Population trends of insect pollinators in a species-rich tropical rainforest: stable trends but contrasting patterns across taxa

Abstract

Recent reports of insect decline have raised concerns regarding population responses of ecologically important groups, such as insect pollinators. Additionally, how population trends vary across pollinator taxonomic groups and degree of specialization is unclear. Here, we analyse 14 years of abundance data (2009–2022) for 38 species of native insect pollinators, including a range of Coleoptera, Lepidoptera and Hymenoptera specialists and generalists from the tropical rainforest of Barro Colorado Island, Panama. We estimated population trends across taxonomic groups to determine whether specialist species with a narrower range of interacting mutualistic partners are experiencing steeper population declines under environmental change. We also examined the relationship between climate variables and pollinator abundance over time to determine whether differences in sensitivity to climate predict differences in population trends among pollinator species. Our analyses indicated that most pollinator populations were stable or increasing, with few species showing evidence of decline, regardless of their degree of specialization. Differences in climate sensitivity varied among pollinator species but were not associated with population trends, suggesting other environmental factors at play for tropical insect pollinators. These results highlight the need for long-term population data from diverse tropical taxa to better assess the environmental determinants of insect pollinator trends.

1. Introduction

Several long-term studies have reported global declines in the abundance and diversity of insects [1–3]. Climate change, habitat loss and pollution are significant drivers of this decline [4,5]. However, the impact of climate change on insect population dynamics is less well understood. Global warming is expected to increase the frequency and intensity of extreme climatic events and alter large-scale climate patterns, such as El Niño Southern Oscillation (ENSO), which affects seasonal rain and temperature patterns [6,7]. Tropical insects are thought to be living at or near their thermal optima and have narrower thermal tolerances than those at higher latitudes, suggesting that such changes in the intensity and frequency of ENSO events are likely to impact their populations significantly.

Insect-mediated pollination is a fundamental ecosystem service linked to angiosperm diversity, evolution [8] and ecosystem functioning in natural and agricultural systems [9] , with more than 80% of approximately 300 000 flowering plants worldwide being pollinated by insects [8,10]. Studies of insect-pollinator decline have often focused on temperate bees [11,12]. Yet, population trends of tropical insect pollinators are unclear, despite a recent global assessment of pollinator abundance concluding that tropical crop production may be jeopardized due to pollinator losses [13]. Furthermore, specialized plant–pollinator interactions are expected to be at the highest risk of decoupling due to environmental change, as the loss of one species in the interaction network may be hard to replace [14,15]. To understand which species or groups of tropical insect pollinators may be more vulnerable to climate change, measuring the effect of climatic variables on pollinator abundances is necessary. Long-term monitoring of population dynamics represents a valuable resource for understanding population trends [16,17], allowing us to visualize changes in insect abundance over time, relate observed trends to climatic events, and thus forecast future responses to different climatic scenarios.

This study quantifies and compares population trends and responses to climate across three major tropical insect pollinator taxa: Coleoptera, Lepidoptera and Hymenoptera. We used 14 years of abundance data in a tropical forest in Panama [17] to estimate the population trends of 38 insect pollinator species that vary in their degree of specialization. Here, we ask: (i) Do population trends in specialist and generalist pollinators differ? (ii) Can sensitivity to climate help estimate population trends in tropical insect pollinators? We predict that specialized pollinators will show steeper population declines since they are more vulnerable to climatic effects on their reduced number of mutualistic partners when compared to generalist species. In a warmer and wetter climate, we predict species that are more abundant in warmer or wetter years should show increasing population trends.

2. Material and methods

(a). Insect taxa and climate data

We used insect abundance data collected by the Smithsonian’s Tropical Research Institute’s (STRI) Arthropod Program during 2009–2022 (https://fgeoarthropods.si.edu/) on Barro Colorado Island (BCI), Panama. We considered 6 species of beetles (Coleoptera: Dynastinae, genus Cyclocephala), 13 species of butterflies within various families, 17 species of orchid bees (Hymenoptera: Apidae, genus Euglossa) and 2 nocturnal sweat bee species (Hymenoptera: Halictidae, genus Megalopta; electronic supplementary material, table S1). The abundance of Euglossa bees was assessed using cineole baits, while the abundance of Megalopta bees and Cyclocephala beetles was assessed using light traps [18]. Butterfly abundance was estimated using Pollard walks [19]. From all species of insect pollinators surveyed on BCI, we selected only species with (i) literature records as potential or verified pollinators or flower visitors and (ii) enough number of records in the STRI’s Arthropod Program database.

We defined pollinator specialization broadly based on our focal group’s interactions with their commonly visited plants. Megalopta bees were considered generalist flower visitors, collecting pollen and nectar from a wide range of flowers [20], and male Euglossa were considered specialists since they primarily visit orchid species to collect volatile compounds used in courtship [21,22]. Butterfly species are generalist flower visitors feeding on nectar from a wide range of flowers, while Cyclocephala beetles have specialized host–plant associations with flowers of early diverging angiosperm groups and dicotyledonous species [23,24]. These beetle species may be destructive or specialized pollinators, in exchange for feeding and mating sites and metabolic boosts from floral heat [23,25–27].

Climate data for BCI were obtained from the Smithsonian physical monitoring programme, including monthly total rainfall (mm) and monthly maximum averages of air temperature (°C), collected in the forest (1 m height) (https://biogeodb.stri.si.edu/physical_monitoring/research/barrocolorado). Additionally, we considered sea surface temperature anomalies (hereafter SST) from the Pacific Ocean in El Niño 3.4 region, obtained from the National Oceanic and Atmospheric Administration (https://www.ncei.noaa.gov/access/monitoring/enso/sst), as this large-scale climate variable is a general proxy for climate around the globe and has been shown to predict insect abundances more accurately than collections of individual climate variables [28,29].

(b). Statistical analysis

We estimated population trends using Bayesian generalized linear models with a negative binomial distribution, including a logarithmic link function using the stan_glm.nb function from the rstanarm package [30] in R v. 4.3.1 [31], following [32]. To estimate population trends over time, we modelled the abundance of each species as a function of the collecting year and month. We included month as a predictor to account for insect and climate seasonality. To estimate the associations between insect abundance and climate, we modelled each species’ abundance as a function of the previously mentioned climate variables. We used a Bayesian statistical approach, which incorporates continuous probabilities. Since our model coefficients were in the logarithmic scale, we exponentiated them to show the independent variables’ calculated multiplicative effect size in insect abundance. As a result, we obtained the percentage change in insect abundance for each unit change on the independent variable. We considered mean posterior probability and 95% credible intervals overlapping one (e.g. a coefficient equal to zero has an exponentiated value of one) as a stable population. A coefficient over one or below one was considered an increasing or declining population, respectively. We used a normal distribution as a prior and the log link function in all models.

We estimated the effects of specialization on temporal trends in abundance of our focal species. We also quantified the association between sensitivity to climate variables and population trends. For specialization, we categorized our focal species into specialists and generalists based on their natural history, with Euglossa bees [22] and Cyclocephala beetles [23,24] as specialists and Megalopta bees [20] and butterflies as generalists. Sensitivity to different climate variables was calculated by the mean of the posterior draws for SST, annual precipitation and average maximum temperature from our Bayesian model (available in electronic supplementary material, table S2). The sensitivity to climate variables quantifies how much the species abundance is expected to change for each unit increase of a climate variable. A linear mixed model used these extracted beta coefficients as predictors of population trends. We fitted two linear mixed-effects models to estimate temporal trends using the lmer function from the R package lme4 v. 1.1.34 [33]. The first model included the binary factor of specialization as a predictor, and the second model included sensitivity to SST, annual precipitation and average maximum temperature as continuous predictors. Both models included insect order as a random effect to account for difference among insect groups. Model diagnostics for all models were performed with the DHARMa package [34], and figures were generated using the ggplot2 package [35]. Additional descriptions of the methods are available in electronic supplementary material, methods.

3. Results

(a). Temporal trends

Generalist species, including Megalopta species and most butterflies (11 out of 13), showed stable trends over time (figure 1a,b ). However, the butterfly species Dryas iulia and Itaballia demophile showed a yearly increase of 8% and 5% in abundance, respectively. Most specialized Euglossa bee species (13 of 17) showed stable trends, but abundances of Euglossa variabilis and E. despecta increased by 9% and 6% annually, respectively. In contrast, E. igniventris and E. dodsoni abundances declined by 35% and 10%, respectively (figure 1a ). Of the six Cyclocephala species, C. brittoni showed a 35% increase in abundance, while the other five species were stable (figure 1c ). Individual time series for all species are detailed in electronic supplementary material, figure S1, and all model coefficients and credible intervals are available in electronic supplementary material, table S2.

Figure 1.

Population trends of (a) 19 species of bees, (b) 13 species of butterflies and (c) 6 species of Cyclocephala beetles during 14 years in BCI. Exponentiated means of the posterior probability (filled circles) and 95% credibility intervals (horizontal lines). The red dashed line in population trend = 1 (exponentiated value of zero) is a reference for stable population trend. Examples of species with increasing abundance (a(i),b(i),c(i)), and stable or decreasing abundance (a(ii),b(ii),c(ii)) over the study period. = Nocturnal bee species.

(b). Responses to climate variables

The abundances of 2 out of 13 butterfly species decreased with each 1°C increase in SST (Parides sesostris by 59% and Heliconius sara by 60%), while both Megalopta species increased with each 1°C rise in SST (Megalopta amoena by 37% and M. genalis by 38%). The abundance of all Cyclocephala species was stable in response to variation in SST. However, the abundance of 5 out of 17 species of Euglossa bees increased (figure 2a ). Cyclocephala stictica and C. carbonaria were the only species showing a slightly increased abundance in response to increased precipitation (0.05% and 0.04%, respectively). Conversely, the bees E. dressleri (0.04%) and E. mixta (1%), as well as the butterfly D. iulia (0.06%), showed a slight decrease in abundance in response to increased rainfall. The rest of the pollinator species were unaffected by precipitation (figure 2b ). Regarding the increased maximum air temperature measured in the forest, E. imperialis showed an increased abundance of 38%, while C. carbonaria, C. sparsa and M. amoena declined 67%, 82% and 35%, respectively. The remaining species appeared to be unaffected (figure 2c ). Model estimates for all species are available in electronic supplementary material, table S2.

Figure 2.

Effects of climate variables on the population trends of focal species: (a) sea surface temperature in the equatorial Pacific, (b) BCI rainfall and (c) maximum average air temperature in BCI forest. Butterfly species are indicated in orange, bees in blue and beetles in green.

(c). Temporal trend relationship to specialization and climate

We found no association between the response of temporal trends to specialization (χ ² = 0.49, β = 0.03, p = 0.48; figure 3a ). Similarly, we did not find a relationship between temporal trends and SST (χ ² = 0.03, β = 0.01, p = 0.86; figure 3b ), precipitation (χ ² = 0.38, β = −4.03, p = 0.53; figure 3c ) or maximum average temperature (χ ² = 1.4, β = −0.04, p = 0.22; figure 3d ).

Figure 3.

Relationship of the temporal trend in species abundance to (a) species specialization, (b) sensitivity to SST, (c) sensitivity to precipitation and (d) sensitivity to maximum temperature. Black lines show a smooth regression line from the linear mixed-effects model fitted values, and the grey shading shows the standard error.

4. Discussion

Our analysis represents one of the few long-term studies focusing on insect pollinators in a tropical rainforest. Contrary to the reports on insect decline from temperate regions, our results suggest that most of the populations of our focal species have been relatively stable over the 14-year study. Only 2 out of the 38 focal species showed a substantial decrease in abundance, and five appeared to increase. We expected that specialized pollinator species (Euglossa and Cyclocephala) would show steeper population declines due to the sensitivity of their specialized ecological relationships to environmental stressors compared to generalists, but our results do not support this hypothesis. Both specialist and generalist pollinator taxa included in our study revealed stable or increasing population trends. In addition, a species’ sensitivity to several influential climate variables (e.g. SST, annual precipitation and average maximum temperature) did not predict differences in trends across insect pollinator species in BCI. This suggests that differences in population trends among insect pollinators at BCI are not driven by changes in the climate but are likely driven by some other environmental factor. The variation in population trends and relationship with climate was larger within insect orders than across orders or flower specialization.

Relative to other climatic predictors, changes in SST had the greatest impact on the abundances of generalist and specialist bee species considered in this study. Increases in SST were associated with population abundance that was either unaffected by SST or increasing in 36 out of 38 pollinator species during the study period. Plant phenology responds to changes in SST, including flowering time, plant biomass growth and acquisition of nutrients [36]. On BCI during El Niño years (elevated SST), flower and seed production increases [37], which is likely beneficial to insect pollinators. Since bees store more resources and food when available [38], they may also benefit from seasons with extended flowering periods. Additionally, several studies have shown that herbivorous insects may benefit from strong El Niño events in central Panama [29,39], potentially affecting the early life stages of butterflies and beetles.

Increased rainfall appeared detrimental to the abundance of two butterflies and one Euglossa species, whereas Cyclocephala beetle populations were unaffected. Rain drastically reduces the activity of diurnal flying insects [40] as raindrops affect insect flight, but chemical and visual cues to locate flowers may also be inhibited [41]. In contrast, adult Cyclocephala beetles may be less sensitive to these effects as they often enter into the enclosed inflorescences of early diverging angiosperms (Nymphaeales, Magnoliales, Arecales, Pandales and Alismatales), where, in exchange for pollination services, they are rewarded with mating sites, food and heat [27]. Overall, an increase in maximum air temperature had a small, negative effect on pollinator species; nevertheless, insects have been shown to implement different thermoregulatory strategies, which may allow them to cope with increasing temperatures [42].

The observed trends are relevant to our focal species and collecting protocols and may not apply to other pollinators, such as moths, flies, other beetles or eusocial wasps and bees. Further, species abundance over time represents only one facet of population dynamics. For example, estimates of population trends based on male bees attracted to baits could be misleading, as high levels of diploid males may reduce effective population size, as shown for E. imperialis in central Panama [43].

Given that our analysis did not support climate as a major predictor of population trends, habitat loss and land use changes may represent the main drivers for the decline in insect populations, including pollinators [44–46]. A recent study in a tropical savannah in Brazil suggested that forest cover and habitat fragmentation predicted species loss and a decline in the abundance of orchid bees [47]. BCI has been a protected area for research since the early twentieth century, with little human disturbance since then [48]. These buffering conditions may mitigate the impact of anthropogenic insect stressors. Our study further concurs with Roubik et al. [49], who monitored the abundance of orchid bees in a lowland tropical forest in Panama for almost 40 years and concluded that 75% out of 33 species surveyed either showed an increase or were stable during the study period. While the abundance of most of our focal species appeared to be stable, this study shows how different insects with diverse ecology, taxonomy and degrees of pollinator specialization respond to temporal changes in climate differently. For this reason, studying multiple pollinator taxa and species with various life histories and ecologies is crucial, especially in tropical rainforests, where the diversity of plants and insect pollinators is higher than in other latitudes [50,51].

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

Observation records for all STRI’s Arthropod Program including focal species of this study are available [52]. The datasets, analytical scripts, supplementary methods and supplementary figures are available from the Dryad Digital Repository [53].

Electronic supplementary material is available online [54].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

E.B.: conceptualization, data curation, formal analysis, investigation, methodology, writing—original draft, writing—review and editing; G.P.A.L.: conceptualization, data curation, funding acquisition, methodology, writing—review and editing; D.S.-V.: conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, writing—review and editing; N.A.P.: conceptualization, formal analysis, investigation, methodology, validation, writing—review and editing; J.A.R.S.: data curation, writing—review and editing; F.P.: data curation, writing—review and editing; R.B.: data curation, writing—review and editing; Y.L.: data curation, writing—review and editing; Y.B.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This study was funded by STRI’s Arthropod Program and the Czech Science Foundation (GAČR 20-31295S to Y.B.). The Sistema Nacional de Investigación, SENACYT, Panamá support G.P.A.L., Y.B. and J.A.R.S.