink colouration. In addition to being the most abundant group, bees and beetles exhibited generalist behavior, with associations with multiple floral traits in both communities.

Conclusions

Our results, while offering partial support for the pollination syndrome hypothesis, revealed predominantly generalized patterns of pollination among the studied plant species. Notably, these patterns were not significantly explained by all the floral traits we analysed. Furthermore, we suggest that within insect groups, especially bees and beetles, species-level variation in floral trait preferences may exist, contributing to the observed generalizability. These findings highlight the need for further research to unravel the intricacies of plant–pollinator interactions and to better understand the mechanisms driving network structure and specialization.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12862-025-02403-w.

Background

Pollination, the transfer of pollen from anther to stigma, is a crucial ecosystem function that ensures both natural plant reproduction and agricultural productivity [1, 2]. Recent estimates indicate that biotic pollination, which is primarily mediated by insects, occurs in 90% of flowering plants [3]. Despite this fundamental importance, our understanding of how plant‒pollinator relationships vary across different ecological contexts remains incomplete, particularly in biodiversity hotspots where these interactions are highly complex.

The relationship between plants and their pollinators represents one of nature’s most studied examples of coevolution, a process spanning 200–300 million years [4]. Through natural selection, plants have evolved diverse floral traits that attract pollinators, including rewards (nectar and pollen), specific colors, and distinct symmetry patterns. Accordingly, insect pollinators have developed specialized adaptations, such as specialized mouthparts, pollen-collecting structures (e.g., scopa), and enhanced sensory abilities for detecting floral signals [5, 6].

Scientists have traditionally used the concept of pollination syndromes—suites of floral traits that have convergently evolved to attract specific pollinator groups—to understand these interactions [7, 8]. However, the predictive power of pollination syndromes has become increasingly debated. While some studies support their validity [9, 10], others have challenged this framework by demonstrating high levels of ecological generalization in plant‒pollinator interactions [11–13]. This debate has highlighted the need to examine how plant‒pollinator relationships vary across different ecological contexts.

The concept of pollination generalization—the diversity of pollinator groups visiting a plant species—has emerged as a key framework for understanding plant‒pollinator interactions [14, 15]. The degree of generalization can vary substantially among plant species and communities and is influenced by both floral traits and the ecological context [16, 17]. At the community level, understanding these generalization patterns is crucial for predicting ecosystem stability and resilience to environmental change [18, 19]. However, we still lack a comprehensive understanding of how generalization patterns vary across different habitat types within the same biogeographic region.

The Mediterranean Basin, particularly Morocco, provides an ideal natural laboratory for investigating these questions. Morocco hosts the second-highest terrestrial biodiversity in the Mediterranean region [20], with exceptional diversity in both plants and pollinators [21–24]. The rich bee fauna of the country is especially noteworthy [23, 25]. Despite this remarkable diversity, studies of plant‒pollinator interactions in Moroccan ecosystems remain surprisingly rare, representing a significant gap in our understanding of Mediterranean pollination ecology, especially in natural ecosystems. Although recent efforts have begun to address this gap through studies of wild pollinators in Moroccan agrosystems [22] and coastal forests of western Morocco [26], much remains to be explored across the country’s diverse ecological landscapes.

The Middle Atlas region of Morocco, with its mosaic of distinct habitat types, offers a unique opportunity to examine how plant‒pollinator relationships vary across contrasting ecological contexts. We selected two representative plant communities for this study: an open canopy cedar forest (Moudmam) and an open area without tree cover (Ras el Ma). These sites, while sharing similar climatic conditions and biogeographic histories, represent distinct habitat types with different plant communities and environmental conditions.

Our study aimed to investigate the associations between 13 floral traits and 10 pollinator functional groups in these plant communities. Specifically, we address three specific objectives: (1) to characterize the floristic diversity of these contrasting plant communities, with particular emphasis on entomophilous flowering plants; (2) to examine how floral traits are related to pollination generalization levels across these different community contexts; and (3) to test the predictive power of pollination syndromes by analysing the relationships between floral traits and pollinator group preferences. By comparing these patterns across contrasting habitats, we can assess the consistency of plant‒pollinator relationships and evaluate the context dependency of pollination syndromes.

We hypothesized that flower structure and color influence pollinator diversity. Specifically, we predicted that open flowers with pale colors (white and yellow) would attract a wider variety of pollinator groups (flies, beetles, ants, etc.), demonstrating greater ecological generalization. In contrast, we expected that closed flowers with bright colors (purple and pink) would attract primarily bees, resulting in lower ecological generalization [7, 27].

This research contributes to both theoretical and applied aspects of pollination ecology. Theoretically, it will help resolve ongoing debates about the validity of pollination syndromes and the drivers of generalization in plant–pollinator networks. From an applied perspective, understanding how habitat type influences plant‒pollinator relationships will inform conservation strategies in the face of ongoing habitat modification and climate change, particularly in the biodiversity-rich but understudied Mediterranean region.

Methods

Study sites

This research was conducted in two natural reserves located within the National Park of Ifrane in the Middle Atlas region of Morocco (Fig. 1). The first site, Moudmam, is situated at 33°23’57.48"N, 5°11’0.87"W, between Azrou and Timahdite and covers an area of 25 hectares. The elevation ranges from 1800 to 1880 m (msl), with a gentle slope (≤ 15°). This site comprises an open canopy cedar forest with a rich diversity of herbaceous plants. In 2003, this natural reserve was created to safeguard cedar populations, promote their natural regeneration, and provide a suitable habitat for numerous flora and fauna species, including several endemic, rare and endangered species, i.e., Anacyclus pyrethrum (Asteraceae), Argyrocitisus battandieri (Fabaceae), Lavandula maroccana, Salvia taraxacifolia (Lamiaceae), Ammotragus lervia (Bovidae), and Macaca Sylvanus (Cercopithecidae) [28].

Fig. 1.

Geographic localization of the study sites (reproduced from https://siredd.environnement.gov.ma)

The second study site, Ras el Ma, is located at 33°30’7.30"N, 5°6’47.58"W, with an elevation of 1690 m (msl) and a slope gradient of less than 15°. It covers an area of 2.84 ha and lies approximately 4 km from Ifrane city and approximately 22 km from the first study site. This area has been designated a conservation zone by the forest administration to protect its ecological integrity [28]. The site is characterized by the absence of arboreal vegetation, consisting predominantly of herbaceous flora, and is situated at the periphery of a mixed forest composed of oak (Quercus rotundifolia, Fagaceae) and cedar (Cedrus atlantica, Pinaceae).

The entire region, including both study sites, experiences a humid Mediterranean climate with an average annual rainfall of 963.69 ± 363 mm. The climate pattern includes a wet season spanning from October to May and a dry season from June to September [28].

Plant sampling

The study sites were sampled monthly from March to August 2023 to capture the main flowering period in the Middle Atlas region. We focused on entomophilous plant species, excluding wind-pollinated families such as Poaceae, due to their lack of insect-attractive floral features.

At each site, we delimited a representative continuous area (At Moudmam, an area of 11.7 ha, and at Ras el Ma, 2.1 ha) that presented high plant density and diversity and captured the typical vegetation characteristics of the habitat (At Moudmam, the sampling area was located within a cedar open-canopy forest and comprised a diverse understorey of shrubs, herbaceous plants, and grasses. In contrast, at Ras el Ma, the delimited area consisted of open grassland without tree cover, dominated primarily by herbaceous vegetation. Within these delimited areas, 2 × 2 m quadrats were randomly placed during each monthly sampling, with a minimum separation of 5 m between adjacent quadrats. Ninety quadrats at Moudmam and 35 at Ras el Ma were established. The quadrat positions were newly randomized for each sampling period to capture temporal variations in the plant community [29]. The difference in quadrat numbers between sites reflects the relative size and heterogeneity of each site, as the more structurally diverse vegetation at Moudmam required more quadrats to capture its variability, whereas the more uniform grassland at Ras el Ma required fewer quadrats.

During each monthly visit, we recorded all flowering plant species within each quadrat and counted the number of individuals per species. The survey included all plant life forms, from herbs to woody species such as Argyrocytisus battandieri (Fabaceae) and Crataegus monogyna (Rosaceae).

Species identification was conducted via the Practical Flora of Morocco [30–32], with the identification validation of Professor Madame Bari Amina, a botanist at the Faculty of Sciences Dhar el Mehraz, Fes.

Voucher samples were prepared and deposited in the herbarium of the Laboratory of Biotechnology, Conservation, and Valorization of Natural Resources (LBCVRN) at the University of Sidi Mohamed Ben Abdellah, Fes. The samples are catalogued under the name ‘MOUDMAM&RASELMA/23–24’ with the following voucher numbers: SV/M01/23/001 - SV/M01/24/046 and SV/R01/23/001 - SV/R01/24/052.

Floral trait description and measurement

For each plant species, 13 floral traits were measured and described. Floral traits included flower clustering (solitary flowers/inflorescences), flower orientation (horizontal/vertical), floral form (open flowers/closed flowers, including semi-closed flowers whose petals partially enclose the reproductive structures, but the flowers are not fully open or fully closed), ovary position (hypogynous/epigynous), flower symmetry (actinomorphic/zygomorphic, including zygomorphic flowers in actinomorphic inflorescences), color as perceived by humans (yellow, white, purple, pink, blue, green, and red), flower shape (including the following categories: bell, bowl, cross, cyathium, disk, funnel, pea, incushion, snapdragon, star, trumpet, tubular, two-lipped, wheel, urn, spurred, and irregular; see Table 1 below for shape descriptions), flower height above the ground (cm), flower diameter (i.e., the length between petals growing in different directions) (mm), corolla tube length (mm), anther position (exposed/hidden, including the semi-hidden, where the anther is partially enclosed or covered by petals or partially inserted in a tubular corolla), flower scent (as perceived by humans, categorized into four categories: Strong/moderate/low/absent) and type of reward (only pollen/pollen and nectar).

Table 1.

Description of flower shapes included in the study

Shape	Description
Bowl	Cup-like flowers with petals curving upwards, forming a shallow, open structure.
Disk	Composite flower heads with a central disk of small tubular florets, typical of the Asteraceae family.
Funnel	Tubular flowers that gradually widen toward the mouth, resembling a funnel. Found in Romuleacea and Boraginaceae.
Trumpet	Flowers with a narrow tube that flares abruptly into a wide, open mouth, similar to a trumpet. Seen in Narcissus sp.
Star	Radially symmetrical flowers with pointed petals radiating outwards, forming a star-like shape.
Cross	Four-petaled flowers arranged in a cross shape, as in the Brassicaceae family.
Pea	Zygomorphic flowers with a banner, wings, and keel, typical of the Fabaceae family.
Two-lipped	Bilabiate flowers with distinct upper and lower lips, often seen in the Lamiaceae family.
Snapdragon	Closed, bilabiate flowers that open when pollinators apply pressure, resembling a snapdragon. Seen In Linaria sp.
Tubular	Cylindrical flowers with a consistent diameter throughout, forming a tube-like shape.
Bell	Bell-shaped flowers with a wide base and flared opening, resembling a bell.
Urn	Urceolate flowers with a swollen base and narrow opening, resembling an urn.
Pincushion	Dense, rounded flowers with protruding styles, giving a pincushion appearance.
Wheel	Flat-faced, radially symmetrical flowers with a short tube and spreading petals, resembling a wheel. Seen in Verbascum sp.
Spurred	Flowers characterized by a tubular extension or “spur,” often projecting backwards from the corolla or petal base, typically containing nectar. Common in species like Delphinium sp.
Cyathium	A unique inflorescence resembling a single flower, characteristic of the Euphorbia genus, consisting of a cup-like structure enclosing several reduced male and female flowers.
Irregular	Asymmetrical flowers that do not fit into the regular shape categories, particularly seen in Reseda sp.

The flower height, flower diameter and corolla tube length of 20 to 30 flowers from at least 5 to 10 individuals of each species were measured via an electronic calliper. For Asteraceae, the diameter of the capitula was measured, and for zygomorphic flowers, we measured the largest dimension of the flower (e.g., in Fabaceae, from the upper limit of the banner to the lower limit of the carina). Floral scents were perceived from fully bloomed flowers from multiple samples of each plant species throughout daylight hours.

Insect survey

To explore the relationships between floral traits and various pollinator groups, we employed the observation plot method [33, 34]. Two sampling sessions were conducted over two consecutive years. In each year, 6 sampling periods were carried out at each site, from March to August, once per month during the second or third week, depending on meteorological conditions. During each period, we randomly positioned ten observation plots of 2 by 1 m within each study area, ensuring a minimum separation of 5 m between plots. Within these plots, we recorded pollinator interactions with plant species during 12-minute observation periods per plot. These observations were conducted throughout the day, from 9:00 to 17:00, under suitable meteorological conditions for insect activity, characterized by temperatures exceeding 22 °C, low wind speeds, and minimal skycover [33, 35]. A visitation event was recorded when an insect landed on the reproductive organs of a flower [36]. We classified each visitor into one of 10 pollinator functional groups, defined as insect groups exhibiting similar behaviors on flowers and generating correlations among floral traits [8, 15, 36]. These functional groups included bumblebees (Bombus sp.), honey bees (Apis mellifera), solitary bees (Andrenidae, Halictidae, Megachilidae, Apidae (Apis mellifera, and Bombus excluded), Melitidae), ants, wasps (predominantly Scoliidae, Ichneumonidae, Pompilidae, and Vespidae), muscoid flies (primarily Anthomiidae, Muscidae, Tachinidae, Asilidae, and others), hoverflies (Syrphidae), butterflies, beeflies (Bombylius sp.), and beetles (photos of some insects visiting flowers are provided in Fig. 2).

Fig. 2.

Pictures from the field of different insects visiting different flowers. a Falsomelyris granulata (Melyridae – Coleoptera) visitng Bellis selvestris (Asteraceae), b An ant – unidentified species (Formicidae - Hymenoptera) visiting Euphorbia helioscopia (Euphorbiaceae), c Scolia hortorum (Scoliidae – Hymenoptera) visiting Scabiosa ochroleuca (Caprifoliaceae), d Aethiessa floralis (Scarabaeidae - Coleopetra) visiting Argyrocytisus battandieri (Fabaceae), e A butterfliy – unidentified species (Nymphalidae – Lepidopetra) visiting Carduus nutans (Asteraceae), f Ancistrocerus sp. (Vespidae – Hymenopetra) visiting Erysimum grandiflorum (Brassicaceae), g Scaeva selenitica (Syrphidae – Dipetra) visitng Gagea liotardi (Liliaceae), h Anthophora dispar (Apidae – Hymenopetra) visiting Linaria sp. (Plantaginaceae)

Data analysis

All the statistical analyses were performed via R (version 2023.06.1; R Core Team, 2023) through RStudio (version 2023.06.1 Build 524; Posit Software, PBC).

Plant community diversity was assessed via the Shannon‒Wiener index (H’) calculated with the diversity() function from the vegan package version (2.6.4). Differences in community composition between sites were tested via permutational multivariate analysis of variance (PERMANOVA) with the adonis2() function. PERMANOVA was based on Bray‒Curtis dissimilarity matrices with 999 permutations. The same approach was applied to test for differences in pollinator assemblages between sites.

The quantitative floral traits—flower height, diameter, and corolla length—were analysed via the Wilcoxon rank-sum test to assess differences between the two study sites. For categorical floral traits (e.g., color, symmetry, shape), proportional distributions were calculated separately for each site and are reported in the text.

The pollination generalization level (PG), a measure of the diversity of pollinator groups visiting each plant species, was calculated using Simpson’s diversity index formula:

where represents the proportion of visits by pollinator group i to a plant species, which is calculated as the number of visits by pollinator group i divided by the total number of visits to that plant species. A PG value of 1 indicates that the plant species is visited by only one pollinator group. This formula accounts for both the diversity and abundance of insect groups and gives greater weight to abundant groups, which prevents the overestimation of incidental visits [37].

To assess the effects of floral traits on pollination generalization (PG), we conducted a generalized linear model (GLM) analysis via the glm function in R. Initial exploratory analysis revealed that the PG values were right skewed (skewness = 1.29), with a mean of 2.78 (range: 1.00–7.63; PG values of zero (n = 11) were excluded). Consequently, we conducted and compared three model families (Gaussian with log-transformed response, gamma with log link, and inverse Gaussian with log link) via the Akaike information criterion (AIC). The Gaussian model with log-transformed PG provided the best fit (AIC = 62.53, compared with 210.44 and 217.08 for the other models, respectively). Diagnostic plots confirmed that this model satisfied the assumptions of normally distributed residuals and homogeneity of variance (see Fig. 3 below).

Fig. 3.

Diagnostic plots for the Gaussian GLM with log-transformed PG response. Residual diagnostics confirm model assumptions: A homoscedasticity across predicted values, B normal distribution of residuals, C consistent variance across the range of predictions, and D absence of influential outliers exceeding Cook's distance thresholds

Our initial predictors included floral traits (clustering, color, height [log-transformed], diameter, symmetry, form, corolla length, ovary position, reward, and scent) and covariates (locality, total pollinator visits [log-transformed], and total plant abundance [log-transformed]). Owing to strong correlations with other traits, flower orientation and anther position were excluded, as they are highly collinear with flower symmetry and flower form, respectively. The remaining predictors had variance inflation factors (VIFs) below 1.6, indicating no concern of collinearity. Flower shape was also omitted because of its complex and highly multicategorical structure, which could lead to unstable model estimates.

To identify the most important factors influencing pollination generalization, we performed backwards stepwise selection on the basis of the AIC. This approach resulted in a simplified model retaining only four significant predictors: clustering, reward, locality and total pollinator visits. We further explored site-specific patterns by conducting separate analyses for each study site (Moudmam: n = 46; Ras el Ma: n = 41).

For the interpretation of model coefficients, we back-transformed the estimates by exponentiation (exp(estimate)) to obtain the multiplicative effect on the original PG scale, thus providing a measure of the percent change in pollination generalization associated with each predictor.

Using multivariate methods, we analysed the relationships between floral traits and pollinator preferences separately for each community and for the pooled data from both communities. Pooling data increased statistical power, allowing for the detection of broader pollinator preference patterns. Flower shape was analysed independently for its complex and multicategorical structure. First, detrended correspondence analysis (DCA) was performed via the decorana() function to determine the appropriate ordination method. The resulting gradient lengths exceeded 4 standard deviations, indicating that unimodal models were most appropriate. Consequently, canonical correspondence analysis (CCA) was conducted via the cca() function of the vegan package. The response variables were the proportional visits by each pollinator group, while floral traits served as explanatory variables. The plant species that received fewer than five visits were excluded from the analyses to prevent the overestimation of specialization [36, 38].

The significance of the CCA model was tested via permutation tests (999 permutations) at two levels: (1) a global test for overall model significance and (2) sequential tests for individual explanatory variables. The results were visualized via the ggplot2 package (version 3.4.2), with separate biplots created for both sites combined and for each site, revealing the relationships between floral traits and pollinator groups.

Results

Plant communities structure and floral characteristics

The two plant communities—Moudmam and Ras el Ma—exhibited similar levels of diversity, with Shannon‒Wiener indices (H’) of 3.126 at Moudmam and 3.314 at Ras el Ma. Permutational multivariate analysis of variance (PERMANOVA) revealed significantly different species compositions (R² = 5.13%, F_1,750 = 41.155, p = 0.001), confirming the distinct compositional differences between them. At Moudmam, 46 plant species belonging to 23 families were recorded. The Asteraceae family is represented by 13 species, followed by Caryophyllaceae with 5 species, Fabaceae with 4 species, and Brassicaceae with 3 species. The Apiaceae, Malvaceae and Liliaceae had 2 species each, while the other families had one species each (Amaryllidaceae, Boraginaceae, Campanulaceae, Caprifoliaceae, Cistaceae, Euphorbiaceae, Geraniaceae, Hypericaceae, Iridaceae, Lamiaceae, Liliaceae, Plantaginaceae, Ranunculaceae, Resedaceae, Rosaceae, and Scrophulariaceae). At Ras El Ma, 52 species from 19 families were recorded. Asteraceae accounted for 10 species, followed by Fabaceae and Lamiaceae, with 8 and 6 species, respectively. Boraginaceae, Caryophyllaceae, Cistaceae, and Ranunculaceae each had 3 species. Brassicaceae, Plantaginaceae, and Rubiaceae had 2 species each, while the remaining families were represented by only one species (Asparagaceae, Liliaceae, Apiaceae, Caprifoliaceae, Euphorbiaceae, Geraniaceae, Linaceae, Malvaceae, Rosaceae, and Scrophulariaceae). See the Supplementary Information (Supplementary Material 1) for more details.

At both study sites, we found a clear predominance of certain floral traits. Compared with those with solitary flowers, plants with inflorescences dominated, representing 86.9% and 88.64% of the species at Moudmam and Ras el Ma, respectively. Similarly, open flowers (Moudmam: 89.1%; Ras el Ma: 78.8%), actinomorphic symmetry (Moudmam: 63.04%; Ras el Ma: 53.84%), horizontal orientation (Moudmam: 76.08%; Ras el Ma: 55.7%), and exposed anthers (Moudmam: 89.13%; Ras el Ma: 67.30%) were the prevalent traits in both communities (see Table 2).

Table 2.

Percentages (%) of floral traits categories in the study sites

Floral traits		Moudmam	Ras el Ma
Clustering	Inflorescences	86.95	88.46
	Solitary flowers	13.04	11.53
Form	Open	89.13	78.84
	Closed or semi-closed	10.86	21.15
Symmetry	Actinomorphic	63.04	53.84
	Zygomorphic	36.95	46.15
Orientation	Horizontal	76.08	55.76
	Vertical	23.91	44.23
Anther position	Exposed	89.13	67.30
	Hidden or semii-hidden	10.86	32.69
Ovary position	Epigynous	39.13	28.84
	Hypogynous	60.86	71.15
Scent	Absent	56.52	48.07
	Low	10.86	15.38
	Moderate	17.39	28.84
	Strong	15.21	7.69
Reward type	Pollen and Nectar	47.82	59.61
	Pollen	52.17	40.38
Shape	Bell	2.17	0
	Bowl	15.2	17.3
	Cross	6.52	3.85
	Cyathium	2.17	1.92
	Disk	28.3	19.2
	Funnel	4.35	5.77
	Irregular	2.17	0
	Pea	8.7	15.4
	Pincushion	2.17	1.92
	Snapdragon	2.17	3.85
	Star	15.2	7.69
	Trumpet	2.17	0
	Tubular	4.35	5.77
	Two-lipped	2.17	11.5
	Wheel	2.17	1.92
	Spurred	0	1.92
	Urn	0	1.92

The flower color distribution did not significantly differ between sites (Fisher’s exact test, p = 0.757). In Moudmam, 44.68% of the plant species presented yellow flowers, followed by 23.4% with white flowers. Purple and pink flowers were displayed by 19.15% and 8.51% of the species, respectively. These proportions remained consistent when the total abundance of each flower color was considered. At Ras el Ma, yellow flowers were still the most common (34.62%), and purple and pink flowers were more prevalent than were Moudmam flowers (26.92% and 13.46%, respectively) (Table 3). In terms of flower height, Moudmam presented a marginally greater range than did Ras el Ma. The median flower height at Moudmam was 24.4 cm (SD = 47.0 cm), whereas at Ras el Ma, it was 18.2 cm (SD = 27.9 cm). However, the difference between these two locations was not statistically significant (p = 0.13) on the basis of the results of the Wilcoxon test. Flower diameter and corolla length did not significantly differ between sites (see Fig. 4). Details about the floral traits of each plant species are provided in the Supplementary Information (Supplementary Material 1).

Table 3.

Flower color distribution in the study sites (percentages of species displaying each color, and percentages of species displaying each color multiplied by their total abundance)

Color		Blue	Green	Pink	Purple	Red	White	Yellow
Moudmam	Proportion of species (%)	2.13	2.13	8.51	19.15	0.00	23.40	44.68
	Proportion * total abundance (%)	0.64	2.31	8.84	17.79	0.00	18.13	52.29
Ras El Ma	Proportion of species (%)	1.92	1.92	13.46	26.92	3.85	17.31	34.62
	Proportion * total abundance (%)	0.17	0.03	2.96	18.14	1.01	14.91	62.78

Fig. 4.

Medians and quantiles of flower height, flower diameter and corolla length of the plant species occurred in the study sites Moudmam and Ras El Ma, including the p value obtained from the Wilcoxn rank sum test

Pollinators’ functional group abundance

Overall, 15 791 visits were recorded at the two sites, with 8 901 visits at Moudmam and 6 890 visits at Ras el Ma. The composition of the pollinator groups differed significantly between the two communities, as revealed by PERMANOVA (R² = 3.74%, F₁₋₈₅=3.31, p = 0.007). As shown in Fig. 5 below, at Moudmam, solitary bees were the most abundant pollinator group, constituting 24.73% of the total visits. Beetles were the second most common, accounting for 18.29% of the visits. Honey bees and flies each account for approximately 10% of visits (10.6% and 10.2%, respectively). However, butterflies and bumblebees also contributed similarly lower percentages (8.95% and 8.32%, respectively). Hoverflies, wasps, and ants are less common, each contributing less than 7% of visits. Beeflies are the least abundant, performing only 1.26% of the visits. At Ras El Ma, solitary bees remained the most frequent pollinator group, accounting for 27.14% of visits. Butterflies are more prevalent than at Moudmam, performing 23.14% of visits. Honeybees contributed 15.05%, followed by beetles at 12.65% and bumblebees at 7.45%. The remaining pollinator groups are less common, each accounting for less than 4% of the total visits.

Fig. 5.

Frequency of visits of different pollinator groups to plant species within the two studied communities

Factors influencing the pollination generalization level

The GLM analysis identified four significant predictors of pollination generalization (PG) in the full model integrating data from both localities (Table 4), two floral traits—flower clustering and reward type—and two other factors included as covariates—total pollinator visits and locality.

Table 4.

Summary of GLMs (full model and site-specific models) showing the influence of flower clustering, reward type, locality, and total pollinator visits on pollination generalization (PG). The percent change values represent back-transformed estimates from the log-scale models, with negative and positive values indicating percentage decreases and increases in the PG relative to the reference levels (inflorescence clustering, pollen-only reward, and Moudmam locality)

Predictor	Estimate β	Std. Error	t value	Pr(>|t|)	Percent Change in PG
Full Model
(Intercept)	−0.308	0.145	−2.118	0.0372
Clustering: Solitary flowers	−0.311	0.121	−2.584	0.0115	–26.8%
Reward: Pollen & Nectar	0.153	0.073	2.101	0.0387	+ 16.5%
Locality: Ras el Ma	−0.415	0.070	−5.914	< 0.0001	–33.9%
log(Total pollinator Visits)	0.316	0.032	9.921	< 0.0001	+ 37.1%
Moudmam Subset Model
(Intercept)	−0.637	0.240	−2.651	0.0113
Clustering: Solitary flowers	−0.322	0.164	−1.956	0.0572	–27.5%
Reward: Pollen & Nectar	0.192	0.105	1.823	0.0755	+ 21.1%
log(Total pollinator Visits)	0.389	0.054	7.207	< 0.0001	+ 47.6%
Ras el Ma Subset Model
(Intercept)	−0.512	0.159	−3.217	0.0027
Clustering: Solitary flowers	−0.223	0.176	−1.269	0.2125	–20.0%
Reward: Pollen & Nectar	0.113	0.096	1.175	0.2474	+ 12.0%
log(Total pollinator Visits)	0.269	0.036	7.394	< 0.0001	+ 30.9%

Compared with those with inflorescences, plants with solitary flowers presented significantly lower pollination generalization compared to those with inflorescences (β = −0.311, t = −2.584, p = 0.011), corresponding to a 26.8% decrease in the PG. Furthermore, plants offering both pollen and nectar as rewards presented 16.5% higher PG values than did plants offering pollen alone (β = 0.153, t = 2.101, p = 0.038).

Overall, total pollinator visits showed the strongest positive relationship with PG (β = 0.316, t = 9.921, p < 0.001), with each unit increase in log-transformed visits corresponding to a 37.1% increase in pollination generalization. Locality emerged as another highly significant predictor, with plants at Ras el Ma exhibiting substantially lower PG values (β = −0.415, t = −5.914, p < 0.001), representing a 33.9% reduction in pollination generalization compared with plants at Moudmam.

When analysing each locality separately (Table 4), we observed consistent directional effects but with varying significance levels. At Moudmam, total pollinator visits maintained a strong positive association with PG (β = 0.389, t = 7.207, p < 0.001), with each unit increase in log-transformed visits corresponding to a substantial 47.6% increase in pollination generalization. The effect of floral clustering approached significance (β = −0.322, t = −1.956, p = 0.057), with solitary flowers showing 27.5% lower PG values than inflorescences did. Similarly, reward type had a marginally significant positive effect (β = 0.192, t = 1.823, p = 0.075), with combined pollen and nectar rewards associated with 21.1% higher PG values.

At Ras el Ma, total pollinator visits remained the only significant predictor (β = 0.269, t = 7.394, p < 0.0001), corresponding to a 30.9% increase in PG per unit increase in log-transformed visits. Neither floral clustering (β = −0.223, t = −1.269, p = 0.212) nor reward type (β = 0.113, t = 1.175, p = 0.247) reached statistical significance at this locality, although their effect directions remained consistent with those of the full model.

The visualization of the predicted values (Fig. 6) clearly illustrates these patterns. Compared with those at Ras el Ma, plants at Moudmam consistently exhibited greater pollination generalization across all categories of floral traits. Compared with solitary flowers, inflorescences were associated with higher predicted PG values at both localities, as were plants offering both pollen and nectar rewards compared with those offering pollen alone. The relationship between total pollinator visits and PG followed an exponential pattern at both sites, with a steeper curve observed at Moudmam, which was consistent with the larger coefficient in the site-specific model.

Fig. 6.

Predicted pollination generalization based on the GLM results. A Predicted pollination generalization (back-transformed from a log scale) by floral clustering and B) reward type across the two study localities (Moudmam, Ras el Ma). The right panel C) shows the relationship between pollination generalization and total pollinator visits (log scale) with fitted curves derived from the Gaussian GLM with a log-transformed response.

Our results indicate that the influence of floral traits on pollination generalization may be modulated by locality-specific ecological factors, particularly insect community abundance. These factors likely drove the greater number of pollinator visits observed at Moudmam, which in turn led to increased PG values. This ecological context explains the consistently greater pollination generalization at Moudmam across all floral trait categories, suggesting that while floral characteristics matter, their importance may be mediated by the local pollinator community structure.

Relationships between pollinator groups and floral traits

The results of the canonical correspondence analysis (CCA) revealed the contributions of different floral traits to pollinator distributions across the studied communities (both communities combined, Moudmam, and Ras el Ma). For both communities, the overall model was significant (𝑝=0.042), with key contributors including orientation (𝑝=0.006) and anther position (p = 0.032). In Moudmam, the model was marginally significant (p = 0.092), with significant contributions from clustering (p = 0.024), orientation (p = 0.006), and color (p = 0.048). For Ras el Ma, the overall model was not significant (p = 0.295), but ovary position (p = 0.028) and diameter (p = 0.014) had significant effects. Flower shape (assessed separately) was significant for both communities (p = 0.047) (Table 5).

Table 5.

Results of the CCA analysis used to assess the relationship between floral traits and pollinator groups preferences, at both communities and at each community

Community	Effect	Df	ChiSquare	F value	Pr(> F)
Both communities	All model	20	0.9505	1.3415	0.042 *
	Locality	1	0.072	2.035	0.029 *
	Orientation	1	0.095	2.679	0.006 **
	Anther position	1	0.082	2.325	0.032 *
	Diameter	1	0.062	1.752	0.08.
	Clustering	1	0.06	1.688	0.093.
	Reward	1	0.059	1.661	0.089.
	Color	6	0.269	1.267	0.172
	Ovary position	1	0.045	1.256	0.266
	Corolla length	1	0.04	1.134	0.34
	Scent	3	0.105	0.984	0.505
	Form	1	0.024	0.672	0.732
	Symmetry	1	0.023	0.637	0.76
	Height	1	0.016	0.44	0.902
	Shape (i)	13	0.7329	1.6143	0.026 *
	Total inertia = 2.8283, constrained = 0.9505, unconstrained = 1.8778.
Moudmam	All Model	17	1.0397	1.2746	0.092.
	Clustering	1	0.132	2.583	0.024 *
	Orientation	1	0.129	2.536	0.006 **
	Color	4	0.359	1.761	0.048 *
	Corolla length	1	0.071	1.392	0.212
	Diameter	1	0.063	1.237	0.3
	Reward	1	0.063	1.232	0.283
	Anther position	1	0.052	1.010	0.413
	Scent	3	0.114	0.744	0.803
	Ovary_position	1	0.036	0.711	0.756
	Symmetry	1	0.032	0.634	0.785
	Form	1	0.026	0.502	0.808
	Height	1	0.015	0.292	0.961
	Total inertia = 2.212, constrained = 1.040, unconstrained = 1.173.
Ras el Ma	All model	18	2.0366	1.107	0.295
	Diameter	1	0.245	2.397	0.014 *
	Ovary position	1	0.198	1.942	0.028 *
	Height	1	0.177	1.732	0.161
	Anther position	1	0.167	1.633	0.142
	Corolla length	1	0.139	1.355	0.182
	Orientation	1	0.121	1.189	0.291
	Reward	1	0.121	1.185	0.302
	Symmetry	1	0.112	1.101	0.393
	Scent	3	0.306	0.997	0.506
	Color	5	0.349	0.683	0.832
	Clustering	1	0.065	0.639	0.512
	Form	1	0.036	0.353	0.94
	Total inertia = 3.570, constrained = 2.037, unconstrained = 1.533.

⁽ⁱ⁾ Flower shape was asseesed separately of other traits for both communities

Significance levels: "." p < 0.1, "*" p < 0.05, "**" p < 0.01

In all the models, the constrained inertia values (0.9505/2.828 for both communities, 1.040/2.212 for Moudmam, and 2.037/3.570 for Ras el Ma) suggest varying degrees of explanatory power for the floral traits, with higher proportions of unexplained variation (unconstrained inertia) indicating complex plant pollinator networks or probably additional factors influencing pollinator group distributions.

In the overall analysis combining both sites, the first two axes explained 58.4% of the variation in floral traits (CCA1: 35.7%, CCA2: 22.7%) (Fig. 7A) and 61.0% of the variation in flower shape (CCA1: 36.8%, CCA2: 25.2%) (Fig. 7B). When analysed separately, the first two axes explained 54.5% of the variation in Moudmam (CCA1: 29.5%, CCA2: 25%) (Fig. 8A) and 45.9% in Ras el Ma (CCA1: 28.7%, CCA2: 17.2%) (Fig. 8B).

Fig. 7.

Biplots corresponding to the CCA analysis conducted to investigate the relationships between floral traits and pollinator group preferences at both sites (pooled data). A All floral traits and (B) floral shape. The gray triangles represent categorical floral traits, the arrows represent continuous traits, and the blue circles represent insect groups

Fig. 8.

Biplots from the CCA analysis used to investigate the relationships between floral traits and pollinator group preferences at Moudmam (A) and Ras el Ma (B). The gray triangles represent categorical floral traits, the arrows represent continuous traits, and the blue circles represent insect groups

The overall analysis (Fig. 7) revealed that bumblebees primarily visit vertically oriented flowers with shapes such as snapdragon and pea flowers, which have zygomorphic symmetry and hidden or semi-hidden anthers. Beeflies showed preferences for pink flowers with tubular shapes. In contrast, flies, hoverflies and ants are commonly found on open flowers with disk or star shapes, where rewards are easily accessible. The butterflies preferred cross-shaped and tubular flowers, particularly those with pink or purple colours. Interestingly, solitary bees, honey bees and beetles presented more generalist behavior, without strong preferences for specific traits.

In Moudmam (Fig. 8A), as shown in the overall analysis, bumblebees were associated with zygomorphic, vertically oriented flowers, whereas beeflies were strongly associated with pink flowers. Ants and flies clustered together, showing preferences for open flowers. At Ras el Ma (Fig. 8B), butterfly flowers were strongly associated with vertically oriented flowers with long corollas and pink colors, and ants, flies and beetles visited large and open flowers more often.

Discussion

Our study revealed a high level of floristic diversity in both communities, reflecting the success of ongoing conservation efforts within the National Park of Ifrane [39]. However, the two communities presented significant differences in species composition, likely driven by distinct habitat characteristics such as soil moisture, light availability, and other microclimatic conditions. Similarly, the composition of pollinator communities differed between sites and was influenced by both local environmental conditions and variations in plant communities. As shown by [NO_PRINTED_FORM] [40], changes in plant species composition can lead to marked shifts in insect community structure, affecting various taxonomic orders and functional groups. These patterns of habitat-specific distributions of both plants and pollinators underscore the importance of preserving habitat heterogeneity to support biodiversity conservation.

The distribution of floral traits showed both habitat-specific and shared patterns. Yellow flowers dominated both communities, followed by white, purple, and pink flowers, whereas blue and green flowers were rare. This color distribution aligns with selection pressures for high reflectance in Mediterranean environments, as light-colored flowers typically attract more pollinators [41–43]. The observed differences in the other floral traits between the two sites are likely shaped by both species turnover and habitat-specific environmental filtering. For example, traits such as floral symmetry, orientation, and anther exposure may be selected in response to differences in light conditions or pollinator behavior in open versus semishaded environments [44, 45]. In addition, trait composition may also reflect the dominant plant lineages adapted to each habitat type.

Our analysis of pollination generalization revealed significant differences between the two communities. This difference is related to the plant community composition and variation in pollinator abundance and diversity. Our results demonstrated that the number of pollinator visits—a proxy of pollinator abundance—significantly increased pollination generalization. Several studies have also shown that greater pollinator availability enhances generalization [14, 46]. The analysis also revealed that flower clustering and reward type were the only traits that significantly influenced generalization levels. Clustered flowers show higher generalization levels than solitary flowers do, likely due to their enhanced visual display and greater resource abundance [47]. This finding aligns with that of [48], who reported a consistent effect of clustering on pollination generalization in an alpine plant community across three years but partially contrasts with that of [36], who reported that the effect of clustering on generalization varied among communities, with solitary flowers being more generalist in some locations. This difference might reflect the distinct ecological contexts between Mediterranean and Nordic communities. The significant effect of rewards on generalization levels highlights the importance of resource availability in shaping plant‒pollinator interactions. This finding is supported by [49] experimental work, which demonstrated that nectar accessibility significantly influences pollinator behavior and visitation patterns. While their study focused on moth pollinators, the principle that reward accessibility affects pollinator choices appears to hold true across different pollinator groups [50, 51].

Notably, we did not find significant relationships between several floral traits (i.e., flower form, color and symmetry) and pollination generalization in either community, differing from previous studies that reported such associations [36, 52–54]. The absence of significant relationships between flower form, color, and symmetry and pollination generalization in our study may reflect context-specific pollinator behaviors. This highlights that generalization patterns can vary across ecological settings, supporting the idea that trait–pollination links are not universally consistent [55].

The CCA revealed pollinator preferences that both support and challenge traditional pollination syndrome concepts. Bumblebees (Bombus terrestris) were strongly associated with closed, zygomorphic flowers, whereas flies and ants preferred open, actinomorphic flowers. These preferences align with the classical pollination syndrome concept proposed by Faegri & Pijl [7] and have been corroborated by various studies [8, 15, 36]. Solitary bees and beetles, however, display remarkable flexibility in their floral choices, likely reflecting their taxonomic and functional diversity, which conflicts with their classification as single pollinator functional groups in the pollination syndrome concept [56]. Long-proboscis insects (butterflies and beeflies) consistently visit tubular flowers, demonstrating classic coevolutionary relationships [57, 58].

The relatively high proportion of unexplained variation in the CCA models suggests that factors other than the studied floral traits may influence pollinator distribution. For example, environmental variables such as microclimatic conditions (e.g., temperature, humidity, and wind exposure) and habitat heterogeneity can shape pollinator activity and foraging patterns [59]. Future studies that integrate these ecological and environmental factors may provide a more comprehensive understanding of the drivers shaping plant–pollinator networks.

One important limitation of this study lies in the qualitative assessment of certain floral traits, particularly color and scent, which are based on human perception. This approach does not fully reflect how pollinators perceive these cues. For example, many insects—especially bees—can detect ultraviolet (UV) wavelengths, which are invisible to the human eye and may rely on UV-reflectance patterns or contrasts that were not captured in our classifications [60]. Similarly, floral scents are recorded without accounting for their chemical composition, intensity, or complexity, all of which can strongly influence pollinator attraction and behavior [61]. These limitations may have contributed to the lack of significant associations between these traits and pollination generalization. Future studies could benefit from more objective and pollinator-relevant methodologies, such as spectrophotometry to quantify floral color across the UV‒visible spectrum and gas chromatography–mass spectrometry (GC‒MS) to chemically profile floral scents. The incorporation of such methods would improve the accuracy of trait characterization and provide deeper insights into the ecological and evolutionary mechanisms shaping plant–pollinator networks [62].

Conclusions

In this study, we investigated the relationships between flowering plants and pollinator groups in two preserved areas of the Ifrane National Park, Middle Atlas, Morocco. Our analysis focused on describing floral traits within plant communities and their influence on pollinator distribution. Our findings reveal complex interactions between habitat structure, floral traits, and pollinator communities. While some aspects of traditional pollination syndromes hold true, the community-specific nature of many plant‒pollinator relationships suggests that the local context strongly influences these interactions.

Study limitations include the single-season sampling period, the focus on only two habitat types, and the reliance on qualitative descriptions for certain floral traits, such as color, reward, and scent. To improve the generalizability of our findings, future research should investigate interannual variation, incorporate a broader range of habitat types, and employ quantitative measurements of floral traits. Additionally, long-term monitoring at finer taxonomic resolutions would be valuable for assessing the temporal stability of plant–pollinator networks, particularly in the context of ongoing climate change.

Authors’ contributions

Conceptualization, original draft writing, reviewing, and editing: Aziz Aitakka, Soufyane Lafraxo, Fatima Zahra Jawhari. Formal analysis, investigations, funding acquisition, reviewing, and editing: Ahlam Sentil, Amina Bari, Raja Guemmouh. Resources, data validation, data curation, and supervision: Mohammed Bourhia, Abdel-Rhman Z. Gaafar, Youssouf Ali Younous.

Funding

This work is financially supported by the Ongoing Research Funding Program (ORF-2025-686), King Saud University, Riyadh, Saudi Arabia

Data availability

All data generated or analyzed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

The current research is approved by the Institutional Research Ethics Committee of the University of Sidi Mohammed Ben Abdellah, Morocco. The collection of plant material complies with relevant institutional, national, and international guidelines and legislation.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.