Insect visitation patterns in diploid Centaurea aspera and its related allotetraploid and triploid hybrids: Similar rates but distinct assemblages

Abstract

Premise

Polyploidy is key to plant evolution by contributing to speciation, diversification, and adaptability. However, the minority cytotype exclusion effect can limit the persistence of polyploids, which can be mitigated by reproductive barriers such as distinct insect visitation between cytotypes. In eastern Spain, the diploid C. aspera, its related allotetraploid C. seridis, and their sterile triploid hybrid C. ×subdecurrens coexist in contact zones. Here we assessed the diversity and behavior of insects visiting these Centaurea taxa, identified factors influencing insect visitation, and explored potential changes in visitor composition and frequency across taxa in the contact zone of El Saler (Valencia, Spain).

Methods

Five natural blocks (25–50 m²) on sand dunes, each with all three plant taxa in proximity, were monitored weekly when their flowering periods overlapped. Insect visitors were identified, and number of visits and of capitula were recorded. Linear modelling was used to identify factors predicting visit frequency and differences in insect composition among plant taxa.

Results

Seventeen flying insect species visited Centaurea plants. The number of florets and the number of capitula were the strongest predictors of visit frequency, showing similar outcomes, rather than plant taxon or date. Although overall visitation rates did not significantly differ among plant taxa, insect assemblages varied. Centaurea seridis attracted a distinct set of insects compared to C. aspera and C. ×subdecurrens, with some insects displaying visitation biases at particular times.

Conclusions

Differences in insect assemblages suggest potential prezygotic barriers that could help C. seridis overcome minority exclusion, supporting its long‐term establishment.

Polyploidy, also known as whole‐genome duplication (WGD), is regarded as a major driving force in the evolution of plants (Wendel, 2015). All plant groups have experienced one or more ancestral WGD events, and approximately half of plant species are recent polyploids that involve one (autopolyploidy) or more (allopolyploidy) parental species (Landis et al., 2018). The long‐term persistence of polyploids has long been debated. On one hand, polyploidy can lead to immediate speciation and has been associated with increased rates of diversification, fixed heterozygosity, hybrid vigor, and genome modifications, all contributing to the success of neopolyploids (Heslop‐Harrison et al., 2023). On the other hand, it has been argued that most neopolyploids are short‐lived and may be repeatedly extinguished or revert to nonduplicated ploidies, with their long‐term survival dependent on specific conditions (Madlung, 2012). One key constraint to the survival of new polyploids is the difficulty of finding compatible reproductive mates of the same ploidy level (cytotype) because these polyploids are often less abundant than their diploid relatives. Crosses between different cytotypes typically produce nonviable progeny, a phenomenon known as a triploid block (Köhler et al., 2010). This reproductive barrier can lead to minority cytotype exclusion (Levin, 1975), a positive frequency‐dependent mechanism that causes the decline of neopolyploids over time, due to their higher proportion of gametes that produce sterile offspring compared to the more widespread lower‐ploidy, ancestral cytotype.

Despite this debate, polyploids are often observed in nature, growing sympatrically with diploids (i.e., Ferriol et al., 2012; Mráz et al., 2012, Sutherland and Galloway, 2021; Vaz de Sousa et al., 2024). In addition to recurrent polyploid formation through high rates of unreduced gamete production (Soltis and Soltis, 1999), several pre‐ and postzygotic reproductive barriers have been identified that allow plants of the same ploidy level to selectively mate, mitigating the minor cytotype exclusion effect, and facilitating their establishment and persistence. Prezygotic barriers include changes in the reproductive system of neopolyploids (e.g., the emergence of asexuality and self‐compatibility, Husband et al., 2013), phenological shifts in flowering times (Segraves and Thompson, 1999; Pegoraro et al., 2019), spatial and ecological segregation of cytotypes even on small scales (e.g. Kobrlová et al., 2022), perenniality (Van Drunen and Husband, 2019), gametic selection due to pollen competition in mixed‐ploidy loads, and mentor effects (Hörandl and Tesch, 2009; Baldwin and Husband, 2011). Postzygotic barriers are typically strong between diploids and tetraploids due to the triploid block, although it is sometimes incomplete (Ramsey and Schemske, 1998). In contrast, reproductive barriers at higher ploidy levels vary among polyploid complexes (Sonnleitner et al., 2013; Sutherland and Galloway, 2021; Šemberová et al., 2023).

Additionally, polyploidization may induce morphological and metabolic shifts that can affect fitness. Genome duplication is often associated with larger cells, leading to bigger plant organs and structures, such as petals, flowers, and inflorescences (also called the gigas effect) (Stebbins, 1971; te Beest et al., 2012). Shifts in ploidy can also affect the morphological match between pollinator species and floral structures (e.g., changes in floral tube length [McCarthy, 2019] or stigma–anther distance [Casazza et al., 2017]), flower color, nectar quantity, and scent chemistry (Gross and Schiestl, 2015; Segraves and Anneberg, 2016; Mortier et al., 2024). Pollinators may respond to small variations in floral signals and rewards, and as a result, polyploidization can alter pollinator assemblage and behavior (Rezende et al., 2020). These changes in pollinator communities can lead to assortative mating and alleviate the minor cytotype exclusion effect (Segraves and Anneberg, 2016).

However, studies dealing with pollinator visitation biases for a given cytotype or competition for pollinators between cytotypes are rare in mixed‐cytotype populations (Segraves and Thompson, 1999; Husband and Schemske, 2000; Kennedy et al., 2006; Thompson and Merg, 2008; Borges et al., 2012; Roccaforte et al., 2015; Laport et al., 2021). In some studies, due to challenges in identifying suitable mixed‐ploidy populations to study pollinator behavior, individuals of different cytotypes were observed in different times and/or different locations (Borges et al., 2012). One alternative approach involves the use of artificial floral arrays, where cut inflorescences of different cytotypes are displayed in alternating patterns in the field or greenhouse (Gross and Schiestl, 2015; Castro et al., 2011, 2020). These arrays allow for controlled observations of insect movement between inflorescences to assess floral isolation, preference, or constancy for specific cytotypes (Jersáková et al., 2010; Gross and Schiestl, 2015; Castro et al., 2020; Sutherland et al., 2020; Schmickl et al., 2024). Nonetheless, natural conditions offer a more accurate representation of pollinator behavior (Segraves and Thompson, 1999; Castro et al., 2011; Borges et al., 2012; Abrahamczyk et al., 2021) because reproductive biology studies have documented notable differences between natural and artificial settings (Cook and Soltis, 1999, 2000).

The genus Centaurea L. is one of the most diverse within the Asteraceae family, comprising ca. 250 species (Susanna and García‐Jacas, 2009), despite its relatively recent evolutionary origin, 4–11 million years ago (Mya) (Hilpold et al., 2014). The high rate of speciation in Centaurea is largely attributed to disploidy, hybridization, and polyploidization, which can lead to the formation of new species in sympatry in one or a few generations (Hellwig, 2004; Romaschenko et al., 2004; García‐Jacas, 2006). Polyploids represent approximately 18% of Centaurea taxa, with allopolyploidization only accounting for 15% of neopolyploids (Barker et al., 2016). Several contact zones have been described where different cytotypes coexist and interact. They involve primarily diploids and autotetraploids (e.g., Hardy et al., 2001; Koutecký et al., 2012; Olsavska and Löser, 2013), along with some allotetraploids (e.g., Mráz et al., 2012). While Centaurea cytotypes in these contact zones often have overlapping flowering periods, facilitating potential crossbreeding, intercytotype hybridization is rare (Koutecky et al., 2012). When hybridization occurs, it predominantly produces tetraploid offspring and less frequently triploid hybrids (Hardy et al., 2000; Koutecký, 2012; Koutecký et al., 2007; Spaniel et al. 2008). Both pre‐ and postzygotic barriers likely contribute to the rarity of intercytotype hybridization in Centaurea. Prezygotic barriers include spatial segregation and microhabitat segregation in contact zones (Hardy et al., 2000; Font et al., 2008; Mráz et al., 2012), phenological shifts (Koutecký, 2012), and mentor effects (Koutecky et al., 2011). Additionally, Centaurea species are highly attractive to both oligolectic and polylectic wild bees (Kuppler et al., 2023), so changes in pollinator behavior due to polyploidization can contribute significantly to prezygotic isolation of cytotypes. Flower and capitulum traits influencing pollinator visits in Centaurea include flower coloration (Renoult et al., 2013), the presence or absence of ray florets (Lack, 1982a), essential oil composition (Novakovic et al., 2019), and nectar production (Lack, 1982b). Postzygotic barriers are typically evident in the low germination rates, low viability, and high sterility of triploid hybrids (Mráz et al., 2012; Olsavska and Löser, 2013).

The diploid Centaurea aspera L. is one of the parental species of the allotetraploid C. seridis L., although the other parent remains unknown. Centaurea seridis exhibits high fixed heterozygosity, supporting its recent origin (Ferriol et al., 2014). Both taxa belong to sect. Seridia (Juss.) Czerep. and coexist in contact zones along the Spanish Mediterranean coast (Ferriol et al., 2014). Unlike other Centaurea polyploid complexes, these species hybridize frequently, producing sterile triploid hybrids (C. ×subdecurrens Pau) that are morphologically intermediate (Ferriol et al., 2012, 2015). The frequent production of sterile triploids suggests that postzygotic barriers among cytotypes are strong, while prezygotic barriers may be weaker. Some of them, such as the 50% overlap of the flowering periods of diploids, triploids, and tetraploids (Ferriol et al., 2015), and the lack of microhabitat segregation of cytotypes (Garmendia et al., 2018), have already been assessed. The neotetraploid C. seridis is highly autogamous, showing no significant difference in seed production between self‐pollination and intraspecific pollinations in controlled conditions. In contrast, the diploid C. aspera is allogamous, resulting in asymmetrical production of sterile triploid hybrids as C. aspera acts as the female parent, and mitigating the minority cytotype exclusion effect (Ferriol et al., 2015). However, the impact of allopolyploidization on pollinator behavior has not yet been studied. The Centaurea aspera polyploid complex offers a tractable system for assessing competition for pollinators among related plant taxa that involve different ploidy levels (cytotypes) and hybridization (allopolyploidy) in natural conditions. As far as we know, no studies have focused on self‐incompatible diploids, self‐compatible allotetraploids, and triploid hybrids. The questions we addressed were (1) What are the identity, diversity, and behavior of insects visiting Centaurea taxa? We hypothesized that the diversity and behavior of insect visitors to Centaurea in the dunes of El Saler will be high and comparable to those observed in other habitats where Centaurea species occur. (2) What are the main factors influencing total insect visits to Centaurea capitula? We expected that differences in the number and size of capitula per plant might influence number of insect visits. (3) Are there differences in flower visitor composition or visitation frequency among Centaurea taxa? Given the differences in capitulum morphology and slight variations in flowering peak timing among the taxa, we posited that these factors could influence both insect abundance and community composition.

MATERIALS AND METHODS

Study system and study area

The diploid Centaurea aspera and the tetraploid C. seridis are perennial herbaceous plants distributed in the Western Mediterranean. In Europe, only diploid populations of C. aspera have been recorded (compiled by Invernón et al., 2013). In eastern Spain, Centaurea aspera is widespread and grows in a wide range of habitats, especially in drylands and nitrophilous habitats, generally at low altitudes (Bolòs and Vigo, 1995). In contrast, C. seridis has a more restricted distribution, primarily on coastal sand dunes that are often subject to significant anthropogenic disturbance and nitrophilic conditions (Costa and Mansanet, 1981). These two species coexist on disturbed, semifixed coastal dunes, where they hybridize to form triploid C. ×subdecurrens plants (Ferriol et al., 2012). Seven contact zones have been identified along the Spanish Mediterranean coast, including one at El Saler (Valencia, Spain) (Ferriol et al., 2014). There is no clear evidence to confirm whether these contact zones are primary or secondary. Genetic data and geographic clustering support primary contact zones, suggesting recurrent formation of C. seridis (Ferriol et al., 2014). However, the allopolyploid origin of C. seridis—with C. aspera as one known parental species and the other still unidentified—challenges the hypothesis of in situ evolution of the tetraploid and instead points toward secondary contact zones (Španiel et al., 2008; Mráz et al., 2012).

The genus Centaurea is insect‐pollinated, mainly by honeybees, bumblebees, and other native bees (Harrod and Taylor, 1995). Although C. aspera is widespread in the Western Mediterranean, insect visitation has scarcely been studied in this species, and studies on C. seridis and C. ×subdecurrens are lacking. Centaurea aspera produces a high quantity of nectar and pollen compared to other co‐occurring species in different families in Spain (Bosch et al., 1997; Aguado et al., 2015). In herbaceous plant communities in Spain, the primary visitors to C. aspera capitula are insects in the order Hymenoptera, particularly bees and ants, followed by Coleoptera, and to a lesser extent, Hemiptera (Heteroptera) (Bosch et al., 1997). Lepidopterans, especially members of the Pieridae and Zygaenidae families, have also been recorded as pollinators of C. aspera in Spain (Aguado et al., 2015).

In El Saler, the three taxa differ in the morphology of florets and capitula (Ferriol et al., 2012) (Figure 1). Centaurea seridis has significantly larger capitula, more and longer bracteae spines, and longer stamens and anthers compared to C. aspera, with C.×subdecurrens having intermediate traits. Additionally, C. seridis has longer inner and outer florets and longer pistils than C. aspera, although no significant differences were found between C. seridis and C. ×subdecurrens for these traits. The hybrid C. ×subdecurrens has more capitulum bracteae and outer florets than C. aspera and C. seridis (Ferriol et al., 2012). Voucher specimens of one representative of each studied taxon are in the Herbarium of the Universitat Politècnica de València: VALA9504 for C. aspera, VALA9505 for C. seridis, and VALA9506 for C. ×subdecurrens. In terms of flowering periods, C. seridis blooms from April to July for 17–21 weeks, C. aspera blooms from May to November for 27–33 weeks, and the hybrids have an intermediate flowering period lasting 23–33 weeks, from mid‐April to October (Ferriol et al., 2015).

Figure 1.

(A) Left: Location of studied Centaurea individuals in blocks (white rectangles) in the sand dunes of El Saler, Valencia, Spain. Orthophotograph 2018 CC BY 4.0 © Institut Cartogràfic Valencià, Generalitat Valenciana (https://creativecommons.org/licenses/by/4.0/deed.ca). Right: Location of the El Saler contact zone (red dot) in the context of other known Centaurea contact zones in eastern Spain (yellow dots). (B) Morphology of capitulum and upper leaves for each studied taxon.

The study was conducted in the dunes of El Saler, located in the Natural Park of l'Albufera in Valencia (E Spain, 39°21′41.67″N, 0°19′05.53″E) within 1.77 ha. In two nearby zones of El Saler, C. aspera, C. seridis, and C. ×subdecurrens individuals were recorded and mapped at a microspatial scale (Garmendia et al., 2018). In the study area, the proportion of individuals was estimated to be 60.1% C. aspera, 31.3% C. seridis, and 8.6% C. ×subdecurrens (Appendix S1). The relatively low number of C. ×subdecurrens individuals, combined with the fact that all three taxa exhibited aggregated distribution patterns at a fine scale—with hybrids generally occurring closer to C. aspera than to C. seridis—made it challenging to select blocks on the sand dunes with an area of 25–50 m² where all three taxa co‐occurred in close proximity. In April 2020, we selected eight candidate blocks, each consisting of three Centaurea plants—one of C. aspera, one of C. seridis, and one of C. ×subdecurrens—with plants spaced less than 3 m apart, except for two blocks in which one plant was spaced up to 7 m from the others, allowing insects to select capitula based on their preferences. However, in three of these blocks, one of the plants failed to reach sufficient size to produce enough capitula throughout the flowering period, and these blocks were subsequently excluded. In the remaining five blocks, each plant was labeled with a neutral sand‐colored string and located using GPS (Figure 1). The diameter of each plant was estimated by averaging the maximum width and the perpendicular width at its midpoint. The distance from each studied plant to the nearest individuals of each Centaurea taxon not included in the same block was also measured (Appendix S2). Due to the COVID‐19 pandemic, the lockdown, and subsequent restrictions on beach access—except for authorized personnel—anthropogenic disturbances were minimized, thereby reducing potential differences among the blocks that might otherwise result from varying levels of human activity affecting insects.

Flower visitation patterns

For 14 weeks, from 23 April (when the first plants began to flower) to 23 July 2020 (when most C. seridis plants stopped setting flowers), the study area was visited weekly. On each visit, we selected the day with the highest forecasted air temperature and the least cloud cover, avoiding rainy days. During each visit, wind intensity (calm, breeze, or wind) and direction were estimated using data provided by AEMET (2020) for La Devesa beach, in El Saler. The mean temperature and relative humidity were recorded using a digital thermometer. Visits took place between 09:30 and 15:00 hours, when pollinator activity was at its peak. However, as spring progressed, the timing of visits was adjusted earlier to avoid excessively high temperatures.

Because phenology varies among Centaurea taxa, the number of open capitula was counted in the field for each plant on every visit. The number of florets per plant was estimated by counting the florets on six random capitula for each taxon. We then multiplied the mean number of florets by the recorded total number of open capitula: 22 ± 3.9 florets per capitulum in C. aspera, 43 ± 3.3 in C. seridis, and 39 ± 2.1 in C. ×subdecurrens.

Each plant was observed for 10 min by two independent observers, including at least one entomologist. Visits from any flying flower visitor were recorded, counting a visit each time an insect landed on a capitulum while actively searching for pollen and/or nectar. If the same insect landed on another capitulum of the same plant, another visit was counted. Unknown insects were photographed, captured with a butterfly net, and brought to the Elytra Agroscience laboratory (Valencia, Spain) for identification by specialized entomologists, who determined the species, genus, or family.

Statistical analyses

Visualization of data

Original data were generated and visualized using packages dplyr (Wickham et al., 2022), tidyr (Wickham and Girlich, 2022), ggplot2 (Wickham, 2016), and ggpubr (Kassambara, 2022) in R version 4.3.3 (R Core Team, 2024).

Identification of factors that predicted the number of insect visits

The correlation between the total number of insect visits and relative humidity (which was not normally distributed) and air temperature was assessed using Spearman's rank correlation and Pearson's correlation analysis, respectively. The effect of categorical weather variables (wind and cloudiness) on the total number of insect visits was analyzed with one‐way ANOVAs. Subsequently, only insect species that visited Centaurea plants at least three times were included in the statistical analysis. To identify factors predicting the number of insect visits per plant, we used linear regression models from the R package MASS (Venables and Ripley, 2002). Data on the number of insect visits were log‐transformed to achieve normality of residuals when used as the dependent (response) variable. The independent (predictor) variables included the number of capitula per plant and the number of florets per plant, and both were log‐transformed. Other covariables were date, block, plant taxon, and their combinations. Data for days and plants that had no capitula and/or no insect visits were excluded. Nineteen linear double‐logarithmic models were analyzed in which the variables were (lm1) log (number of capitula per plant); (lm2) plant taxon; (lm3) date; (lm4) block; (lm5) log (number of florets per plant); (lm6) log (number of florets per plant) and plant taxon; (lm7) log (number of florets per plant) and date; (lm8) log (number of florets per plant) and block; (lm9) log(number of florets per plant), plant taxon, and date; (lm10) log (number of florets per plant), plant taxon, and block; (lm11) log (number of florets per plant), date, and block; (lm12) log (number of florets per plant), plant taxon, date, and block; (lm13) log (number of florets per plant) and interaction with plant taxon; (lm14) log (number of florets per plant) and interaction with date; (lm15) log (number of florets per plant) and interaction with block; (lm16) log (number of florets per plant) and interaction with plant taxon and date; (lm17) log (number of florets per plant) and interaction with plant taxon and block; (lm18) log (number of florets per plant) and interaction with date and block; and (lm19) log (number of florets per plant) and interaction with plant taxon, date, and block.

The normality of the data distribution was assessed using the Shapiro–Wilk normality test. The models were evaluated using the corrected Akaike information criterion (AIC) and Bayesian information criterion (BIC) as model selection criteria (Wit et al., 2012).

For the best‐fitting model, estimates, standard errors, and P‐values for the intercept, coefficients of each predictor variable, and P‐value for the model as a whole were calculated. Flower visitors with significant P‐values when the best linear model was applied to them individually were also identified.

Differences in composition of species of flower visitors and frequency of visits among Centaurea taxa

To examine whether the larger capitula of C. seridis attracted more insects than those of C. aspera, differences in insect visits per capitulum and floret were assessed across the different Centaurea taxa. Specifically, we analyzed both metrics to determine whether adjusting for the number of florets—accounting for the potentially greater attractiveness of larger capitula—would improve the model fit. A Kruskal–Wallis test was performed using the residuals from the models that best fit the observed data and that were previously selected: lm1 and lm5, where the predictor variables were log (number of capitula per plant) and log (number of florets per plant), respectively.

To test for statistical differences in flower visitor composition among plant taxa, we used a general linear model with negative binomial distribution because it fit the data better than normal and Poisson models. To identify the insects that were significantly associated with a given plant taxon, we used a multilevel pattern analysis of the association between insect species frequencies and plant taxa. Additionally, multivariate‐response generalized linear mixed models (MGLM) were performed using insect abundance data with the package mvabund in R (Wang et al., 2012). This analysis provided a multivariate test and pairwise comparisons to evaluate the significance of factors affecting pollinator composition (plant taxa, interactions between plant taxa and number of capitula per plant, plant taxa and date, plant taxa and block, and plant taxa with the combination of number of capitula per plant, date, and block). For each of these models, univariate tests were conducted to estimate the significance of the association between a given insect species and each Centaurea taxon for each of the factors evaluated. Interactions between pollinators and plant species were visualized using a bipartite network created with the R package bipartite (Dormann et al., 2008).

Additionally, ANOVAs were performed for each pollinator and date separately to check for differences in number of insect visits between plant taxa.

RESULTS

Identification and temporal appearance of insects visiting Centaurea plants

During the 14 weeks of sampling, 17 different insect species were observed visiting Centaurea capitula (Table 1, Figure 2). The majority of the pollinators were hymenopterans, accounting for 79.4% of the visits. The most common species were Halictus scabiosae, followed by Bombus terrestris, Heriades crenulatus, and Rhodanthidium sticticum (Table 1). Five insect species were observed only once or twice: hymenopterans Bembix zonata, Dasypoda visnaga, and Euodynerus variegatus; dipterans Lucilia sericata; and lepidopteran Lampides boeticus. These five species were removed from subsequent statistical analyses.

Table 1.

Identification of insects visiting Centaurea capitula on the dunes of El Saler, Valencia, Spain. Generalist (G) or specialist (S) behavior is assigned based on literature for insect species that visited Centaurea at least five times.

Order	Family	Genus	Species	N
Hymenoptera	Apidae	Bombus	B. terrestris (G)	148
	Apidae	Anthophora	A. plumipes (G)	5
	Crabronidae	Bembix	B. zonata	1
	Halictidae	Halictus	H. quadricinctus (G)	37
	Halictidae	Halictus	H. scabiosae (G)	275
	Halictidae	Lasioglossum	L. discum (G)	18
	Megachilidae	Rhodanthidium	R. infuscatum (G)	24
	Megachilidae	Rhodanthidium	R. sticticum (G)	66
	Megachilidae	Pseudoanthidium	P. lituratum (S)	21
	Megachilidae	Heriades	H. crenulatus (S)	102
	Melittidae	Dasypoda	D. visnaga	1
	Vespidae	Euodynerus	E. variegatus	2
Coleoptera	Cetoniidae	Oxythyrea	O. funesta (G)	8
	Melyridae	Psilothrix	P. viridicoerulea	3
Diptera	Muscidae	Lucilia	L. sericata	2
	Tabanidae	Pangonius (Melanopangonius)	P. haustellatus (G)	5
Lepidoptera	Lycaenidae	Lampides	L. boeticus	1

Note: N, number of total visits.

Figure 2.

Insects observed visiting Centaurea capitula. (A) Bombus terrestris. (B) Anthophora plumipes photographed on Teucrium dunense. (C) Bembix zonata. (D) Halictus quadricinctus. (E) Halictus scabiosae (male individual on female individual). (F) Lasioglossum discum. (G) Rhodanthidium infuscatum. (H) Rhodanthidium sticticum. (I) Pseudoanthidium lituratum. (J) Heriades crenulatus. (K) Dasypoda visnaga. (L) Euodynerus variegatus. (M) Oxythyrea funesta. (N) Psilothrix viridicoreruela. (O) Lucilia sericata. (P) Pangonius (Melanopanganius) haustellatus photographed on Scabiosa atropurpurea L. (Q) Lampides boeticus.

Considering all plants and insects, a total of 719 visits to Centaurea capitula were recorded (Table 1). The most insects visited between 18 May and 23 June, with 70 to 106 insect visits per sampling day. The peak in insect visitation coincided with the peak flowering period of the three Centaurea taxa (Figure 3). On the remaining observation days, Centaurea plants received fewer than 50 visits. No clear relation was found between number of insect visits and cloudiness (ANOVA P = 0.341), wind intensity or direction (ANOVA P = 0.937), air temperature (Pearson correlation r = 0.15, N = 14, P = 0.60), or relative humidity (Spearman's rank correlation ρ = –0.22, N = 14, P = 0.44) (Table 2).

Figure 3.

Number of insect visits and number of capitula of the studied Centaurea plants growing in El Saler and recorded over time.

Table 2.

Number of Centaurea capitula/number of total visits for each sampled plant and sampling day on the dunes of El Saler, Valencia, Spain. Weather (cloudiness, wind, atmospheric temperature and relative humidity) and plant diameter (cm) of each plant are also provided.

Block	Plant taxon	Plant diameter	Date
			23 Apr	29 Apr	6 May	11 May	18 May	26 May	2 Jun	11 Jun	17 Jun	23 Jun	2 Jul	8 Jul	14 Jul	23 Jul
			Sun/cloud	Sun	Sun	Sun/cloud	Sun	Sun	Sun/cloud	Sun	Sun	Sun	Sun	Sun/cloud	Clouds	Sun
			Calm	W wind	Breeze	W wind	Calm	Calm	Breeze	W wind	Calm	Breeze	Breeze	Breeze	Calm	Calm
			15.7°C	19.5°C	20.0°C	20.1°C	19.1°C	20.0°C	20.9°C	25.0°C	20.6°C	24.0°C	25.7°C	24.9°C	23.3°C	25.9°C
			81.7%	44.8%	71.8%	53.9%	71.1%	72.6%	72.4%	46.9%	77.3%	69.2%	73.6%	78.1%	78.1%	54.9%
1	aspera	85	2/0	3/0	12/6	10/0	14/8	16/0	16/0	14/5	18/2	17/7	20/7	18/33	9/5	3/0
	seridis	40	0/0	1/1	1/1	3/0	5/0	4/1	3/0	1/0	2/0	0/0	0/0	0/0	0/0	0/0
	subdec	80	0/0	1/0	1/0	2/0	3/1	3/1	2/0	2/0	2/0	2/0	1/0	1/0	0/0	0/0
2	aspera	55	0/0	0/0	0/0	0/0	1/0	4/0	5/2	5/4	12/12	4/0	4/0	4/0	7/0	4/1
	seridis	45	1/0	1/1	3/1	4/0	5/11	10/18	5/8	4/5	2/1	3/1	0/0	0/0	0/0	0/0
	subdec	75	0/0	1/2	6/0	4/1	4/3	7/4	3/2	6/4	11/2	6/4	4/4	2/1	2/0	0/0
3	aspera	30	0/0	0/0	0/0	0/0	2/5	4/2	3/0	2/1	6/0	2/0	4/2	4/0	5/0	0/0
	seridis	60	1/0	7/2	6/0	7/11	11/11	14/0	5/1	10/17	2/1	1/1	0/0	0/0	0/0	0/0
	subdec	50	0/0	0/0	1/0	1/0	2/0	3/0	9/0	5/3	7/4	½	1/0	1/1	1/0	0/0
4	aspera	55	2/0	2/0	5/2	4/1	4/0	4/1	2/1	2/1	2/0	4/2	7/5	3/0	4/0	2/1
	seridis	110	4/1	0/0	7/7	8/5	14/4	18/20	23/39	18/20	17/16	4/4	1/0	0/0	0/0	0/0
	subdec	110	10/0	5/3	19/5	28/29	64/31	72/20	114/29	73/30	70/26	33/40	40/32	30/2	40/17	16/8
5	aspera	50	0/0	0/0	1/0	7/0	9/12	10/0	12/3	13/6	12/9	5/6	5/0	3/0	4/1	8/3
	seridis	60	1/0	4/0	2/0	4/1	6/7	2/3	4/1	2/2	1/1	1/2	0/0	0/0	0/0	0/0
	subdec	80	1/0	0/0	5/3	2/0	7/9	4/0	6/0	7/8	11/0	10/6	3/0	1/0	1/0	0/0
Total			22/1	25/9	69/25	84/48	151/102	175/70	212/86	164/106	175/74	93/75	90/50	67/37	73/23	33/13

Note: aspera: Centaurea aspera, seridis: C. seridis, subdec: C. ×subdecurrens.

Insect visitors had their peak number of visits at different times. The earliest peak was observed for R. sticticum, which had the most visits on 11 May (Figure 4). A week later, on 18 May, B. terrestris, A. plumipes, and P. viridicoerulea reached their peak visits. Visits by Halictus scabiosae and P. haustellatus peaked on 2 June, while O. funesta and P. lituratum reached their highest visitation on 11 June, followed by H. quadricinctus on 17 June. Lasioglossum discum reached its peak on 23 June, while R. infuscatum and H. crenulatus peaked in July. The length of the period of visits also varied among the pollinators. Some species, such as A. plumipes and H. quadricinctus, were observed for only 2 or 3 weeks, whereas others, like H. crenulatus, were present throughout the sampled flowering period (Figure 4).

Figure 4.

Number of visits along time of the insects that visited at least three times the capitula of the studied Centaurea plants growing in El Saler, with no differentiation of plant taxa.

Identification of factors that predicted the number of insect visits

Of the 19 linear models tested to predict number of visits per plant, three did not meet the normality assumption for residuals (Table 3). These three models included only discrete predictive variables (plant taxon, date, and block). Among the remaining models, the one with the lowest AIC and BIC values was lm5, which predicted the log (number of visits) using a single predictor: log (number of florets per plant) (Table 3). Model lm1, which used log (number of capitula), and model lm6, which used log (number of florets) + plant taxon, obtained very similar results.

Table 3.

Accuracy metrics for the tested regression models used to predict insect log (number of visits) on Centaurea plants in El Saler (Valencia): number of model parameters (k), corrected Akaike information criterion (AICc), ΔAICc related to the smallest AICc as percentage, Bayesian information criterion (BIC), coefficient of determination (R ²), and Shapiro Wilk test value (SW). Models are sorted in descending order of AICc.

Model	Variables	k	AICc	ΔAICc	BIC	R 2	SW
lm5	log (No. florets)	3	44.673	0,000	52.173	0.636	0.209
lm6	log (No. florets) + plant taxon	5	45.091	0,936	57.364	0.643	0.279
lm1	log (No. capitula)	3	45.650	2,187	53.150	0.632	0.161
lm8	log (No. florets) + block	4	46.514	4,121	56.424	0.633	0.378
lm7	log (No. florets) + date	4	46.564	4,233	56.474	0.633	0.292
lm13	log (No. florets)*plant taxon	7	46.580	4,269	63.431	0.647	0.167
lm10	log (No. florets) + plant taxon + block	6	46.884	4,949	61.471	0.641	0.465
lm9	log (No. florets) + plant taxon + date	6	47.312	5,907	61.898	0.639	0.348
lm15	log (No. florets)*block	5	48.063	7,588	60.336	0.632	0.396
lm11	log (No. florets) + date + block	5	48.474	8,508	60.746	0.630	0.458
lm14	log (No. florets)*date	5	48.786	9,207	61.059	0.629	0.290
lm12	log (No. florets) + plant taxon + date + block	7	49.174	10,075	66.024	0.637	0.485
lm17	log (No. florets)*plant taxon*block	13	54.556	22,123	83.827	0.648	0.302
lm18	log10(No. florets)*date*block	9	54.849	22,779	76.068	0.626	0.582
lm16	log (No. florets)*plant taxon*date	13	57.193	28,026	86.464	0.639	0.253
lm19	log (No. florets)*plant taxon*date*block	25	73.275	64,025	119.843	0.663	0.293
lm4	block	3	141.617	217,008	149.116	0.021	<0.001
lm2	plant taxon	4	143.045	220,205	152.954	0.018	<0.001
lm3	date	3	144.428	223,300	151.928	–0.008	<0.001

The formula for the best model was (Figure 5): Number of visits = 0.0468264 × Number of florets^0.8205434.

Figure 5.

Best double logarithm linear model indicating the direct effect of the number of florets in each Centaurea plant on the number of insects visits. In the formula y = log (number of visits) and x = log (number of florets). Each point represents the log (number of visits) for each date and plant. Different colors represent the different plant taxa, n.flowers is number of florets, and n.visits is number of visits.

This model revealed a significant relationship (P < 0.001) between log (number of visits) and log (number of florets), with both the intercept and log (number of florets) showing statistically significant effects (Table 4).

Table 4.

Significance of the regression model that best predicted the log (number of insect visits) on Centaurea plants growing in El Saler, Valencia using the variable log (number of florets per plant), which had the lowest values of corrected Akaike information criterion (AICc) and Bayesian information criterion (BIC).

Model parameters	Estimate	SE	t	P	df	Adjusted R 2
Intercept	–1.32951	0.14938	–8.90	<0.001
log (No. florets)	0.82054	0.06285	13.05	<0.001
Residual		0.2974			93
Model						0.3659
Model F				<0.001

Linear regressions predicting log (number of visits) for each pollinator, using log (number of florets) as the predictor, were performed for insects that visited Centaurea plants at least 10 times during the flowering period. Four of the eight insects showed a significant association: B. terrestris (P = 0.004), H. scabiosae (P < 0.001), H. crenulatus (P = 0.025), and R. sticticum (P = 0.006). The remaining four insects, Halictus quadricinctus, Lasioglossum discum, Pseudoanthidium lituratum, and Rhodantidium infuscatum, showed no significant association (Appendix S3). Similar results were obtained when log (number of capitula) was used as the predictor instead of log (number of florets), with the same insect species showing significant associations (not shown).

Relationship between insect visits and plant taxon

Differences in number of insect visits per capitulum and floret among Centaurea taxa

According to its larger capitulum size, a single capitulum of C. seridis attracted more visitors than predicted by model lm1. Conversely, C. aspera capitulum attracted fewer visitors than expected, while that of C. ×subdecurrens closely matched the linear regression predictions (Appendix S4). However, differences among plant taxa were not statistically significant (Kruskal–Wallis test P = 0.060). When considering florets instead of capitula, C. seridis, which had the highest number of florets per capitulum, received the expected number of visits per floret according to lm5. Centaurea aspera received more visits than predicted, while C. ×subdecurrens received fewer visits than expected (Appendix S4). Again, differences among plant taxa were not significant (Kruskal–Wallis test P = 0.181).

Differences in flower visitor composition among Centaurea taxa

Visitation biases for a given Centaurea taxon in the eight species of insects that visited Centaurea plants at least 10 times during spring are shown in Figure 6. Halictus scabiosae and H. crenulatus accounted for 66.7% of the visits to C. aspera (56 and 43 visits, respectively, of 153 total visits). Halictus scabiosae accounted for 58.8% of the visits to C. seridis, while Bombus terrestris accounted for 24.1% (127 and 52 visits, respectively, of 216 total visits). For C. ×subdecurrens, H. scabiosae and B. terrestris made up 55% of the visits (92 and 85 visits, respectively, of 322 total visits).

Figure 6.

Relation between Centaurea taxa growing in El Saler and insect species that visited Centaurea plants at least 10 times during the flowering period. The width of the curves that relate taxa correspond to the number of visits (frequency).

When considering the pollinators, B. terrestris most frequently visited C. ×subdecurrens (85 of 148 visits or 57.4% of its visits). Similarly, H. quadricinctus, P. lituratum, R. sticticum, and R. infuscatum most commonly visited C. ×subdecurrens (32 of 37 visits or 86.5%; 15 of 21 visits or 71.4%; 44 of 66 visits or 66.7%; and 15 of 21 visits or 71.4%, respectively). Among them, H. quadricinctus and R. infuscatum did not visit C. seridis. Centaurea aspera was dominant in visitation frequency for L. discum (10 of 18 visits or 55.6%). Similarly, H. crenulatus most frequently visited C. aspera (46 of 102 visits or 41.5%), and to a lesser extent C. ×subdecurrens (32 of 102 visits or 31.4%). Finally, H. scabiosae primarily visited C. seridis (127 of 275 visits or 46.2%) followed by C. ×subdecurrens (92 of 275 visits or 33.5%), particularly as the number of capitula on C. seridis declined (Appendix S5).

The multivariate test revealed significant differences in flower visitor composition among the Centaurea taxa (P = 0.003). Pairwise comparisons indicated that the flower visitor composition of C. seridis was significantly different from both C. ×subdecurrens (P = 0.004) and C. aspera (P = 0.026), but no significant differences were found between C. aspera and C. ×subdecurrens. However, the multilevel pattern analysis only showed a significant association between P. lituratum and C. ×subdecurrens (P = 0.0431). There was also weak evidence of associations between R. sticticum and H. scabiosae with C. seridis (P = 0.0605 and P = 0.0606, respectively) and between B. terrestris and C. seridis and C. ×subdecurrens (P = 0.0926).

There was a significant interaction between Centaurea taxa and the number of capitula per plant on flower visitor composition (P = 0.022). Even when the interactions were not considered, the number of capitula per plant and the Centaurea taxa still had significant effects on flower visitor composition (P = 0.001 and P = 0.004, respectively). Visits of H. scabiosae were significantly influenced by the number of capitula per plant and its interaction with Centaurea taxa (P = 0.001 and P = 0.041, respectively), while visits of B. terrestris and H. crenulatus were only affected by the number of capitula per plant (P = 0.001 and P = 0.009, respectively).

In contrast, no significant interaction effect was found between Centaurea taxa and date on flower visitor composition. However, when the interaction was not considered, date had a strong significant effect on the insect composition (P = 0.001), and Centaurea taxa also had a significant effect when date was also excluded (P = 0.002). Specifically, date significantly influenced the visits of R. sticticum and R. infuscatum (P = 0.001 and P = 0.009, respectively).

Regarding the block, its interaction with Centaurea taxa significantly affected flower visitor composition (P = 0.002) as did block alone (P = 0.002) when the interaction was not considered, and Centaurea taxa alone (P = 0.003) when both the interaction and the block were excluded. Only visits of B. terrestris were significantly affected by the block (P = 0.003), likely due to the higher number of capitula produced by C. ×subdecurrens in block 4 compared to the other blocks (Table 2). Notably, on 18 May 2020, the day B. terrestris activity peaked, C. ×subdecurrens in block 4 received 25 visits from this bumblebee, while the remaining blocks collectively received only eight visits.

Finally, when all factors—Centaurea taxa, number of capitula per plant, date, and block—were included in the linear model, all significantly influenced flower visitor composition (P = 0.004, P = 0.001, P = 0.001, and P = 0.002, respectively). Univariate analysis revealed that visits of H. scabiosae, B. terrestris, R. sticticum, and H. crenulatus were significantly affected by the number of capitula per plant (P = 0.001, P = 0.003, P = 0.001, and P = 0.037, respectively). Visits of B. terrestris, R. sticticum, and R. infuscatum were significantly affected by the date of observation (P = 0.004, P = 0.002, and P = 0.004). Only visits by B. terrestris were affected by the block (P = 0.002).

DISCUSSION

In this study, we identified an assemblage of 17 insect species that varied in visiting behaviors to coexisting plants of diploid Centaurea aspera, allotetraploid C. seridis, and sterile triploid C. ×subdecurrens. The primary flower visitors were the bumblebee Bombus terrestris and several bee species, notably Halictus scabiosae. Number of florets per plant was the primary determinant of total insect visitation frequency, outweighing the effects of Centaurea taxon identity. Although overall visitation rates did not differ significantly among plant taxa, the composition of insect visitors varied, with C. seridis attracting a distinct pollinator community.

For a robust test of these visitation patterns, it was critical to use a methodological framework capable of capturing the temporal and behavioral complexity of insect visitation in natural populations. Accordingly, we combined long‐term field observations (98 days) during the overlapping flowering periods of the three Centaurea taxa with detailed taxonomic identification and quantitative recording of individual visits. Since some insects visited Centaurea plants within narrow time frames and at different periods, our long‐term observations allowed us to identify a high number of visitors and treat date as a statistical variable affecting visitation rates and assemblages. Considering this extended flowering period, it is also important to note that, due to the species’ own phenology, at the beginning of the period some individuals of C. aspera and, to a lesser extent, C. ×subdecurrens did not produce any capitula. Consequently, pollinators could only visit the other Centaurea taxa in the block. At the end of the period, this situation occurred with C. seridis and, to a lesser extent, again with C. ×subdecurrens. Our study contrasts with some previous studies, where observation periods were much shorter, ranging from 4 to 8 d (Husband and Schemske, 2000; Jersáková et al., 2010; Abrahamczyk et al., 2021; Oliveira et al., 2022). However, although the study was conducted over a single year—similar to most studies examining insect visitation in polyploid complexes (Segraves and Thompson, 1999; Husband and Schemske, 2000; Hirsch et al., 2003; Jersáková et al., 2010; Castro et al., 2020; Sutherland et al., 2020; Abrahamczyk et al., 2021; Oliveira et al., 2022; Schmickl et al., 2024)—multiseasonal studies would help further reinforce and validate our findings (Thompson and Merg, 2008; Gross and Schiestl, 2015; Roccaforte et al., 2015; Laport et al., 2021).

Moreover, challenges such as insect identification, limited observation periods, and the need for statistical simplification, have led others to categorize insects into broad groups, such as visitation time (e.g., noctuids and hawk moths) (Jersáková et al., 2010), pollination effectiveness (e.g., legitimate pollinators, occasional pollinators, nectar thieves) (Borges et al., 2012; Oliveira et al., 2022), or specialization (e.g., pollen specialists, generalists) (Laport et al., 2021). In other studies, only the presence of the pollinator was recorded but not their frequency (Borges et al., 2012). Our fine‐scale insect identification and visitation quantification provides a finer resolution of pollinator behavior, revealing ecological interactions and reproductive barriers that may be masked by broad functional categories (Blüthgen and Klein, 2011., Christie et al., 2022). This level of detail also improves the accuracy of community‐level analyses and facilitates comparisons across systems and years (Vasiliev et al., 2023). Moreover, it strengthens the ecological relevance of pollination studies, particularly in the context of biodiversity and conservation assessments (Diaz, 2024).

Diversity of pollinators visiting Centaurea taxa and its effect on mixed‐ploidy Centaurea population of El Saler

A total of 17 insect species visited Centaurea flowers in El Saler, a number comparable to or greater than that observed for other Centaurea species in the United States and Europe. For example, 18 species visited C. scabiosa, 17 visited C. diffusa, 13 visited C. maculosa, 13 to 20 visited C. solstitialis, and 5 to 14 visited C. jacea (Harrod and Taylor, 1995; Hirsch et al., 2003; De Groot, 2006; McIver et al., 2009; Leong et al., 2014; Abrahamczyk et al., 2021). The predominant flower visitors in these previous studies belonged to the same insect orders as those observed in El Saler: Hymenoptera (bumblebees of genus Bombus, the honeybee Apis mellifera, various other bee species, and wasps), Diptera (flies), Lepidoptera (butterflies), and Coleoptera (beetles). However, the species of pollinators varied depending on the Centaurea species and geographic location (Lack, 1982b; Harrod and Taylor, 1995; Hirsch et al., 2003; Kirchner et al., 2005; De Groot, 2006; Bilisik et al., 2008; Albrecht et al., 2009; McIver et al., 2009; Leong et al., 2014; Farris et al., 2018). Specifically, for Mediterranean habitats, our study aligns with previous research, in which high insect abundance throughout the season and short activity periods of many pollinators have been observed (Bosch et al., 1997; Farré‐Armengol et al., 2015). The three Centaurea taxa—diploid C. aspera, tetraploid C. seridis, and triploid C. ×subdecurrens—attracted a wide range of insect species, consistent with previous findings that C. aspera shows a low level of selectiveness (Bosch et al., 1997; Cortés‐Fernández et al., 2022).

In El Saler, the primary flower visitors of Centaurea included Bombus terrestris, Halictus scabiosae, Heriades crenulatus, Rhodanthidium sticticum, and other bee species such as Lasioglossum discum, Pseudoanthidium lituratum, and R. infuscatum. These species were recorded for the first time in La Albufera Natural Park, and L. discum, H. crenulatus, P. lituratum, and R. infuscatum are also newly documented in the Valencian Community (BDBCV, 2024). Our findings also confirmed the presence of both generalist and specialist pollinators, with species like Pseudoanthidium lituratum visiting only Centaurea and Cirsium species, and species belonging to Halictus and Lasioglossum collecting nectar and pollen from various plant families (Aguado et al., 2015; Romero, 2020).

Flower visitors observed on C. aspera in El Saler overlapped partially with those observed on the same species in other Mediterranean dune habitats. In the Balearic Islands, five species were identified, with Oxythyrea funesta and Bombus terrestris being the only coincident insects (Cortés‐Fernández et al., 2022), while in Catalonia some common shared flower visitors included Lasioglossum, Halictus, Bombus, and Anthophora species (Bosch et al., 1997). However, in El Saler, we did not observe Apis mellifera, which was prominent in studies from the Balearic Islands and Catalonia, likely due to the absence of managed honeybee colonies in the study area.

Despite the high diversity of insect visitors, four species—Halictus scabiosae, Bombus terrestris, Heriades crenulatus, and Rhodanthidium sticticum—accounted for more than 75% of the visits, which is consistent with the common pattern of a few dominant species and several minor ones in studies in natural conditions (Lack, 1982b; Segraves and Thompson, 1999; De Groot, 2006; Thompson and Merg, 2008).

Factors that contribute to insect visiting

We expected that number and size of capitula per plant may affect number of insect visits. Results showed that the best linear model predicting the number of insect visits per plant included only the number of florets or capitula, with plant taxa, date, block, and their interactions providing no significant improvement in AICc and BIC. This finding aligns with our expectations as C. seridis produced fewer capitula than C. aspera, but they were larger and contained more florets, while floret size remained consistent between the two species (Ferriol et al., 2012), a pattern also observed in Tragopogon allotetraploids (Vamosi et al., 2007).

This result also aligns with findings in Chamerion angustifolium (L.) Holub, where autopolyploids received significantly more pollinator visits than diploids, likely due to their larger and taller inflorescences, bigger floral structures, and greater number of open flowers per inflorescence (Husband and Schemske, 2000; Kennedy et al., 2006). This increase of total display area per capitulum may increase attractiveness to pollinators and compensate for the lower number of capitula (Harder et al., 2004), as suggested by our results where the contribution of the number of florets and capitulaand the interaction of number of florets and plant taxon on AICc and BIC were very similar.

Our results may be related to the cues flowers use as attractants. Insects vary in their ability to perceive floral signals such as visual, olfactory, and tactile stimuli (Jersáková et al., 2010), with bees primarily relying on visual cues for long‐distance attraction (Lack, 1982a, b). In our study system, Centaurea flowers are uniformly pink and contrast sharply with the predominant yellow hues of the El Saler dunes. If pink coloration functions as an effective visual attractant, individuals bearing more capitula—and therefore a larger pink floral display—may have attracted more insect visitors. However, the presence of nearby Centaurea plants outside the experimental plots could also have influenced visitation patterns. Notably, we observed that many of these insects also visited Scabiosa atropurpurea L., which also has pink flowers (M. Ferriol and A. Garmendia, personal observations). This interesting issue warrants further investigation. Moreover, our study could not fully disentangle evolved interactions from functional foraging patterns, highlighting the need for additional research in this area.

Relationship between insect visits and Centaurea taxon

Although we expected that both the number of insect visits and the composition of the insect assemblage might vary among Centaurea taxa, the number of visits per floret or capitulum did not differ significantly between plant taxa. However, insect assemblages to C. seridis differed slightly but significantly from those to C. aspera and to C. ×subdecurrens. Notably, H. scabiosae, the most frequent visitor, had a visitation bias toward C. seridis. However, as flowering of C. seridis began to decline, H. scabiosae mostly visited C. ×subdecurrens. In contrast, the less common R. infuscatum, P. lituratum, L. discum, and H. quadricinctus visited C. aspera and C. ×subdecurrens but not C. seridis. Iberian Halictus and Lasioglossum species are known to be polylectic, while R. infuscatum shows some preferences, and P. lituratum is oligolectic (Aguado et al., 2015; Romero, 2020). These findings suggest a subtle yet potentially meaningful visitation biases of certain insect species toward specific Centaurea taxa. However, these visitation biases cannot be interpreted as ecological or evolutionary specialization, which would require further study (Armbruster, 2017).

In this study, we did not assess pollination efficacy, which depends on factors such as pollen‐carrying capacity, morphology, and foraging behavior. Some authors emphasize the importance of this evaluation because not all floral visitors contribute equally to pollination, and the same species can vary in pollination efficiency depending on the cytotype or plant taxon (Lindsey, 1984; Segraves and Anneberg, 2016). Pollination efficacy can be measured by seed production per visit, visitation rate, or a combination of both, but such assessments are challenging in the field (Thompson and Merg, 2008). Although previous studies indicate that halictids and solitary bees are effective pollinators of Centaurea (Harrod and Taylor, 1995; Kirchner et al., 2005), our results should be interpreted with caution.

While differences in insect visitation rates were linked to the number of florets or capitula, variations in insect assemblages among plant taxa may be influenced by sensory cues beyond color, such as floral scent (Chittka and Raine, 2006). Additionally, previously reported morphological differences among Centaurea taxa (Ferriol et al., 2012) could influence pollinator attraction. The larger capitula and longer floral structures (stamens, anthers, and pistils) of tetraploid C. seridis compared to diploid C. aspera may have affected interactions with pollinators and other insect visitors, as observed in other polyploid complexes (Segraves and Thompson, 1999; Vamosi et al., 2007).

Similar patterns have been documented in Heuchera grossularifolia Rydb. (Thompson and Merg, 2008), Libidibia ferrea (Mart. ex Tul.) L.P. Queiroz (Borges et al., 2012), and Larrea tridentata (DC.) Coville (Laport et al., 2021); diploids and tetraploids were visited by the same species of floral visitors, but with some significant differences in the frequency of visitation rates to the different plant taxa. However, our findings contrast with two other trends observed in polyploid complexes. The first is a lack of preference for a given cytotype, as reported in Chamerion angustifolium (Husband and Schemske, 2000), Aster amellus L. (Castro et al., 2011), Gladiolus communis L. (Castro et al., 2020), and Gymnadenia conopsea (L.) R.Br. (Jersáková et al., 2010). The second is strong differentiation in insect assemblages with little overlap between cytotypes, as observed in Erythronium spp. (Roccaforte et al., 2015). All of these polyploid complexes involve autopolyploids, with the exception of Erythronium albidum Nutt., which is likely an allopolyploid. These results suggest that allopolyploidy may have a greater impact on pollinator differentiation relative to related diploids than autopolyploidy does. However, further research is needed due to the limited number of studies focused on allopolyploids.

In El Saler, strong postzygotic barriers exist between diploid C. aspera and tetraploid C. seridis, as evidenced by the sterility of their triploid hybrids (triploid block). However, the subtle flower visitor biases observed for certain Centaurea taxa also suggest the presence of prezygotic barriers. The differentiation in insect assemblages between C. seridis and C. aspera or C. ×subdecurrens, but not between the last two, may help C. seridis mitigate the minority cytotype exclusion effect. This effect is further reinforced by the self‐compatibility of C. seridis, in contrast to the strictly allogamous diploid C. aspera. As a result, hybridization is asymmetric, with C. aspera consistently acting as the maternal parent, leading to more energy wastage and competition for space with C. ×subdecurrens, since Centaurea achenes disperse over short distances (Ferriol et al., 2015; Garmendia et al., 2018). These three factors—self‐compatibility, reduced competition for space, and lower competition for pollinators with sterile triploid hybrids—may have played a crucial role in the long‐term persistence of C. seridis.

CONCLUSIONS

This study highlights that while visitation rates did not differ significantly, floral visitor composition varied among Centaurea diploid, allotetraploid, and triploid related taxa, suggesting that subtle shifts in insect assemblages may contribute to prezygotic reproductive isolation. The observed visitation biases of key pollinators such as Halictus scabiosae and Bombus terrestris toward certain Centaurea taxa illustrate how pollination networks can respond to floral trait variation associated with polyploidy and hybridization, particularly differences in capitulum size and number per plant. Our documentation of these dynamics in a natural mixed‐ploidy system underscore the ecological role of pollinators in maintaining reproductive boundaries and facilitating the coexistence of closely related plant taxa.

AUTHOR CONTRIBUTIONS

M.F. and A.G. conceived and coordinated the study. M.F., A.G., and H.M. acquired funding. M.F. and A.G. contributed to the fieldwork. A.G. and P.L. performed the statistical analysis. A.G. and P.L. interpreted the analyses. M.F. wrote the manuscript and prepared the figures and tables. All authors discussed and edited the manuscript and approved the final version.

Supporting information

Appendix S1. Location of Centaurea individuals studied in El Saler, Valencia, Spain.

Table S1. Geographical coordinates of individuals of C. aspera, C. seridis, and C. ×subdecurrens.

Figure S1. Map of location of individuals in Table S1.

Appendix S2. Distance from each studied plant growing in El Saler to the nearest individuals of each Centaurea taxon not included in the same block.

Appendix S3. Effect of the log (number of florets per plant) on the log (number of visits) of each insect species that visited Centaurea plants at least six times during the flowering period in El Saler (Valencia, Spain).

Appendix S4. Violin plots representing magnitude of residuals of lm1 (top) and lm5 (botttom), in which log (number of visits) was predicted using log (number of capitula per plant) and log (number of florets per plant) respectively, for each Centaurea taxon growing in El Saler (Valencia, Spain). Red points represent means.

Appendix S5. Number of visits over time by insects that visited the capitula of Centaurea aspera, C. seridis, and C. subdecurrens at least three times, with plant taxa differentiated.

Garmendia, A. , Merle H., Lucio‐Puig P., and Ferriol M.. 2025. Insect visitation patterns in diploid Centaurea aspera and its related allotetraploid and triploid hybrids: similar rates but distinct assemblages. American Journal of Botany 112(9): e70103. 10.1002/ajb2.70103

DATA AVAILABILITY STATEMENT

All the data collected in this research and R scripts can be found in Zenodo: https://doi.org/10.5281/zenodo.14905080.