Poli-Stability: a comprehensive assessment of stability in pollination communities

Species and ecological interactions are disappearing at alarming rates with unknown effects on key ecosystem functions basic for human well-being, such as pollination. This project aims to address one of the most relevant problems in ecology nowadays: how plant-pollinator communities respond to environmental changes. By bridging the classical divide between the empirical and theoretical frameworks to study ecological stability, and using as a case study detailed data on 12 communities in the Doñana natural reserve (southern Spain) across a gradient of landscape fragmentation monitored over seven years, this project will put forwards solutions to comprehensively quantify the response of pollination communities to environmental perturbations, and elucidate the mechanisms by which pollination communities withstand global change pressures and achieve different axes of stability.

Research is divided into three Work packages, through which we will advance towards a more comprehensive quantification of stability in plant-pollinator communities:

  • WP1 aims to characterise and quantify the stability of the empirical communities under study.
  • WP2 centres on the validation of mechanistic approaches: first measuring stability in model plant-pollinator communities, and then comparing it with the stability of empirical communities.
  • Finally, WP3 deals with Interaction structure and its consequences for stability. We will study it in two ways: first by looking at the effect of landscape fragmentation on stability, and then on the effect that the dynamical nature of the interactions has on the stability of the plant-pollinator communities.

For the duration of the project, I will be working inside the Bartomeus Lab In Doñana Biológical Station, Seville.

Keep tuned, this page will be regularly updated with the ongoing results of the project.

This project is funded by the European Union under the Marie Skłodowska-Curie grant agreement 101064340 to V.D.-G.

Methodology:

Our study sites are forest islands located in the Doñana Natural Area that look like this

The 12 sites correspond to forest islands following a fragmentation gradient in which plant-pollinator networks in the shrubland understory have been gathered from 2015 to 2021. Each year we sampled twice a month for at least seven rounds during the full flowering season, gathering data on plant-pollinator interaction frequency and flower abundance. Sites were at least 3 km apart, which is larger than the foraging distances of most pollinators45 and hence can be viewed as independent. In each site, and during each round, we walked a 100 m straight line for 30 minutes in which we wrote down every plant–pollinator interaction seen, differentiating between the number of pollinator individuals and the number of visits per individual. This resulted in a total of more than 250 hours of field observations.

Figure 1: The 12 sites in the PoliS project are distributed along a landscape fragmentation gradient.

Our theoretical framework is based on the structuralist approach. The main prediction of the structuralist approach is that the network of biotic interactions among species composing an ecological community determines their opportunities to coexist. Communities with larger coexistence opportunities, which in natural plant–pollinator communities correspond to more cohesive structures of biotic interactions (that is, nested)14, are probably more persistent because they can tolerate a larger difference in intrinsic growth rates (that is, performance) among species. Importantly, species’ performances are expected to be driven by the environment where they live, providing a phenomenological link with the current environmental conditions. The degree of structural stability of an ecological community (that we will refer to as 𝜔) can be rigorously assessed by coupling mathematical theory to population models that describe the dynamics of the species based on these two ingredients: species’ performance and their interactions. Hence, it is expected that the larger the range of species’ performance compatible with its persistence, the larger the environmental fluctuations that the community can withstand, and the more probable its persistence (Fig. 2).

Figure 2: Species’ persistence probability in model communities: 1) Empirical observations determine the mutualistic plant-pollinator interactions (𝛾) of our dynamical model. 2) From γ we obtain a matrix describing the intra-guild effects among pollinators (𝛼′), called “effective interaction matrix” (see Methods). 3) 𝛼′ is then used to quantify the structural stability (see Methods), which gives the average probability that any species in the community persists (𝜔(𝛼 ′ )). At the species level, the different distribution of growth rates compatible with a given species persistence (Anthophora bimaculata in this example) in sites A and B (colour coded) will determine the likelihood of that species persistence when reproductive rates are sampled randomly (r1, r2). Note that fewer values of r are compatible with species persistence in site B (i.e. it has a lower expected persistence probability, 𝜔𝑖 )

Results: Validating mechanistic approach

Following the structuralist approach, we investigated whether empirical observation of species persistence (that is, the average number of years across sites a species is observed in the community) is explained by the structural stability of the plant–pollinator communities emerging from the species interaction networks. In its probabilistic interpretation, structural stability represents the average probability that a randomly chosen species of the community persists despite varying its performance (Fig. 1). Therefore, we predict that larger values of structural stability promote higher persistence of a pollinator community. To test our main prediction across levels of biological organization (both community and species level) we recorded the following across the 12 independent communities: (1) the abundances of pollinator species over time, with which we quantified empirical persistence; and (2) the mutualistic network of plant-pollinator interactions, with which we quantified the structural stability (that is, the expected average persistence probability of a species in a given community, ω). According to our main hypothesis, we observe that those pollinator communities with higher opportunities for species to coexist (that is, with higher values of ω) strongly correspond with higher mean
pollinator persistence observed in the field (Spearman correlation coefficient, ρ, of 0.85), as shown in Fig. 2a. This result indicates that the structure of species’ interactions in the studied communities contains information on how species, on average, can persist under changing environments. In agreement with our main prediction, we also find a positive relationship between the expected pollinator persistence probability in the model (ωi) and the number of years that a pollinator is found in the field (ρ ranges from 0.37 to 0.65 depending on the site, and is always statistically significant), meaning that species that are more frequently expected to be locally extinct in the theoretical models also tend to have lower observed persistence in real communities. Some of the species more frequently predicted to be extinct are comparatively rare species.

Theoretical expectation versus empirical values. A) Expected average persistence probability of pollinators (𝜔) versus mean persistence observed in the field for the 12 sites in the study. The shaded area represents the 95% confidence interval of the regression estimate. B) Expected pollinator persistence probability (𝜔𝑖 ) versus the number of years this pollinator is present in the field. Each point represents one pollinator species in one site, and all 12 sites are plotted together but analysed independently. The dotted blue lines represent the 50th quantile regression for each of the 12 sites, Spearman’s rank correlation coefficient (𝜌) is shown for A, and its range depending on the site in B. Note the differences in the x axis between A and B: while A indicates an average persistence in the field between 1.5 and 2 years across species, B represents the number of years each pollinator species was observed in the field (1–6 years).

You can read the detailed results in our pre-print here!

Update!: See the video with our recent results!

Update!: These results have been published and can be accessed via the journal page, digital CSIC, or here.
The associated Database and Code to quantify the stability of ecological communities can be accessed via Zenodo:

DOI

Results: Landscape intensification & pollinator persistence

Interestingly, our results additionally suggests that landscape fragmentation matters for determining the observed variation in the expected persistence probability with communities in larger habitat patches tending to exhibit higher persistence (correlation ρ ≈ 0.6) and a more nested structure (ρ ≈ 0.7) than those in smaller patches, ), but not necessarily more diverse in species (Fig. 3).

Figure 3: Effect of landscape intensification (patch size) on expected pollinator persistence (𝜔), nestedness of the interaction networks (𝜈), and pollinator species richness (𝑁). The shaded area represents the 95% confidence interval of the regression estimate. Not significant correlations appear in a dashed line.

Fig.2 proves that the network of interactions seems to be very important in predicting the ability of pollinators to persist in the field. Following this idea, we also found that landscape determines pollinator persistence mostly through an indirect path, by changing the interaction structure between plants and pollinators (Fig.4).

Figure 4: Path analysis showing the direct and indirect effect of landscape intensification (% of cultivated land) on the average persistence of pollinators in the field. Red arrows depict negative effects, and black arrows positive effects. The number on the arrows shows the estimates of the model with the statistical significance in brackets (p-values).

More to come soon!

Results: Interaction flexibility affects pollinator persistence

In the last part of the project, we investigate how the re-organization of the plant-pollinator interaction network affects the ability of pollinators to persist in the field. For that we compare the network of year t, with that of year t+1, and quantify the interaction changes.

The plant-pollinator networks are highly dynamic, with most pollinators not being present in consecutive years. Only a relatively small number of pollinators are permanent (∼37%), meaning that on average roughly two-thirds of pollinators do not remain in the network from one year to the next. For plants, the situation was the reverse, with most plants (∼58%) remaining present through consecutive years (Fig.5C). Most inter-annual interaction changes are due to species turnover (∼60% between permanent and transient species and ∼20% between two transient species), while only ∼20% of the interaction changes are due to rewiring (Fig. 5D). This is in line with previous results showing that ~70% of interaction changes were due to species turnover. Despite this, most of the interactions are concentrated in the core formed by the permanent species (Fig. 5E).

Figure 5: C) Distribution of permanent species richness (i.e. number of species that remain two consecutive years, in deep green) and of transient species richness (in light green) for pollinators and plants in the annual networks in our study. D) Distribution of ecological interaction changes according to the permanent/transient nature of the interaction partners. E) Density of interactions (i.e. connectance) in the permanent species core (formed by the permanent species) compared to its value in the transient periphery (formed by transient species).

To learn how this interaction reorganization affects species persistence we compared the structural stability (𝜔) of the empirical networks with that of different null models and quantified the z-scores of the empirical networks. We found that, in general, the empirical re-organization is acting as a boost of pollinator persistence (Fig. 6), with values peaking around 1 and beyond. However, the empirical reorganization is not optimal, as we can find synthetic networks that have larger values of 𝜔 than those we can find in the empirical networks.

Figure 6: Distributions of Z-scores measuring how far from the mean of the null model lie the empirical value of structural stability for each year and site. Vertical linea area a help to the eye to identify Z=0 (continuous line), and Z=+-1 (dashed lines).

Finally, we search to find which are the drivers of this empirical reorganization in the core species. We found that the amount of rewiring in the core was highly correlated with the changes in abundance of the core species, and with the phenological changes in the core species (Fig. 7A and 7B). We also found that more changes in the persistent core were accompanied by more changes in the outer part of the net network (Fig. 7C).

Figure 7: Drivers of interaction rewiring. Number of interaction changes caused by rewiring vs A) changes in abundance of permanent species, B) amount of phenological changes in the permanent species, and C) number of interaction changes caused by species turnover. Pearson correlation coefficient and its p-value are indicated in the upper left corner of each panel. The shaded area represents the 95%confidence interval of the regression estimate (black lines).

More soon!

Outreach & Materials

Noche Europea de l@s investigadore/as 2022, 2023 & 2024

Feria de la ciencia 2024

With this interactive workshop, you can play with the web of life: Sink the net!
We have developed this interactive activity as a way to show how relevant are ecological interactions to maintain the biodiversity.

To celebrate this year biodiversity week we are organizing a Bio-Blitz in Tablada, take a look at the event!

More information in the event webpage: https://ic1.ugr.es/members/virginia/i-bioblitz-de-invertebrados-de-tablada/

Final Update!!

Overview of the project:

As project PoliS comes to an end, here is a graphical summary of the more relevant things we have learned during the 24th month of the action.