Journal list menu

Volume 104, Issue 4 e4004
REPORT
Open Access

Flowering phenology influences butterfly nectar foraging on non-native plants in an oak savanna

Stephanie A. Rivest

Corresponding Author

Stephanie A. Rivest

Department of Biology, University of Ottawa, Ottawa, Ontario, Canada

Correspondence

Stephanie A. Rivest

Email: [email protected]

Search for more papers by this author
E. M. Wolkovich

E. M. Wolkovich

Department of Forest & Conservation Sciences, University of British Columbia, Vancouver, British Columbia, Canada

Search for more papers by this author
Heather M. Kharouba

Heather M. Kharouba

Department of Biology, University of Ottawa, Ottawa, Ontario, Canada

Search for more papers by this author
First published: 17 February 2023
Citations: 1
Handling Editor: Joseph B. Yavitt

Funding information: Natural Sciences and Engineering Research Council of Canada; University of Ottawa

Abstract

The negative impacts of non-native species have been well documented, but some non-natives can play a positive role in native ecosystems. One way that non-native plants can positively interact with native butterflies is by provisioning nectar. Relatively little is known about the role of phenology in determining native butterfly visitation to non-native plants for nectar, yet flowering time directly controls nectar availability. Here we investigate the phenological patterns of flowering by native and non-native plants and nectar foraging by native butterflies in an oak savanna on Vancouver Island, British Columbia, Canada. We also test whether native butterflies select nectar sources in proportion to their availability. We found that non-native plants were well integrated into butterfly nectar diets (83% of foraging observations) and that visitation to non-natives increased later in the season when native plants were no longer flowering. We also found that butterflies selected non-native flowers more often than expected based on their availability, suggesting that these plants represent a potentially valuable resource. Our study shows that non-native species have the potential to drive key species interactions in seasonal ecosystems. Management regimes focused on eradicating non-native species may need to reconsider their aims and evaluate resources that non-natives provide.

INTRODUCTION

The negative impacts of non-native species on ecological communities are well documented (e.g., Bradley et al., 2019). However, some non-native species can play a positive role in the persistence of native species and even contribute to achieving conservation goals. For example, non-native species can provide food and habitat to native species, act as ecosystem engineers, and provide ecosystem services (Schlaepfer et al., 2011). As more non-native species are introduced into biological communities, interactions with native species are predicted to increase (Pearse & Altermatt, 2013). Therefore, evaluating the roles that non-native species play in ecosystems is critical. In environments impacted by human activities (e.g., urban areas), non-native species tend to be more dominant (e.g., plants; Pyšek et al., 2010), so their role in supporting native species in these habitats may be particularly important.

Native butterflies are likely to interact with non-native plants because they rely on plant resources during all life stages. Nectar provisioning is one understudied yet potentially important role of non-native plants since many adult butterflies rely on nectar as a source of nutrients and energy for flight and reproduction (Gilbert & Singer, 1975; Mevi-Schütz & Erhardt, 2005). Previous work has shown that native butterflies can forage for nectar on non-native plants (e.g., Bergerot et al., 2010; Thomas & Schultz, 2016), but it is not clear to what extent and under what conditions. Knowing how often native butterflies visit non-native plants for nectar is essential for maximizing the effectiveness of potential conservation actions.

Flowering phenology may be an important factor influencing native butterfly visitation to non-native plants since flowering time directly controls nectar availability, but few studies have addressed this. Theoretical predictions (Wolkovich & Cleland, 2011) and empirical evidence (Gerlach & Rice, 2003; Pearson et al., 2012) suggest that non-native plants can flower longer and/or at different times compared to natives. For native butterflies, this could mean there are times during their flight period when non-native flowers are more available than native flowers and/or that their flight periods could lengthen if nectar is available for longer. Yet only a handful of studies have explored phenological patterns in the availability of butterfly nectar resources (e.g., Szigeti et al., 2018), and to our knowledge, only one study specifically considered the flowering of non-native plants (Thomas & Schultz, 2016).

It is also unclear how butterflies would respond to differences in native and non-native flowering phenology if they were to occur. Since adult butterflies are typically generalist nectar feeders (Hardy & Dennis, 2008), they may forage based on nectar availability, regardless of plant characteristics or origin (i.e., null hypothesis from resource selection studies; Manly et al., 2002). Indeed, flower abundance can explain butterfly visitation for certain butterfly–plant species pairs (Grundel et al., 2000; Szigeti et al., 2018). Alternatively, butterflies may choose plants based on other factors (e.g., sugar quantity; Thomas & Schultz, 2016). As past work has focused on nectar resource selection by single butterfly species (Szigeti et al., 2018; Thomas & Schultz, 2016), we have a limited understanding of foraging patterns at the community level.

We investigate native butterfly nectar foraging in relation to the flowering phenology of plants in an oak savanna ecosystem and ask: (1) How much are native butterflies using non-native plants as a source of nectar? (2) How does phenology influence nectar availability? (3) How does phenology influence nectar foraging by native butterflies, and do they select nectar sources in proportion to their availability? For our third question, we predict that butterfly visitation to non-native plants will increase proportionally with non-native flower availability.

MATERIALS AND METHODS

Study area

We conducted field surveys of native butterflies and flowering plants in 10 sites of Garry oak savanna in Victoria, British Columbia, Canada (Appendix S1: Figure S1). Remnant patches of Garry oak savanna in this region are characterized by a diverse forb understory and an open canopy of trees including Quercus garryana (Garry oak). The climate is Mediterranean-like, with wet winters and dry summers ending in late-season drought (Klassen et al., 2015). Since 1840 when Europeans settled, human activities (e.g., urbanization, agriculture) have led to a 90% decline in the extent of Garry oak savanna in this region (Lea, 2006). Today, this ecosystem contains more than 100 species at risk (BC red list) including seven at-risk butterflies (BC Conservation Data Centre, 2021). It also contains a high diversity of non-native flowering plants (Lilley & Vellend, 2009), making it an ideal system to consider our research questions. Since we were primarily interested in phenological variation, we selected our 10 sites of Garry oak savanna to minimize differences in site-level factors (Appendix S2: Section S1).

Field surveys

We surveyed sites in rotation approximately once per week from 6 May to 12 August 2019, which represents the main period of butterfly activity in this region (Zand et al., 2017; H.M. Kharouba, unpublished). Weather conditions during the field season were similar to climate normals (1981–2010) (Appendix S2: Section S2). At the start of each visit to a site, two observers recorded butterfly nectar foraging for 60–90 min by walking a 300-m-long route divided into five transects of 60 m each arranged side by side 15 m apart. Since butterfly activity tended to be low, we prioritized areas of the transect where butterflies were active. Butterflies were identified on the wing and could be confidently resolved to species. Individuals were not distinguished as marking butterflies is time consuming and disturbs short-term behavior. However, when a single individual was observed, the number of foraging observations recorded was restricted to five (<30% of total observations). Following standard protocols (Pollard & Yates, 1993), surveys were conducted only when butterflies were fully active: 9 a.m. to 4 p.m., temperatures of 13–40°C or >17°C when cloud cover >75%, wind <30 km/h, and no rain.

Next, we estimated nectar availability from all flowering forbs and shrubs either on the same or following day as the butterfly surveys by placing a 1-m2 habitat quadrat at five locations along the same 300-m route described previously. For each site visit, quadrat locations were selected using an algorithm that randomized the distance from the start (0–300 m) and the side of the transect (left or right). However, we ensured the placement of quadrats adequately covered the floral resources present in a site. For example, if a quadrat was located on a rocky outcropping, we moved the quadrat forward on the transect until the next patch of flowers. Within each quadrat, we counted the number of open flowering units for each plant species. Depending on the species, a flowering unit could be a single flower, cluster of flowers (inflorescence), or branching stem (Appendix S1: Table S1). We assumed only open flowers produced nectar and that flowers that were shriveled or dying did not produce nectar. It is possible that some open flowering units may have been older and no longer producing nectar, which we were unable to detect without sampling nectar directly. Then we estimated nectar availability by multiplying the number of open flowering units counted with their mean surface area in square millimeters (mm2), thereby accounting for differences in flower size (Appendix S2: Section S3). We also determined plant species origin and whether each species was known to produce nectar rewards (Appendix S2: Section S3).

Statistical analyses

We structured our analysis into two sections and fit a series of generalized linear models (GLMs) in each. The same modeling approach was followed in both sections (Appendix S2: Section S4). All predictor variables were standardized by subtracting the mean and dividing by the standard deviation. All statistical analyses were performed in R version 4.2.0 (R Core Team, 2022).

Influence of phenology on nectar availability

To estimate trends in the phenology of nectar availability, we summed our floral area data across the five habitat quadrats within each site visit (n = 114) and analyzed the proportion of floral area from non-native plants compared to native plants. Given this was a continuous proportion, we fit GLMs with the beta probability distribution (Hastie, 2019; Wood, 2011) after rescaling the values to satisfy model assumptions (Appendix S2: Section S5). To understand phenological trends, we included day of year as a predictor. We also included two site-level covariates: site area and the proportion of the total flowering plant species richness across the season that was non-native. Although we minimized site-level variation in these factors through site selection, we included the covariates to account for any remaining variation.

Influence of phenology and availability on nectar foraging

To explore trends in the phenology of nectar foraging and test the hypothesis that native butterflies select nectar sources in proportion to their availability, we fit two models. We first combined observations of nectar foraging across butterfly species within each site visit. Nectar foraging was only observed during 58 out of a possible 114 site visits (n = 58). We analyzed the proportion of nectar foraging visits made to non-native flowers compared to native flowers. Since this was a discrete proportion with minor overdispersion (dispersion parameter ~4), we fit the GLMs with the quasibinomial probability distribution (Zuur et al., 2009). We included the predictors day of year, proportion of floral area from non-native plants compared to native plants, and two site-level covariates: site area and the proportion of the total flowering plant species richness across the season that was non-native.

To test our hypothesis, we used the same nectar foraging model as earlier but excluded times when butterflies did not have a choice between native and non-native flowers (i.e., when only non-native flowers were available; after 18 July). As such, we fit a second model on a reduced data set that only included observations before 18 July (Appendix S1: Figure S2b). After this point in the season, 100% of foraging visits were to non-native plants and relative availabilities of non-native floral area were ≥99% for 80% of our visits to sites (see Results).

RESULTS

Summary of descriptive patterns

Across the season, we observed nectar foraging on 1143 occasions by 14 native butterfly species from five families (Appendix S1: Table S2). Summed across the season, 83% (946/1143) of all foraging visits were to non-native plants. We found that 64% (9/14) of butterfly species visited non-native flowers most often, 29% (4/14) visited native flowers most often, and one species visited native and non-native flowers equally (Appendix S1: Table S2). However, some butterflies were rare in our sites (e.g., Papilio zelicaon [Anise Swallowtail] is more common in woodlands than in savannas), leading to lower confidence in our estimates for these species. Out of a potential seven at-risk butterfly species in this region, we only observed two: Erynnis propertius (Propertius Duskywing) and Coenonympha tullia insulana (Vancouver Ringlet). These two butterflies made 60% (121/201) and 67% (6/9) of their nectar foraging visits to non-native flowers, respectively (Appendix S1: Table S2).

We observed 81 species of flowering plants with nectar rewards from 23 plant families (Appendix S1: Table S1). Of these species, 55% (45/81) were non-native. The amount of floral area recorded for each plant species varied across species, sites, and the season (Appendix S1: Figures S3 and S4). Some plants were highly available: eight species (three native, five non-native) made up 87% of all floral area recorded (Appendix S1: Figure S5). Other plant species were rare: nine species (three native, six non-native) were only recorded in a single site visit (their inclusion in the analysis did not impact results; Appendix S2: Section S6). Among our sites, the total flowering plant species richness varied from 40% to 71% non-native (Appendix S1: Figure S1).

The number of foraging visits was unevenly distributed among flowering plants with only 36% (29/81) of available plant species being visited by butterflies, 59% (17/29) of which were non-native (Appendix S1: Figure S6). Indeed, 71% (812/1143) of all foraging visits were to just four flowering plant species including the non-native species Hypochaeris radicata (hairy cat's ear), Rubus armeniacus (Himalayan blackberry), and Vicia sativa (common vetch), as well as the native species Camassia quamash (common camas) (Appendix S1: Figure S6).

Influence of phenology on nectar availability

The availability of native and non-native flowering plants fluctuated across the season. While non-native floral area stayed at relatively low, but constant, levels across the season, native floral area started the season at high levels and then declined over time until eventually reaching zero in several of our sites (Appendix S1: Figure S2a). In the last 3–4 weeks of our surveys (18 July–12 August), we recorded relative availabilities of non-native floral area that were ≥99% in 80% of our visits to sites (31/39 site visits).

The proportion of non-native floral area was higher in sites with a higher proportion of non-native plant species, but the magnitude of this effect was dependent on day of year (χ2 = 4.6, df = 1, p = 0.032, n = 114 site visits; Figure 1a) (Appendix S1: Table S4). Later in the season when native floral area declined, all sites had high relative availabilities of non-native floral area, regardless of the proportion of non-native plant species richness (Figure 1a). The proportion of non-native floral area was also higher in larger sites (β = 0.21 [0.10 SE], χ2 = 4.1, df = 1, p = 0.043, n = 114 site visits; Figure 1b) (Appendix S1: Table S4).

Details are in the caption following the image
Effects of significant predictors (x-axes) on (a, b) proportion of non-native (NN) floral area using beta generalized linear models (GLMs) (n = 114 site visits) and (c, d) proportion of nectar foraging visits made to NN flowers using quasibinomial GLMs (n = 58 site visits). Shown are raw data (points), lines of best fit (solid lines), and 95% CIs (a: ribbons; b–d: dotted lines). Also shown are model estimates (β), SE, and p-values from either (a, b) likelihood ratio tests (χ2-value) or (c, d) F-tests (F-value). Panel a shows the interaction between proportion of NN plant species richness (SR) and day of year. To visualize the interaction, the season was divided into equal thirds (early = purple squares; mid = green triangles; late = yellow circles), and model predictions were obtained for the midpoint of each third (22 May = purple lines; 24 June = green lines; 27 July = yellow lines).

Influence of phenology and availability on nectar foraging

The proportion of nectar foraging visits made by native butterflies to non-native flowers compared to native flowers increased later in the season (β = 1.4 [0.5 SE], F1,55 = 10.5, p = 0.002, n = 58 site visits; Figure 1c) (Appendix S1: Table S4). During the beginning of the season, butterflies visited both native and non-native flowers, whereas in the last 3–4 weeks of field surveys, 100% of foraging visits were to non-native flowers (Appendix S1: Figure S2b). The proportion of nectar foraging visits made to non-native flowers was also higher in sites with a higher proportion of non-native plant species richness (β = 1.6 [0.4 SE], F1,55 = 23.1, p < 0.001, n = 58 site visits; Figure 1d) (Appendix S1: Table S4).

When we tested butterfly selection of nectar sources when both native and non-native flowers were available (i.e., before 18 July), we found that native butterflies increased their visitation to non-native flowers when the relative availability of non-native flowers increased (β = 1.1 [0.4 SE], F1,31 = 11.0, p = 0.002, n = 33 site visits; Figure 2) (Appendix S1: Table S4). However, this increase was not in proportion to availability as predicted by the null hypothesis. Instead, butterflies selected non-native flowers more often than expected based on their availability (i.e., predicted curve is above null expectation curve; Figure 2). For example, 50% of foraging visits were to non-natives when only 27% of floral area was non-native (Figure 2).

Details are in the caption following the image
Test of null hypothesis that native butterflies forage for nectar in proportion to nectar availability. Presented is the effect of the proportion of non-native (NN) floral area on the proportion of nectar foraging visits made to non-native flowers using quasibinomial generalized linear models (n = 33 site visits). Shown are raw data (dots), the line of best fit (solid line), the 95% CI (dotted black lines), and the prediction from the null hypothesis (dotted red line), which was derived by fitting a quasibinomial model to data where foraging was in perfect proportion to availability. Also shown are the model coefficient (β), SE, and p-value from an F-test (F-value). Model was fit using only observations before 18 July (see Materials and Methods).

DISCUSSION

Relatively little is known about how differences in native and non-native flowering phenology can influence butterfly nectar foraging, yet this knowledge is essential to maximize the effectiveness of conservation actions. Our study contributes three key findings. First, we show that across the entire season, non-native nectar was well integrated into native butterfly diets (83% of all foraging observations). Second, we found that seasonal shifts in flower availability had a large impact on butterfly nectar foraging patterns: Visitation to non-native plants increased later in the season when native plants were no longer flowering. Lastly, we show that when both native and non-native flowers were available, butterflies visited non-native plants more often than expected, suggesting that these plants represent a potentially valuable resource.

Across the entire season, we found that non-native flowers were a heavily used source of nectar for native butterflies. Our estimate of usage may even be an underestimate since our sites were adjacent to suburbs, public gardens, and other urban land uses that could have provided additional non-native nectar sources that were not measured here. Even still, our low-end estimate is higher than previous reports, which documented low to moderate usage of non-native plants (i.e., 20%–50% of foraging visits to non-natives [Bergerot et al., 2010; Hardy & Dennis, 2008; Thomas & Schultz, 2016], but see 71% of visits in Jain et al. [2016]). We may have found higher estimates of non-native usage than were found in other studies due to site-level characteristics, such as the high species richness of non-native flowering plants that we found here. Differences across studies could also be due to a combination of site-level factors and resource selection since we know butterflies can select certain resources over others despite patterns in resource quantity (e.g., Thomas & Schultz, 2016). Future studies should examine how site-level factors, in particular land-use history (e.g., urbanization), influence the usage of non-native nectar by butterflies.

Non-native flowers were the main source of nectar in the last 3–4 weeks of the butterfly flight period due to a seasonal decline in native flowers, and butterflies responded by increasing their visitation to non-native plants during this time. This result is consistent with past work showing that insects can increase their visitation to non-native flowers when they are relatively more available (e.g., Salisbury et al., 2015). Our results also support previous work done in this system (Simon et al., 2021) and in others (Gerlach & Rice, 2003; Pearson et al., 2012) showing that non-native plants flower later in the season than native plants. However, given that our work is based on a single year, future studies could determine whether the patterns we observed here are sensitive to interannual variability (e.g., weather, population dynamics).

Since late-season plant community dynamics in Garry oak savannas and other Mediterranean-like ecosystems are driven by drought (Klassen et al., 2015), flowering later than natives could be a result of increased drought tolerance in certain non-native plant species. For instance, in this system, non-native floral area was found to be negatively correlated with soil moisture (Simon et al., 2021). Since historical data on flowering times in Garry oak savannas are unavailable, it is unknown whether non-native plants are occupying vacant temporal niches and/or if they have become more abundant than native plants that were historically present late in the season. Current plant species lists compiled for this ecosystem indicate that 10 native plant species are expected to flower, at least in part, in August (GOERT, 2023). We only observed one of these 10 species (i.e., Symphoricarpus albus [common snowberry]), while a previous study done at the same sites observed three (Lilley & Vellend, 2009). More work is needed to determine what changes are occurring in the native plant community. Regardless, the later flowering of non-native plants relative to native plants could mean that the nectar foraging period, and thus the flight season, for butterflies has lengthened with the introduction of non-native plants. This could allow some late-season butterfly species (e.g., Ochlodes sylvanoides) to accumulate more resources and produce additional generations, which could accelerate population growth (Altermatt, 2010). Whether this occurs will also depend on late-season environmental conditions remaining suitable for butterflies and their larval host plants.

We found that butterflies selected non-native flowers more often than expected based on non-native flower availability. This suggests that non-native nectar is a valued resource and that butterflies may prefer non-native flowers over native flowers in some contexts, though a choice experiment could be used to determine this with greater confidence. Our results are similar to a single-species study that found ovipositing females preferred non-native plants (Euphydryas editha; Haan et al., 2021) but are unlike another single-species study that found that nectar foraging adults preferred native flowers (Plebejus icarioides fenderi; Thomas & Schultz, 2016). This reinforces the notion that studies of multiple species offer insights into general tendencies of the community. Of the top five most visited plant species, three were also among the top five species that provided the greatest amounts of total floral area (Appendix S1: Figures S5 and S6). Yet other plant species that provided very high amounts of floral area (e.g., native Plectritis congesta) were not highly visited by butterflies, suggesting that factors other than flower availability are influencing foraging decision-making. Overall, our study suggests that non-native plants can be a valued resource for native butterfly communities.

Several possible factors could have influenced why butterflies chose non-native nectar sources so often, such as flower color and shape (Tiple et al., 2006), or nectar characteristics like sugar concentration (Pivnick & McNeil, 1985) and amino acid content (Mevi-Schütz & Erhardt, 2005). For example, we observed that native butterflies visited yellow flowers most often (52% of visits; 598/1143) (Appendix S1: Figure S7), which is consistent with past work (Grundel et al., 2000). Since most yellow flowering plant species in our study were non-native (75% of species; 9/12) (Appendix S1: Figure S7), this may help to explain why we observed such high rates of visitation to non-native plants. Future work could evaluate differences between native and non-native flowers to determine why butterflies are selecting non-natives so frequently.

We found that non-native flowers were well integrated into native butterfly nectar diets, but there was also variation across butterfly species in the degree of usage of non-native flowers. This is consistent with community-level work where interspecific variation in nectar foraging is thought to be influenced by differences in proboscis lengths among butterfly species (Bergerot et al., 2010). Interspecific variation could also be related to the timing of butterfly flight periods (Appendix S1: Figure S8). For instance, the high number of visits to non-native flowers by Ochlodes sylvanoides (woodland skipper) could be influenced by its high abundance in July and August, when native flowers were least available. In contrast, the high number of visits to native flowers by Celastrina echo (echo azure) could be explained by its high abundance in May and June, when native flowers were most available. To inform regional conservation practices, future studies could assess nectar foraging in relation to flower availability for individual butterfly–plant species pairs, especially for at-risk butterflies.

Although we showed that non-native nectar was well integrated into native butterfly diets, it could be of lower quality than native nectar and lead to negative effects on butterfly fitness. Since nectar quality differences between native and non-native plants have not been well studied, it is difficult to predict potential outcomes for butterflies. In theory, butterflies should select nectar sources that provide sufficient rewards of adequate quality. But previous work showed that butterflies could sometimes have maladaptive strategies (i.e., developmental traps), such as when females oviposit on non-native plants that are toxic to their larvae (Sands, 2008). This may occur because toxic non-native host plants are closely related to native hosts or because they possess similar chemical cues (e.g., glucosinolate profiles as cues for ovipositing Pieris napi macdunnoughii; Rodman & Chew, 1980). Alternatively, if non-native nectar is of similar or better quality than native nectar, then non-native plants may help buffer native butterflies against losses in native flowers. An important next step will be to evaluate the consequences of non-native nectar diets on native butterflies.

Our findings have important implications for the conservation of butterflies in seasonal invaded ecosystems. Our work shows that non-native flowers can be an important resource for butterflies throughout the season, and particularly at key times. This is at odds with current management regimes, including those for Garry oak savannas (e.g., Trowbridge et al., 2017), that typically recommend the eradication of non-native plants, which could leave butterflies and other pollinators without sufficient nectar resources, especially during periods of drought later in the season. To mitigate the removal of potentially important nectar resources, managers could combine their efforts with supplemental plantings of native flowers that provide nectar at the necessary times. We also caution that the presence of non-native plants may help to achieve conservation goals where native plants are less common or less likely to persist (e.g., disturbed habitats). Overall, our work highlights the need for evaluating the phenological patterns of non-native plants to better understand their role in the diets of native insects.

ACKNOWLEDGMENTS

We thank Annelise Grey for help in the collection of field data. We thank Patrick Lilley and Andrew Simon for consultation about the study system. We thank Jeremy Kerr, Joseph Bennett, and members of Kharouba Lab for guidance and advice. We are also grateful to the Albert Head Cadet Training Centre and the Districts of Esquimalt and Saanich for permission to perform research at their sites. Funding was provided to Heather M. Kharouba by the University of Ottawa and a Natural Sciences and Engineering Research Council of Canada Discovery Grant.

    CONFLICT OF INTEREST STATEMENT

    The authors declare no conflict of interest.

    DATA AVAILABILITY STATEMENT

    Data (Rivest et al., 2023) are available in Dryad at https://doi.org/10.5061/dryad.2547d7wvp.