Species responses to global change in a community framework
A classical definition of ecological communities could be a definition from Robert Whittaker (Whittaker 1975), an assemblage of populations of different organisms that live in an environment and interact with one another, forming together a distinctive system with its own composition, structure, environmental relations, dynamic and functions, but here we simply define ecological communities as a collection of interacting species found in a particular location at a given time period (Morin 1999). Community ecology is the study of patterns and processes involving at least two species co-occurring in space and time and interacting (Morin 1999), a very broad definition that embraces a large part of Ecology4. Since communities are constituted from a set of species and a set of interactions among species, which depends on species co-occurrence in space and time, any changes in spatial or temporal distributions of species is supposed to affect ecological communities by affecting not only the species composition of ecological communities but also interactions among them. Reciprocally, because species are linked by interactions, communities affect eco-evolutionary trajectories of species, and thus the way that they respond to global change. Here I will first briefly review the current knowledge on the properties of ecological communities, mainly focusing on their stability and ecological dynamics. Then I will present what we know about how species responses to global change combine to affect communities. Finally, I will review our knowledge about how species responses to global change are affected by inter-specific interactions.
Community stability and structure
Community stability and the structure of ecological interaction networks
Based on Darwin’ statement on the struggle for existence among species, the competitive exclusion principle predicting species exclusion when ecological niche are too close (Volterra 1928; Gause 1934) laid the foundation stone of a key challenge of Ecology: understand what stabilizes ecological communities and prevent them from collapse. In this topic, few papers have stimulated as much debate and research in Ecology as Robert May’s paper in Nature in 1972, Will a Large Complex System be Stable?, at the origin of one of the most beautiful research story of Ecology, the ongoing Diversity-Stability debate. Based on a work from Gardner & Ashby, which shows by using numerical simulations that the probability of a system to be stable decreases with its connectance (i.e. number of realized links divided by the number of possible links) and its number of entities (Gardner & Ashby 1970), May completed this work by an analytical approach, showing that connectance and diversity predict pretty well the probability of stability of communities (May 1972).
These theoretical conclusions then seem to contradict the previous ones, which suggest that the more diverse the ecological systems, the less they are affected by invasions and extinctions (MacArthur 1955). In an attempt to resolve the issue, many empirical studies have been conducted and find contrasting results (McCann 2000; Ives & Carpenter 2007), but 69 % of the 64 studies reviewed by Ives & Carpenter (2007) have found a positive relationships between diversity and stability. However, such apparent contradiction between theoretical and empirical results is partly due to the diversity of stability concepts (Table I-1) rather than real oppositions among these results. While MacArthur, and more generally empirical studies, uses stability to refer to the variability in species abundances, May uses stability to refer to the probability to reach a stationary attractor (McCann 2000). Considering this mismatch between empirical and theoretical studies, it is almost impossible to merge both and to draw conclusions, that reemphasizes the need to statistically fit models to data5 (Ives & Carpenter 2007). The theoretical approach of stability of May also neglects an important aspect of the diversity-stability relationship: the feasibility, which is the fact that all species have strictly positive abundances at equilibrium. Considering only feasible equilibria in the analysis reverses the relationship found by May, leading to a positive relationship between diversity and stability (Roberts 1974). However, feasibility in diverse system also exhibit a strong negative relationships with diversity (Dougoud et al. 2018). Considering feasibility thus reinforce the diversity-stability debate rather than solve it (Dougoud et al. 2018) and focusing on the theoretical definitions of the stability, mainly equilibrium stability and resilience (Table I-1), the original question asked by May’s article remains unchanged: why are diverse ecological communities stable while theory predicts that they should not be?
A few elements of answer are given at the end of May’s article, which says: “within a web species which interact with many others […] should do so weakly […], and conversely those which interact strongly should do so with but a few species” and “our model multi-species communities, for given average interaction strength and web connectance, will do better if the interactions tend to be arranged in ‘block’” (May 1972). Indeed, Ecology has found some mechanisms promoting stability of complex ecological communities by remembering that Nothing in Biology Makes Sense Except in the Light of Evolution (Dobzhansky 1973) and that natural systems are evolved systems, not random associations of species. To do so, ecological communities are often represented by using networks, a mathematical representation of interacting species in which species are the nodes and interactions the links. First, stabilizing patterns have been found in the non-random distribution of interactions strengths among species (Yodzis 1981). Food webs are largely stabilized by an over-representation of weak interactions (McCann et al. 1998), especially in long loops (Neutel et al. 2002, 2007), and by the fact that interaction strengths are distributed following constraints determined by species body sizes (Otto et al. 2007; Jacquet et al. 2016). Second, echoing May’s prediction, empirical food webs tend to exhibit a high modularity, food webs exhibit a highly modular compartmentation of interactions, which are organized in blocks of interactions (Melián & Bascompte 2004; Fontaine et al. 2011). Despite opposite results found by abstract and theoretical models (Allesina & Tang 2012; Allesina et al. 2015) a majority of studies found that the structure of empirical networks, defined by the distribution of interaction strength and the organization of occurring interactions, provide higher stability than random ones (Kondoh 2008; Thébault & Fontaine 2010; Tang et al. 2014), allowing diverse but stable systems existing. However, those researches first focused only on trophic interactions, while many other kinds of interaction shape communities, such as mutualistic ones, in which both partners benefit from the interaction.
The study of mutualistic interactions in communities has been achieved using bipartite graphs to represent two distinct guilds of organisms interacting together (Figure I-4), instead of a representation of communities as undirected graphs as for food webs, allowing to represent several trophic levels (Bascompte & Jordano 2006). Bipartite graphs, or bipartite networks, can be represented as an interaction matrix, which differs from the classical adjacency matrix used for food webs, in which each column represents a species from the first guild, while each row represents a species from the second guild (Figure I-4). Using such representation, it has been shown that mutualistic networks are nested, a structure that occurs when specialist species tend to interact with a proper set of the species that interact with more generalist ones (Bascompte et al. 2003). In mutualistic networks, the analysis of network structure has mainly been conducted on pollination networks, although other kinds of networks have been considered, such as seed dispersal networks. This might be because the interaction among pollinators, mainly insects, birds and bats, and flowering plants imply very diverse groups in terms of species and because 87% of terrestrial plants, a basal trophic level of a majority of terrestrial ecosystems, depends on this interaction to reproduce (Ollerton et al. 2011). In contrast bipartite 15 antagonistic networks, which describe predation interactions, are less nested than expected and highly modular (Fontaine et al. 2011). Nestedness and modularity have been shown to promote stability (Thébault & Fontaine 2010) and diversity (Bastolla et al. 2009; Thébault & Fontaine 2010) of mutualistic and antagonistic networks respectively, showing that mechanisms stabilizing networks depends on the kind of interactions constituting them.
Figure I-4: Bipartite networks. (A) An example of quantified mutualistic interaction network: a bee-plant interaction network in a forest of the Colombian Caribbean (Flórez-Gómez et al. 2020). Extracted from Flórez-Gómez et al. (2020). (B) Schematic representation of perfectly nested (a) and (c) and modular (b) and (d) binary bipartite networks. (a) and (b) Matrix representation, where each row and column represents a species, and the intersections of rows and columns are black when the species interact. (c) and (d) Network representation, where each circle (or node) represents a species, which are connected by edges when the species interact. Extracted from Fontaine et al. (2011).
Traits and phenology as determinants of network structure
Since the community structure plays an important role in stability-diversity relationships, the linkage rules (i.e. rules defining species interactions) at the base of the structural properties of networks have been investigated during the last decade. Although some studies have found that the abundance of species is an important determining factor of species interactions (Vázquez et al. 2007), food webs and mutualistic networks also exhibit interaction distributions that depend on the evolutionary history of species (Rezende et al. 2007; Eklöf et al. 2012; Laigle et al. 2018), suggesting that long-term coevolution processes among species traits is at the base of patterns observed. In pollination networks, the first studies have mainly focused on the importance of morphological traits in shaping the structure of pollination networks, maybe because of an historical heritage, since Charles Darwin predicted the existence of a unknown pollinator from floral morphology (Darwin 1862). These works on morphological traits have shown they participate to structure pollination networks more than species abundance and promote nestedness (Santamaría & Rodríguez-Gironés 2007; Chamberlain et al. 2014; Watts et al. 2016; Biddick & Burns 2018). These studies also highlight that forbidden links, defined as impossible interactions among species due to incompatible traits, play an important role in pollination network structure (Santamaría & Rodríguez-Gironés 2007; Vizentin-Bugoni et al. 2014).
However, those studies used empirical data, aggregating observed interactions in time and space, while recent studies have shown that interaction networks are dynamic over space and time. Indeed, the few studies with seasonal sampling in pollination networks show that the seasonal turnover of interactions is high (CaraDonna et al. 2017; Souza et al. 2018; Rabeling et al. 2019), leading to seasonal changes in network structure (Rabeling et al. 2019; CaraDonna & Waser 2020), and suggesting that phenology is an important factor in community structure. Indeed, several studies including phenology in addition to morphological traits have shown that phenology is often the best predictor of plant-pollinator interactions (Junker et al. 2013; Maruyama et al. 2014; Gonzalez & Loiselle 2016; CaraDonna et al. 2017). Despite the importance of phenology for pollination interactions, the way that species are organized along the season, henceforth the seasonal structure of ecological communities, remains widely overlooked, theoretically as well as empirically. There are only few preliminary works on pollination networks linking seasonal structure to community diversity, network structure and robustness to extinctions (Encinas-Viso et al. 2012; Ramos–Jiliberto et al. 2018).
Thus, we currently know that ecological communities can be diverse and stable because of structural patterns that are non-random and stabilizing. However, linkage rules underlying these patterns remain unclear and taking into account the seasonal dynamic of communities seems an important step to highlight mechanisms structuring interactions networks, and thereby maintaining diversity.
Global change effects: from species to communities
Persistence, physiology, geographic range and phenology of species are changing because of global change. Communities are susceptible to be affected by these species’ responses to global change due to changes in species composition (nodes of the networks), because of 17 species extinctions, species invasions or variation in species abundances. In addition, communities can be affected by these species’ responses to global change through changes in interactions among species (links of the networks), because of a modification of interaction strengths, a break of existing interactions or emergence of new interactions. However, few empirical studies have focused on this question, our knowledge on the topic remaining mainly theoretical. Here, I will briefly review studies integrating species responses to global change in a community framework, focusing on geographic range and phenological shifts and not considering physiological responses that imply very different approaches.
Spatial and temporal mismatches between interacting species
Since species are moving in space and time and that those responses can vary in direction and strength (Parmesan 2007; Lenoir et al. 2020), interacting species might no longer co-occur in space and time, leading to spatial and temporal mismatch respectively. Evidence for temporal mismatch among trophic levels have been accumulated, trophic levels shifting their phenology with different strengths, higher trophic levels advancing less their phenology than lower ones (Thackeray et al. 2016). The most documented example is the increasing temporal mismatch between the peak of caterpillar abundance and the peak of nestling demand of birds, a recent review showing that, although birds respond adaptively in time, they fail to follow the advance of the caterpillar peak (Radchuk et al. 2019). Such temporal mismatch among interacting species is also occurring in marine ecosystems (Edwards & Richardson 2004), between insect egg hatching and plant bud burst (Visser & Both 2005; Asch et al. 2013), between herbivore mammals and plant growth (Post & Forchhammer 2008), flowering periods and pollinator activity periods (Burkle et al. 2013; Schmidt et al. 2016), etc.. Such mismatches even happen within species, as for example the phenological mismatch occurring between males and females of squirrels in their hibernation phenology, leading to a delay of the reproduction (Williams et al. 2017). Although there is no evidence that geographic range shifts differ among trophic levels, interspecific variations can also induce spatial mismatch, similarly to phenological shifts. However, most of the literature considers future projections of spatial mismatch, which I do not consider here, and very few empirical studies based on historical data show that those spatial mismatches among interacting species are currently occurring (Grunsven et al. 2007; Zang et al. 2020). Understanding the response to climate change of communities from past to present is however the first step to predict the future.
On the need to consider the heterogeneity of species responses
Many studies focus on average mismatch among functional or taxonomical groups of species, neglecting the substantial variation of responses among species within these groups. Such variation can lead to important interaction mismatches, even when the functional groups that interact are on average responding with the same direction and strength to global change. For instance, a species interacting with several species not shifting in space/time with the same direction or strength will be unable to follow all its partners (Memmott et al. 2007). This overlooked variance is also key to understand how species responses combine to affect the seasonal structure of communities. Few empirical studies have indeed shown that the seasonal structure of given functional groups was strongly modified because of heterogeneous phenological shifts (Diez et al. 2012; CaraDonna et al. 2014; Theobald et al. 2017; Carter et al. 2018; Hällfors et al. 2020). Not only phenological shifts can vary among species, thus affecting the structure of communities, but species phenological shifts can also vary over space, leading to the fact that similar communities at different locations can exhibit very distinct changes in their seasonal structure (Figure I-5). Changes in the seasonal structure of ecological communities affect the overall structure of the networks, which is likely to affect their stability and robustness to perturbations, stressing the need to investigate how changes in the seasonal structure of a community, a multi-dimensional object, affect community diversity, stability and functioning.
Table of contents :
Ecology in face of the global change
Species responses to global change
Shifts in geographicl range and phenology
Species persistence in a changing world
Species responses to global change in a community framework
Community stability and structure
Global change effects: from species to communities
From communities to species, a neglected eco-evolutionary feedback
Knowledge gaps and thesis questions
Characterizing species responses to global change drivers
Integrating species responses in a community framework
II. Methods & Results
Assessing species responses to global change
Modelling pollination networks with a seasonal structure
Changes in the seasonal structure of pollinator assemblages
Geographic range shifts are linked to species persistence
Eco-evolutionary mechanisms of phenological shifts
Consequences of species phenologies in a community framework
Temporal dynamics of biodiversity
Long-term temporal dynamics, over years
Short-term temporal dynamics, over the season
On the importance of competition in mutualistic networks
Competition and the seasonal structure
A trace of competition in species response to global change?
Opportunistic data vs protocoled data
VI. List of figures
VII. List of tables