Perspectives
Free Access

Beyond species loss: the extinction of ecological interactions in a changing world

Alfonso Valiente‐Banuet

Corresponding Author

Departamento de Ecología de la Biodiversidad, Instituto de Ecología, Universidad Nacional Autónoma de México, México, DF, México

Correspondence author. E‐mail: E-mail address:avalient@ecologia.unam.mx
Search for more papers by this author
Marcelo A. Aizen

Laboratorio Ecotono‐CRUB, Universidad Nacional del Comahue and INIBIOMA, San Carlos de Bariloche, Río Negro, Argentina

Search for more papers by this author
Julio M. Alcántara

Departamento de Biología Animal, Biología Vegetal y Ecología, Universidad de Jaén, Jaén, Spain

Search for more papers by this author
Juan Arroyo

Departamento de Biología Vegetal y Ecología, Universidad de Sevilla, Sevilla, Spain

Search for more papers by this author
Andrea Cocucci

Laboratorio de Biología Floral, Instituto Multidisciplinario de Biología Vegetal (IMBIV), CONICET‐Universidad Nacional de Córdoba, Córdoba, Argentina

Search for more papers by this author
Mauro Galetti

Departamento de Ecologia, Universidade Estadual Paulista (UNESP), Rio Claro, Sao Paulo, Brazil

Search for more papers by this author
María B. García

Instituto Pirenaico de Ecología, CSIC, Zaragoza, Spain

Search for more papers by this author
Daniel García

Departamento de Biología de Organismos y Sistemas, Universidad de Oviedo, Unidad Mixta de Investigación en Biodiversidad (UMIB CSIC‐UO‐PA), Oviedo, Spain

Search for more papers by this author
José M. Gómez

Departamento de Ecología Funcional y Evolutiva, Estación Experimental de Zonas Aridas (EEZA‐CSIC) Ctra, La Cañada de San Urbano, Almería, Spain

Departamento de Ecología, Facultad de Ciencias, Universidad de Granada, Granada, Spain

Search for more papers by this author
Pedro Jordano

Integrative Ecology Group, Estación Biológica de Doñana (EBD‐CSIC), Sevilla, Spain

Search for more papers by this author
Rodrigo Medel

Departamento de Ciencias Ecológicas, Facultad de Ciencias, Universidad de Chile, Santiago, Chile

Search for more papers by this author
Luis Navarro

Department of Plant Biology and Soil Sciences, Faculty of Biology, University of Vigo, Vigo, Spain

Search for more papers by this author
José R. Obeso

Departamento de Biología de Organismos y Sistemas, Universidad de Oviedo, Unidad Mixta de Investigación en Biodiversidad (UMIB CSIC‐UO‐PA), Oviedo, Spain

Search for more papers by this author
Ramona Oviedo

Instituto de Ecología y Sistemática de La Habana, Carretera Varona 11835 e/ Oriente y LinderoLa Habana 19, Calabazar, Boyeros, Cuba

Search for more papers by this author
Nelson Ramírez

Facultad de Ciencias, Instituto de Biología Experimental, Universidad Central de Venezuela, Caracas, Venezuela

Search for more papers by this author
Pedro J. Rey

Departamento de Biología Animal, Biología Vegetal y Ecología, Universidad de Jaén, Jaén, Spain

Search for more papers by this author
Anna Traveset

Terrestrial Ecology Group, Laboratorio Internacional de Cambio Global (LINC‐Global), Institut Mediterrani d'Estudis Avançats (CSIC‐UIB), Esporles, Mallorca, Balearic Islands, Spain

Search for more papers by this author
Miguel Verdú

Centro de Investigaciones sobre Desertificación (CIDE CSIC‐UV‐GV), Moncada, Valencia, Spain

Search for more papers by this author
Regino Zamora

Departamento de Ecología, Facultad de Ciencias, Universidad de Granada, Granada, Spain

Search for more papers by this author
First published: 28 November 2014
Cited by: 115

Summary

  • The effects of the present biodiversity crisis have been largely focused on the loss of species. However, a missed component of biodiversity loss that often accompanies or even precedes species disappearance is the extinction of ecological interactions.
  • Here, we propose a novel model that (i) relates the diversity of both species and interactions along a gradient of environmental deterioration and (ii) explores how the rate of loss of ecological functions, and consequently of ecosystem services, can be accelerated or restrained depending on how the rate of species loss covaries with the rate of interactions loss.
  • We find that the loss of species and interactions are decoupled, such that ecological interactions are often lost at a higher rate. This implies that the loss of ecological interactions may occur well before species disappearance, affecting species functionality and ecosystems services at a faster rate than species extinctions. We provide a number of empirical case studies illustrating these points.
  • Our approach emphasizes the importance of focusing on species interactions as the major biodiversity component from which the ‘health’ of ecosystems depends.

Introduction

Understanding the factors behind biodiversity loss is one of the most urgent tasks faced by scientists at present. Fast‐paced rates of habitat loss and fragmentation, and large‐scale disturbances (e.g. biological invasions, habitat degradation, recurrent defaunation) are presently the main anthropogenic drivers of species extinctions. Current estimates of species extinctions are three to four orders of magnitude above background extinction rates (Barnosky et al. 2011). Recent estimates of extinction risk from a wide range of climate impacts and ecosystems types indicate that real empirical data often exceed predicted extinctions (Maclean & Wilson 2011), which suggests that we are still far from having adequate tools to predict species losses attributable to major anthropogenic disturbances such as habitat loss (He & Hubbell 2011). Furthermore, many species now survive at such low densities that they can be considered practically extinct from an ecological point of view (Janzen 2001). These limitations preclude an adequate assessment of ecosystem functions decays, and their derived services, an emerging research field where species richness is the diversity metric (Balvanera et al. 2006; Cardinale et al. 2012).

Biodiversity loss has been largely assessed through species extinctions. However, an important yet frequently missed component of biodiversity loss that goes along or may even precede species extinctions is the extinction of ecological interactions in which those species are engaged (Janzen 1974; Tylianakis et al. 2008; Aizen, Sabatino & Tylianakis 2012). Given that many key functional aspects of ecosystems closely depend on biotic interactions, their loss may have pervasive effects accelerating species local extinction and decay of ecosystem functions, ultimately collapsing the derived services provided to humans (Díaz et al. 2013). Moreover, extinctions of interactions and their associated ecological functions may frequently precede the complete disappearance of species, that is, when a species’ abundance is so reduced as to represent its functional extinction (Redford 1992; Janzen 2001; Wilkie et al. 2011; Galetti et al. 2013; Säterberg, Stellman & Ebenman 2013). Biotic interactions are thus a major but often neglected component of biodiversity that needs to be considered in order to assess the ‘health’ of ecosystems and to define critical indicators providing early diagnosis of environmental problems (Tylianakis et al. 2010; Aizen, Sabatino & Tylianakis 2012; Dirzo et al. 2014). Several studies have already demonstrated that species interactions are particularly sensitive to different anthropogenic drivers often shifting their frequency or becoming disrupted due to large environmental changes (Tylianakis et al. 2008). In consequence, there is an urgent need to develop robust analytical tools, such as generic mechanistic models, to understand, foresee and manage the impact of anthropogenic drivers on biotic interactions.

Under this conceptual framework, and based on a set of proposed definitions, here, we model the relationship between both the diversity of species and interactions along a gradient of habitat loss (as a generic form of environmental degradation) and examine the consequences of interaction extinctions for the maintenance of ecological functions. We assume that species extinctions increase with increasing habitat loss, the signature conservation problem of the twenty‐first century (He & Hubbell 2011). Other scenarios where species gains occur after habitat loss (e.g. invaded sites having higher levels of diversity than pristine sites; Lekberg et al. 2013; Stouffer, Cirtwill & Bascompte 2014) are not considered here, but extensions of the proposed model could account for these situations. In our model, we apply the concept of extinction debt of species to ecological interactions [see original definitions in Tilman et al. (1994) and Wearn, Reuman & Ewers (2012)]. Accordingly, we define the extinction debt of ecological interactions as any future interaction loss that has to be realized due to a current or past environmental disturbance. Our concept refers to those interactions ‘committed to extinction’ owing to lags between the habitat loss and the complete extinction of the interaction. We refer to extinction of an interaction when its ecological outcomes are no longer functional, for example collapse of fruit set and/or fruit removal, loss of long‐distance dispersal. For example, extensive declines of megafauna frugivores due to habitat fragmentation might represent future functional losses of seed dispersal services (Bueno et al. 2013), even when the species themselves and their interactions remain sporadically recorded in biodiversity inventories. Here, we show that interaction‐based extinction debts emerge as a consequence of the different pace for species and interactions extinctions. The mismatch between species and interaction extinction curves determines the response of ecological functions, depending on how crucial is their functional effect in local communities.

Shifting the focus from species to ecological interactions

The extinction of species inevitably causes the loss of interactions in which they are involved. In this situation, the number of interactions disappearing due to habitat loss can be assumed to be proportional to the number of species going extinct, a situation we call directly proportional interaction loss. However, this assumption may not be necessarily well warranted as it implies that (i) species are equivalent in terms of the number and quality of their interactions, (ii) species cannot survive, at least temporarily, despite becoming mostly disconnected from others and (iii) the emergence of new interactions does not occur. In this oversimplified scenario, which can serve as a null hypothesis, the set of interactions is viewed as static and homogeneously structured. To explore the broad array of situations where the number of interactions lost does not follow linearly the number of species disappearing from the community, we have developed a model that takes into account the way interactions are lost as a function of habitat loss (Box 1). Our model allows examining the consequences of an increased or decreased loss of interactions relative to the number expected under directly proportional interaction loss, that is interaction balance (Fig. 1c). The interaction balance, in turn, depends on a single parameter, which we call delta (∆L), a property of the community that reflects its resistance or susceptibility to interaction losses regarding species loss. In other words, delta indicates how much faster or slower the speed of interaction loss is with respect to a proportional loss. When delta is negative, interactions are lost at a lower rate than species along the gradient of habitat loss, and thus, the interaction balance is positive (Fig. 1c). We call this situation interaction surplus. In contrast, when delta is positive, interactions are lost at a higher rate than species along the gradient of habitat loss, and the resulting interaction balance is negative (Fig. 1c). We call this situation interaction deficit.

Box 1. Predicting interaction losses in a changing world

We define an axis of habitat loss (x) that ranges from a pristine, undisturbed stage (zero) to a stage (one) of complete disturbance where all species present in the zero stage have lost their habitat. This axis may represent a reversed area axis, widely used to indirectly predict extinction rates with species–area curves (He & Hubbell 2011). It may also represent any axis of habitat deterioration, such as increased isolation of habitat fragments, degree of desertification and intensity of grazing. Along this axis, we represent the pattern of species loss as a function of species richness decrease (He & Hubbell 2011), S(x), with α as the slope of this loss (Fig. 1a). To study how the pattern of species extinctions may affect the interaction extinctions, we translated species richness into an interaction or link loss function, L(x), through the expression derived by (May 1972) and empirically corroborated in recent studies (Ings et al. 2009). It then follows that L(x) = aS(x)b. Where ‘a’ is a proportionality factor which represents the per species number of links in the undisturbed stage,that is when habitat loss equals zero. According to the value of ‘b’, link loss from one stage to the next is faster or slower than species loss. Only when b = 1 is interaction loss proportional to species loss, thus, the amount of interaction loss proportionally contributed by species loss (directly proportional loss) may be dissected out from total interaction loss, that is for an L(x) function with b = 1 + δ (Fig. 1b). Accordingly, we derive interaction balance (ΔL(x)),that is the number of interactions in a community with a given degree of habitat loss in excess or deficit with respect to the expected number of interactions under exclusively directly proportional interaction loss. Interaction balance is given from the above expressions by ΔL(x) = a[S(x)1 + δ−S(x)] which is represented as a unimodal curve (Fig. 1c). Interaction balance may represent an interaction surplus (ΔL(x) > 0) when the community is resilient to interaction losses, whereas it may represent an interaction deficit (ΔL(x) < 0) when the community is prone to cascading interaction losses. The stage of habitat loss where deficit or surplus is greatest shows critical points in the process of habitat degradation. Species loss or retention that is due to the effect of interaction balance we call species balance and is calculated as ΔS = ΔL(x)/k(x), where k(x) is the grade (mean number of interactions per species) at a given stage of habitat loss. This quantity tells about the number of species lost or retained as a consequence of accelerated or decelerated link loss.

(a) Species richness functions along an axis of increasing habitat loss (x) for three slope values, α = 0·1 (dotted), α = 0·2 (dashed) and α = 0·5 (solid). (b) Interaction richness functions corresponding to a species richness function with α = 0·5, number of species (S0) and mean number of links (k0) at = 0 equal 100 and 5, respectively. For each, functions with values of b decreasing and increasing by δ > 0 and δ < 0 are represented above (dashed) and below (dotted) the directly proportional function (b = 1, solid line), respectively. (c) Interaction balance functions, ΔL(x), corresponding to the interaction richness functions in b. Vertical lines show the critical values of x where interaction deficit (below) or superavit (above) are greatest.

Interaction surplus occurs when the average number of interactions per species (degree) of a community that has been subjected to habitat loss is greater than before habitat loss. This phenomenon may result from two non‐mutually exclusive mechanisms. First, if the surviving species build up new interactions (i.e. ‘rewiring’ in network terminology), then the directly proportional interaction loss may be overcompensated, increasing the linkage density (Pimm, Lawton & Cohen 1991). For example, many plants are visited by few and efficient pollinators that competitively exclude less effective pollinators (Mitchell et al. 2009). In these systems, the loss of a main pollinator entails that less efficient and otherwise competitively excluded animals broaden their list of floral resources by pollinating new plant species or populations (Pérez‐Barrales, Vargas & Arroyo 2006; Pérez‐Barrales & Arroyo 2010). Likewise, if a magnet plant species is removed from a community, pollinators may compensate its suppression by building up new interactions with other plant species to which they were not attracted before (Kaiser‐Bunbury et al. 2010). Secondly, when species extinction is biased towards the less connected species and the connections among the surviving species remain unaltered after disturbance, we might expect a surplus of interactions relative to the situation before disturbance. For instance, pollination and seed dispersal webs can remain with surplus connectivity in isolated and fragmented areas, as species number decreases and generalists survive better than specialists (Hagen et al. 2012). Similarly, the mean number of plant–plant facilitation interactions may increase in a community after the local extinction of the less connected nurse plants because facilitated species are able to recruit under generalist nurse species (Verdú & Valiente‐Banuet 2008; Valiente‐Banuet & Verdú 2013). Finally, interaction surplus may occur when the loss of interactions caused by the extinction of a key species promotes the emergence of new interactions of a different type. In many communities, competition between plant species is cancelled by the presence of a generalist keystone herbivore that keeps plant abundance below a critical threshold. Removal of the keystone herbivore may cause pairwise competitive interactions among all or most plant species to become manifest (Olff & Ritchie 1998). In this new situation, the number of competitive interactions in the herbivore‐less community increases with respect to the original community.

In a second, non‐mutually exclusive mechanism, interaction deficit may occur when habitat loss causes a reduction in the abundance of species down to a threshold below which the interactions are no longer maintained. For instance, many plants can escape predation and herbivory from generalist animals by living at very low densities (Greenwood 1985; Rooney 1997; Underwood & Halpern 2012). In such circumstances, the low densities of partners, making the actual encounter unlikely, may cancel potential antagonistic interactions. Likewise, interaction deficits may also occur in systems where many weak interactions are maintained by a core of strong mutualists (super‐generalists) that enhance their abundance up to a threshold above which the weak interactions spontaneously arise (e.g. pollination and seed dispersal market). In this scenario, with many weak indirect interactions, the loss of one of such interactions will surely have only minor overall consequences. In contrast, the weakening of strong positive interactions between core species may trigger a decrease in the species’ abundances, causing the loss of co‐occurring weak interactions (Aizen, Sabatino & Tylianakis 2012). These interaction deficits are likely attributable to subtle variation in the abundance of generalized plant–animal mutualisms, such as some pollination and dispersal interactions, as they imply consumer–resource use interactions with marked frequency dependency (Holland & DeAngelis 2010).

Other systems where multiple interactions arise by the effect of a single species are likely to produce interaction deficits. For example, in communities shaped by facilitation, a benefactor (e.g. nurse plant species) facilitates the establishment of other species, causing spatially aggregated clumps of species where multiple interactions are established (Castillo, Verdú & Valiente‐Banuet 2010). Human overexploitation, by reducing the abundance of a number of key nurses, has triggered the loss of a disproportionately high number of interactions between nurses and facilitated species (Valiente‐Banuet & Verdú 2013). Another source of interaction deficit occurs through phenological mismatches between interacting species when environmental changes disrupt the synchronization in the activity periods of interaction partners (Arroyo & Dafni 1995; Visser & Holleman 2001). The speed of this detrimental effect may be strongly variable relying on the indirect dependence of species to interactions. For instance, an increased fruit set in small remnant forests has been detected when spare populations of native bee pollinators are displaced by a high abundance of introduced honeybees, reducing the number of mutualistic interactions (González‐Varo, Arroyo & Aparicio 2009). Increased geitonogamy by honeybees, however, conveyed higher selfing rates and reduced progeny performance in such forests (González‐Varo et al. 2010). In some cases, anthropogenic disturbances leading to habitat fragmentation affecting pollinator and seed disperser communities have been shown to promote evolutionary change in populations rather than extinction (Murúa et al. 2010; Brys & Jacquemyn 2012; Galetti et al. 2013), although the generality and trajectory of the change is still poorly understood (Jacquemyn et al. 2012). In general, interaction deficits can be envisioned under many different scenarios of anthropogenic habitat loss where interaction loss precedes species extinction (Sabatino, Maceira & Aizen 2010). This is probably a very frequent situation, where interactions – more labile in their origination and maintenance – are more prone to failure than species under the continuous deterioration of the environment (Quental & Marshall 2013).

From interaction extinctions to functional loss

Interspecific interactions provide the raw material for relevant ecological functions that determine the dynamics of populations, communities and ecosystems. For instance, pollination, seed dispersal and plant–plant facilitation and competition may drive the trajectory and speed of ecological succession (Connell & Slatyer 1977; Verdú et al. 2009) and the co‐evolutionary dynamics of multispecies assemblages (Guimarães, Jordano & Thompson 2011). Moreover, the functional outcomes of biotic interactions, mainly pollination, seed dispersal or host–parasitoid relationships, have been recently signalled as relevant ecosystem services (Kremen et al. 2007). The effects of interaction loss on the delivery of ecological functions will depend on the relative contribution of each interaction type to the global function provided by the whole interaction assemblage. From this perspective, we identify here three types of interaction functionalities:

  1. Additive interactions: An interaction is called additive when its loss impacts at least one species involved in the interaction, but has no indirect effects on the remaining links of the interaction network. This category is based on the functional singularity of interactions within communities. In other words, every interaction adds its singular contribution to the ecological function and contributes equally to the function maintenance. This is the case of extremely specialized interactions. The fig–fig wasp interaction is a good example of additive interactions (Weiblen 2002), as the pollination service of the fig‐dominated community depends on the addition of singular contributions of every species‐specific fig–fig wasp interaction. When interactions are additive, the functional loss is expected to be gradual, decreasing linearly relative to the interaction loss (See Box 2, Fig. 2a). Thus, we expect that an increasing intensity of the anthropogenic driver will trigger a proportional response of interaction loss and functional loss (Fig. 2b).
  2. Keystone interactions: An interaction is called keystone when it controls community or ecosystem function to a disproportionately higher extent than predicted from the abundance of interacting organisms (Helfield & Naiman 2006). This interaction type is expected in communities showing strong differences in the contribution of interspecific interactions to ecological function. Such is the case when generalized mutualists (often called super‐generalists) are removed from plant–pollinator or plant–disperser networks. For example, scatter‐hoarding rodents play a pivotal role in the dispersal of large‐seeded species in neotropical rain forests, especially in those once dispersed by the extinct Pleistocene megafauna (Guimarães, Galetti & Jordano 2008; Jansen et al. 2012). In Amazonian forests, the overharvesting of the Brazilian nut (Bertholetia excelsa) may have important consequences not only for the recruitment of B. excelsa and agouti populations, but also for the entire scatter‐hoarding dispersal service of other large‐seeded tree species (Peres et al. 2003; Galetti et al. 2006). Similar examples have been reported for flying foxes, gorillas and elephants for which overhunting and habitat loss has resulted in a recruitment reduction of a significant number of large‐seeded species (McConkey & Drake 2006; Beaune et al. 2013; Haurez, Petre & Doucet 2013). Moreover, predator–prey interactions can have effects at different levels affecting preys directly and plants indirectly in many different ways (Novaro, Funes & Walker 2000; Ripple & Beschta 2006; Anderson et al. 2011). Strong reduction of grazing by a few large domestic herbivores in the alpine habitats of European mountains during the last decades is causing the encroachment of man‐made meadows and the shift upwards of the tree line (Améztegui, Brotons & Coll 2010), which inevitably implies the loss of a countless number interactions associated to the very rich alpine meadows. The loss of a keystone interaction has a disproportionate impact on the associated ecological function (Fig. 2a). Thus, when a keystone interaction is disrupted along an increasing gradient of anthropogenic impact, the involved ecological function will decline at a faster rate than the loss of interactions (Fig. 2b).
  3. Redundant interactions: Ecological interactions are redundant when their contributions to the community functioning are equivalent and replaceable, as most interactions are able to sustain the optimal function by themselves (Lawton & Brown 1993). Thus, all pollinators are interchangeable as they have the same effect on fruit set (Gómez & Zamora 1999; Fründ et al. 2013). This seems to be the case of several Sonoran desert cacti, such as cardon (Stenocereus thurberi) and saguaro (Carnegiea gigantea), in which pollination either by bats (cardon) or by birds and honeybees (saguaro) alone accounts for most reproductive success (Fleming et al. 2001). Similarly, this occurs for the pollination of the endemic alpine Borderea pyrenaica, which is successfully pollinated by the scarce resident ants, the years when the most abundant lady beetles do not arrive on time (García, Antor & Espadaler 1995). When an interaction is redundant, its loss will not reverberate in the ecological function of the interaction network, as long as other interactions substitute the role left by the lost one. Thus, ecological functions will be largely unaffected when the intensity of the anthropogenic driver is increased (Fig. 2b). Only when most interactions are lost, at high levels of the anthropogenic impact, will the ecological function finally collapse (Fig. 2b). The lag between environmental change and functional decay represents, therefore, an extinction debt for ecological functions (Fig. 2b).

Box 2. Interaction loss and functional decay

Once we have derived the functions of link loss along an axis of habitat loss (Box 1), we may translate link loss into functional loss given the theoretical relationships between functionality and link richness. These relationships can be described by means of a power law function of accumulative functionality with increasing link richness, λ. This function is given by F(λ)∝λγ where γ modulates the interaction functionality in the network such that when γ = 1, γ > 1 or 0 < γ < 1 the functional role is additive, keystone or redundant, respectively (Fig. 2a). By replacing in this function λ with L(x) from Box 1, we obtain a relationship of functional loss along an axis of habitat loss as F(x)∝L(x)γ or F(x)∝S(x). In other words, the speed of functional loss along the axis of habitat loss equals that of link loss only when the role is additive (γ = 1), whereas functional loss may be exacerbated if its role is keystone (γ > 1) or delayed if it is redundant (γ < 1; Fig. 2b). In analogy to interactions balance, we may also define functional balance to denote situations when functionality is maintained above or below values expected by link loss, buffering or exacerbating functional collapses or functional debts.

(a) Ecological function F(λ) for interaction networks of variable size (λ), depending on three functional interaction roles (from top: additive, redundant and keystone). (b) Decay of ecological function (in magenta) along an axis of increasing habitat degradation, (x) depending on the additive, redundant and keystone role of interactions. The response of link richness to habitat loss L(x) in a scenario of null interaction balance is shown, highlighting a functional collapse (for keystone role, the functional loss is faster than expected from interaction loss) or functional debt (for redundant role, the functional loss is slower than expected from interaction loss). (c) Functional decay for additive, redundant and keystone roles of interactions under different interaction balances (null in magenta, negative in black, positive in blue).

Once we have established that the response of ecological function to anthropogenic change depends on the functional role of interactions, we may predict the functional loss expected under different scenarios of balance between interactions loss and species loss, that is, under situations of interaction surplus, deficit or null balance (Fig. 2c). Accordingly, positive and negative interaction balances will, respectively, exacerbate functional debts and functional collapses along the gradients of anthropogenic change.

The expected response of ecological functions along the habitat loss gradient would thus depend on the type of interaction scenario: surpluses in keystone scenarios vs. interaction deficits in redundant scenarios. This alone may buffer both functional losses and functional debts along the gradient, due to the multiplicative effect of interaction functionality and interaction balance.

For example, the redundancy among frugivores of Prunus avium in central Germany seems to maintain seed dispersal function across human‐impacted landscapes despite changes in frugivore richness (Breitbach et al. 2012). On the other hand, an even stronger functional collapse is expected when the loss of keystone interactions is associated to concomitant interaction losses (i.e. interaction deficit). For example, the loss of the interaction of the palm Maximiliana maripa and its main seed disperser, the tapir Tapirus terrestris, in the Brazilian Amazonia leads to an additional loss of functionality in the secondary seed dispersal accounted by rodents feeding on seeds from tapir latrines (Fragoso 1997). On the other hand, a surplus of redundant interactions may lead to a longer functional debt. This seems to be the case in the interaction network between fleshy‐fruited trees and frugivorous thrushes (Turdus spp.) in northern Spain. There, the decay of frugivory on the dominant tree Ilex aquifolium in low‐crop years is compensated with stronger frugivory on the tree Crataegus monogyna, whose widespread distribution leads to an increased resilience of seed dispersal to habitat loss (García et al. 2013). A fast‐paced loss of interactions ahead of species extinctions (i.e. an empty‐forest syndrome; Janzen 1974; Redford 1992) will be clearly pervasive whenever extinction hits the keystone interactions.

Concluding remarks

In a changing world, habitat loss, climate change, severe defaunation and invasion of exotic species are the main drivers of the present biodiversity crisis; the effects of which have been largely assessed by quantifying the loss of species. Despite considerable efforts of conservationists to secure viable populations of threatened species, their ecological extinction may have already occurred (Janzen 1974; Traveset & Riera 2005; McConkey & Drake 2006; Aslan et al. 2013; Galetti et al. 2013; Säterberg, Stellman & Ebenman 2013). Here, we applied a mechanistic model to demonstrate that the extinction of interactions can be decoupled from the extinction of species, fine‐tuning the concept of co‐extinction cascades of interacting species as a major form of biodiversity loss (Koh et al. 2004; Rezende et al. 2007; Dunn et al. 2009). More specifically, we suggest that such decoupling can be negative or positive, which means that interactions can be lost faster or go along the loss of species. In consequence, the rate of species extinction may be delayed or accelerated depending on the balance of interaction losses.

Our model shows that a critical aspect of how the loss of interactions impacts biodiversity depends on the balance between the rate at which interactions are lost relative to the loss of species. Losing interactions mean losing ecological functions, and the magnitude of the extinction debt due to the extinction of interactions can exceed the proportional level depending on the mismatch between species extinctions and interaction extinctions. Situations of interaction surplus occur whenever interactions are lost at a lower rate than species along the disturbance gradient. This is expected, for example, when species ‘rewire’ their ecological links in absence of missing partners (Pires et al. 2014). Situations of interaction deficit are also frequent because they relate to the loss of species functionality that precedes species extinction. Reduced abundances of supergeneralist species due to logging (trees) or hunting (megafauna vertebrates) may cause ecological extinction and trigger the loss of multiple interactions, resulting in interaction deficits and a sizeable interactions–extinction debt.

Therefore, although most previous studies on the biodiversity–ecosystem functioning paradigm have focused exclusively on species richness, our novel analytical approach reveals the relevance of considering the species interaction balance as a critical indicator of ecosystem health (see Table S1, supporting information for a summary of examples). To ensure the long‐term provision of ecosystem services that depend upon biodiversity, the greatest attention should be focused on those components of biodiversity, such as species interactions, that can be affected by the new scenarios emerging in a changing world. Anthropogenic impacts are the most evident causes of biodiversity loss, but other causes can also be at work. The consequence of species interactions loss for biodiversity is just an emerging field. Box 3 outlines future avenues and issues for research on the consequences of species interaction loss for biodiversity reductions. Thus, the focus on the health of biotic interactions provides a deeper, process‐oriented understanding of the functional consequences of biodiversity in a changing world. Using interaction losses as early warning signals of critical transitions is necessary to identify and quantify extinction debt and would allow us detecting early on‐time thresholds of habitat loss from which collapse of ecosystem functions is imminent.

Box 3. Key issues for future research

Drivers for change in the interaction balance

  1. What are the main drivers of interaction debts?

  2. How do changes in species diversity influence changes in the diversity of interactions and their balance?

Interaction network structure

  1. How does the loss of interactions impact the structure of interaction networks?
  2. Are different types of interaction loss (surplus vs. deficit) associated with different types of interactions (e.g. mutualistic vs. antagonistic)?

Ecological services

  1. What is the impact of losing different types of interactions (i. e. additive, keystone, redundant) upon ecological services?
  2. To what extent ecological services depend on interaction balance?

Acknowledgements

Four workshops that led to this publication were organized with financial support provided by Programa Iberoamericano de Ciencia y Tecnología para el Desarrollo (CYTED Project 409AC0369). AV‐B thanks to Alejandro de Avila and co‐workers for facilities during the workshop held in the Botanical Garden of the Centro Cultural Santo Domingo, Oaxaca, Mexico. MBG thanks the funding received from the Spanish MEC (CGL2011‐13456‐E). PJ was supported by a Junta de Andalucía Excellence Grant (RNM‐5731), as well as a Severo Ochoa Excellence Award from the Ministerio de Economía y Competitividad (SEV‐2012‐0262). RM thanks the funding support from FONDECYT 1120155. MG received a fellowship from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Fundação de Amparo 'a Pesquisa do Estado de São Paulo (FAPESP). Marc Johnson, Angela Moles and one anonymous reviewer greatly improved the manuscript with their comments.

    Authorship

    RZ, MAA, MBG and AV‐B organized the workshops in Granada (Spain), Bariloche (Argentina), Jaca (Spain) and Oaxaca (Mexico), respectively. All the authors attended the workshops, outlined the manuscript and contributed substantially to its writing. AC, JMG and PJ prepared all figures.

    Number of times cited: 115

    • , Do asynchronies in extinction debt affect the structure of trophic networks? A case study of antagonistic butterfly larvae–plant networks, Oikos, 127, 6, (803-813), (2017).
    • , Zero‐sum landscape effects on acorn predation associated with shifts in granivore insect community in new holm oak (Quercus ilex) forests, Diversity and Distributions, 24, 4, (521-534), (2018).
    • , Persisting in defaunated landscapes: Reduced plant population connectivity after seed dispersal collapse, Journal of Ecology, 106, 3, (936-947), (2017).
    • , Interaction frequency, network position, and the temporal persistence of interactions in a plant–pollinator network, Ecology, 99, 1, (21-28), (2017).
    • , Landscape effects on pollination networks in Mediterranean gypsum islands, Plant Biology, 20, S1, (184-194), (2017).
    • , Reflections on, and visions for, the changing field of pollination ecology, Ecology Letters, 21, 8, (1282-1295), (2018).
    • , Alien plants have greater impact than habitat fragmentation on native insect flower visitation networks, Diversity and Distributions, 24, 1, (58-68), (2017).
    • , Responses of interspecific associations in mixed‐species bird flocks to selective logging, Journal of Applied Ecology, 55, 4, (1637-1646), (2018).
    • , Seed‐dispersal interactions in fragmented landscapes – a metanetwork approach, Ecology Letters, 21, 4, (484-493), (2018).
    • , Fishing‐down within populations harms seed dispersal mutualism, Biotropica, 50, 2, (319-325), (2017).
    • , Defaunation leads to interaction deficits, not interaction compensation, in an island seed dispersal network, Global Change Biology, 24, 1, (e190-e200), (2017).
    • , New approaches to enhance pollutant removal in artificially aerated wastewater treatment systems, Science of The Total Environment, 627, (1182), (2018).
    • , Species co-occurrence analysis predicts management outcomes for multiple threats, Nature Ecology & Evolution, 10.1038/s41559-017-0457-3, 2, 3, (465-474), (2018).
    • , Bird-mediated effect of deforestation on potential seed dispersal does not increase the seed recruitment of Fuchsia magellanica , Trees, 32, 1, (245), (2018).
    • , Ecological Networks in Changing Tropics, Ecological Networks in the Tropics, 10.1007/978-3-319-68228-0_11, (155-169), (2018).
    • , Bat diversity in Carajás National Forest (Eastern Amazon) and potential impacts on ecosystem services under climate change, Biological Conservation, 218, (200), (2018).
    • , Tropical Biodiversity: The Importance of Biotic Interactions for Its Origin, Maintenance, Function, and Conservation, Ecological Networks in the Tropics, 10.1007/978-3-319-68228-0_1, (1-13), (2018).
    • , Spatial distance and climate determine modularity in a cross‐biomes plant–hummingbird interaction network in Brazil, Journal of Biogeography, 45, 8, (1846-1858), (2018).
    • , Trait‐based approaches to analyze links between the drivers of change and ecosystem services: Synthesizing existing evidence and future challenges, Ecology and Evolution, 7, 3, (831-844), (2017).
    • , Landscape simplification weakens the association between terrestrial producer and consumer diversity in Europe, Global Change Biology, 23, 8, (3040-3051), (2017).
    • , Global Risks, Conservation, and Criminology, Conservation Criminology, (1-23), (2017).
    • , Unifying concepts of biological function from molecules to ecosystems, Oikos, 126, 10, (1367-1376), (2017).
    • , Organic enrichment simplifies marine benthic food web structure, Limnology and Oceanography, 62, 5, (2179-2188), (2017).
    • , Predicting the consequences of disperser extinction: richness matters the most when abundance is low, Functional Ecology, 31, 10, (1910-1920), (2017).
    • , Considering the unintentional consequences of pollinator gardens for urban native plants: is the road to extinction paved with good intentions?, New Phytologist, 215, 4, (1298-1305), (2017).
    • , Effect of habitat degradation on competition, carrying capacity, and species assemblage stability, Ecology and Evolution, 7, 15, (5784-5796), (2017).
    • , Trait‐matching and mass effect determine the functional response of herbivore communities to land‐use intensification, Functional Ecology, 31, 8, (1600-1611), (2017).
    • , Bridging ecology and conservation: from ecological networks to ecosystem function, Journal of Applied Ecology, 54, 2, (371-379), (2016).
    • , Increased resource availability prevents the disruption of key ecological interactions in disturbed habitats, Ecosphere, 8, 4, (2017).
    • , Credit of ecological interactions: A new conceptual framework to support conservation in a defaunated world, Ecology and Evolution, 7, 6, (1892-1897), (2017).
    • , The new nature: Limitations and prospects of the paleoenvironmental tradition in biogeography in the 21st century, The Canadian Geographer / Le Géographe canadien, 61, 1, (41-51), (2017).
    • , Defaunation and biomass collapse of mammals in the largest Atlantic forest remnant, Animal Conservation, 20, 3, (270-281), (2016).
    • , Fruit biomass availability along a forest cover gradient, Biotropica, 49, 1, (45-55), (2016).
    • , Seed dispersal by macaws shapes the landscape of an Amazonian ecosystem, Scientific Reports, 10.1038/s41598-017-07697-5, 7, 1, (2017).
    • , Forest cover influences occurrence of mammalian carnivores within Brazilian Atlantic Forest, Journal of Mammalogy, 98, 6, (1721), (2017).
    • , Microbial diversity and ecological networks as indicators of environmental quality, Environmental Chemistry Letters, 15, 2, (265), (2017).
    • , Differential pollination modes between distant populations of Unonopsis guatterioides (Annonaceae) in Minas Gerais and Amazonas, Brazil, Flora, 232, (39), (2017).
    • , Behavioral Diversity (Ethodiversity): A Neglected Level in the Study of Biodiversity, Frontiers in Ecology and Evolution, 5, (2017).
    • , Role of small rodents in the seed dispersal process: icrocavia australis consuming rosopis flexuosa fruits, Austral Ecology, 42, 1, (113-119), (2016).
    • , Spectators or participants: How can SETAC become more engaged in international climate change research programs?, Environmental Toxicology and Chemistry, 36, 8, (1971-1977), (2017).
    • , Ecological and genetic consequences of fragmentation in a semiarid Mediterranean urban forest, Urban Ecosystems, 20, 5, (1161), (2017).
    • , A structural equation modeling approach for formalizing and evaluating ecological integrity in terrestrial ecosystems, Ecological Informatics, 41, (74), (2017).
    • , Challenges and opportunities for rewilding South American landscapes, Perspectives in Ecology and Conservation, 10.1016/j.pecon.2017.10.002, 15, 4, (245-247), (2017).
    • , Ecosystem quality in LCIA: status quo, harmonization, and suggestions for the way forward, The International Journal of Life Cycle Assessment, (2017).
    • , Preliminary results on snow surveys of Pyrenean grey partridge (Perdix perdix hispaniensis) in the Cantabrian Mountains, European Journal of Wildlife Research, 63, 5, (2017).
    • , Habitat Losses and Conservation of Mutualisms, Mutualisms and Insect Conservation, 10.1007/978-3-319-58292-4_10, (193-223), (2017).
    • , Environmental warming accelerates extinctions but does not alter extinction debt, Basic and Applied Ecology, 24, (30), (2017).
    • , Deforestation drives functional diversity and fruit quality changes in a tropical tree assemblage, Perspectives in Plant Ecology, Evolution and Systematics, 28, (78), (2017).
    • , Pollinator importance networks illustrate the crucial value of bees in a highly speciose plant community, Scientific Reports, 7, 1, (2017).
    • , Can dung beetles (Scarabaeinae) indicate the status of Amazonia’s ecosystems? Insights integrating anthropogenic disturbance with seasonal patterns, Animal Biology, 67, 3-4, (301), (2017).
    • , Analysis and effect of the use of biofertilizers on Trifolium rubens L., a preferential attention species in Castile and Leon, Spain, with the aim of increasing the plants conservation status, AIMS Microbiology, 3, 4, (733), (2017).
    • , How do land-use legacies affect ecosystem services in United States cultural landscapes?, Landscape Ecology, 10.1007/s10980-017-0545-4, 32, 11, (2205-2218), (2017).
    • , Not equal in the face of habitat change: closely related fishes differ in their ability to use predation-related information in degraded coral, Proceedings of the Royal Society B: Biological Sciences, 10.1098/rspb.2016.2758, 284, 1852, (20162758), (2017).
    • , Keystone Species, Forest and Landscape: A Model to Select Protected Areas, Environmental Management, 59, 6, (1017), (2017).
    • , Phylogenetic Paleoecology: Tree-Thinking and Ecology in Deep Time, Trends in Ecology & Evolution, 32, 6, (452), (2017).
    • , Best Practices Toward Sustainable Ecotourism, Ecotourism’s Promise and Peril, 10.1007/978-3-319-58331-0_10, (153-178), (2017).
    • , References, Mixed-Species Groups of Animals, 10.1016/B978-0-12-805355-3.16001-4, (169-198), (2017).
    • , The friendship paradox in species-rich ecological networks: Implications for conservation and monitoring, Biological Conservation, 209, (245), (2017).
    • , Fire modulates the effects of introduced ungulates on plant–insect interactions in a Patagonian temperate forest, Biological Invasions, 19, 8, (2459), (2017).
    • , Synergistic effects of seed disperser and predator loss on recruitment success and long-term consequences for carbon stocks in tropical rainforests, Scientific Reports, 10.1038/s41598-017-08222-4, 7, 1, (2017).
    • , Fine-scale determinants of conservation value of river reaches in a hotspot of native and non-native species diversity, Science of The Total Environment, 10.1016/j.scitotenv.2016.09.097, 574, (455-466), (2017).
    • , Rewilding the Atlantic Forest: Restoring the fauna and ecological interactions of a protected area, Perspectives in Ecology and Conservation, 10.1016/j.pecon.2017.09.004, 15, 4, (308-314), (2017).
    • , Are disturbance gradients in neotropical ecosystems detected using rove beetles? A case study in the Brazilian Amazon, Forest Ecology and Management, 405, (319), (2017).
    • , Frugivore diversity increases frugivory rates along a large elevational gradient, Oikos, 125, 2, (245-253), (2015).
    • , Leveraging nature's backup plans to incorporate interspecific interactions and resilience into restoration, Restoration Ecology, 24, 4, (434-440), (2016).
    • , Patch size, shape and edge distance influence seed predation on a palm species in the Atlantic forest, Ecography, 39, 5, (465-475), (2015).
    • , Complexity of leaf miner–parasitoid food webs declines with canopy height in Patagonian beech forests, Ecological Entomology, 41, 5, (599-610), (2016).
    • , Forest loss increases insect herbivory levels in human-altered landscapes, Acta Oecologica, 77, (136), (2016).
    • , Informational Mismatches: A Neglected Threat of Climate Change to Interspecific Interactions, Frontiers in Ecology and Evolution, 4, (2016).
    • , How soil and elevation shape local plant biodiversity in a Mediterranean hotspot, Biodiversity and Conservation, 10.1007/s10531-016-1113-y, 25, 6, (1133-1149), (2016).
    • , Temporal changes in the structure of a plant-frugivore network are influenced by bird migration and fruit availability, PeerJ, 4, (e2048), (2016).
    • , What are We Learning about Speciation and Extinction from the Canary Islands?, Ardeola, 63, 1, (15), (2016).
    • , Assessing ecological integrity: A multi-scale structural and functional approach using Structural Equation Modeling, Ecological Indicators, 10.1016/j.ecolind.2016.07.006, 71, (258-269), (2016).
    • , The signatures of Anthropocene defaunation: cascading effects of the seed dispersal collapse, Scientific Reports, 10.1038/srep24820, 6, 1, (2016).
    • , Removing interactions, rather than species, casts doubt on the high robustness of pollination networks, Oikos, 125, 4, (526-534), (2015).
    • , Landscape‐scale deforestation decreases gene flow distance of a keystone tropical palm, Euterpe edulis Mart (Arecaceae), Ecology and Evolution, 6, 18, (6586-6598), (2016).
    • , Forest Cover and Bird Diversity: Drivers of Fruit Consumption in Forest Interiors in the Atlantic Forest of Southern Bahia, Brazil, Tropical Conservation Science, 9, 1, (549), (2016).
    • , Declining relict plants: Climate effect or seed dispersal disruption? A landscape-scale approach, Basic and Applied Ecology, 10.1016/j.baae.2015.08.003, 17, 1, (81-91), (2016).
    • , Management of Protected Areas and Its Effect on an Ecosystem Function: Removal of Prosopis flexuosa Seeds by Mammals in Argentinian Drylands, PLOS ONE, 11, 9, (e0162551), (2016).
    • , Sampling networks of ecological interactions, Functional Ecology, 30, 12, (1883), (2016).
    • , The diverse effects of habitat fragmentation on plant–pollinator interactions, Plant Ecology, 10.1007/s11258-016-0608-7, 217, 7, (857-868), (2016).
    • , Patterns, Causes, and Consequences of Anthropocene Defaunation, Annual Review of Ecology, Evolution, and Systematics, 10.1146/annurev-ecolsys-112414-054142, 47, 1, (333-358), (2016).
    • , Birds in Ecological Networks: Insights from Bird-Plant Mutualistic Interactions, Ardeola, 63, 1, (151), (2016).
    • , Loss of seed dispersal before the loss of seed dispersers, Biological Conservation, 10.1016/j.biocon.2016.06.024, 201, (38-49), (2016).
    • , Habitat loss exacerbates regional extinction risk of the keystone semiarid shrub Ziziphus lotus through collapsing the seed dispersal service by foxes (Vulpes vulpes), Biodiversity and Conservation, 25, 4, (693), (2016).
    • , Increasing land use drives changes in plant phylogenetic diversity and prevalence of specialists, PeerJ, 4, (e1740), (2016).
    • , Selective logging in tropical forests decreases the robustness of liana–tree interaction networks to the loss of host tree species, Proceedings of the Royal Society B: Biological Sciences, 283, 1826, (20153008), (2016).
    • , Comparing diversity to flower-bee interaction networks reveals unsuccessful foraging of native bees in disturbed habitats, Biological Conservation, 202, (110), (2016).
    • , Endangered plant-parrot mutualisms: seed tolerance to predation makes parrots pervasive dispersers of the Parana pine, Scientific Reports, 6, 1, (2016).
    • , Disrupted learning: habitat degradation impairs crucial antipredator responses in naive prey, Proceedings of the Royal Society B: Biological Sciences, 10.1098/rspb.2016.0441, 283, 1830, (20160441), (2016).
    • , Forest succession and population viability of grassland plants: long repayment of extinction debt in Primula veris, Oecologia, 181, 1, (125), (2016).
    • , Predators, Prey and Habitat Structure: Can Key Conservation Areas and Early Signs of Population Collapse Be Detected in Neotropical Forests?, PLOS ONE, 11, 11, (e0165362), (2016).
    • , Human impact gradient on mammalian biodiversity, Global Ecology and Conservation, 10.1016/j.gecco.2016.01.004, 6, (79-92), (2016).
    • , Making dispersal syndromes and networks useful in tropical conservation and restoration, Global Ecology and Conservation, 10.1016/j.gecco.2016.03.002, 6, (152-178), (2016).
    • , Stability and generalization in seed dispersal networks: a case study of frugivorous fish in Neotropical wetlands, Proceedings of the Royal Society B: Biological Sciences, 10.1098/rspb.2016.1267, 283, 1837, (20161267), (2016).
    • , Implications of Habitat Loss on Seed Predation and Early Recruitment of a Keystone Palm in Anthropogenic Landscapes in the Brazilian Atlantic Rainforest, PLOS ONE, 10, 7, (e0133540), (2015).
    • , Towards an Integration of Biodiversity–Ecosystem Functioning and Food Web Theory to Evaluate Relationships between Multiple Ecosystem Services, Ecosystem Services - From Biodiversity to Society, Part 1, 10.1016/bs.aecr.2015.09.001, (161-199), (2015).
    • , The effect of land-use on the diversity and mass-abundance relationships of understory avian insectivores in Sri Lanka and southern India, Scientific Reports, 5, 1, (2015).
    • , Native and Non-Native Supergeneralist Bee Species Have Different Effects on Plant-Bee Networks, PLOS ONE, 10, 9, (e0137198), (2015).
    • , Does mixed-species flocking influence how birds respond to a gradient of land-use intensity?, Proceedings of the Royal Society B: Biological Sciences, 282, 1811, (20151118), (2015).
    • , Seven Shortfalls that Beset Large-Scale Knowledge of Biodiversity, Annual Review of Ecology, Evolution, and Systematics, 10.1146/annurev-ecolsys-112414-054400, 46, 1, (523-549), (2015).
    • , Meta-analysis of anthropogenic habitat disturbance effects on animal-mediated seed dispersal, Global Change Biology, 21, 11, (3951), (2015).
    • , Overfishing disrupts an ancient mutualism between frugivorous fishes and plants in Neotropical wetlands, Biological Conservation, 10.1016/j.biocon.2015.06.019, 191, (159-167), (2015).
    • , Conceptualizing Forest Degradation, Trends in Ecology & Evolution, 30, 10, (622), (2015).
    • , Frugivore downsizing and the collapse of seed dispersal by fish, Biological Conservation, 191, (809), (2015).
    • , Persistence of Plants and Pollinators in the Face of Habitat Loss, Ecosystem Services - From Biodiversity to Society, Part 1, 10.1016/bs.aecr.2015.09.005, (201-257), (2015).
    • , The structure of mixed-species bird flocks, and their response to anthropogenic disturbance, with special reference to East Asia, Avian Research, 10.1186/s40657-015-0023-0, 6, 1, (2015).
    • , Complexity and insidiousness of cryptic function loss mechanisms, Trends in Ecology & Evolution, 30, 7, (371), (2015).
    • , Resolving the life cycle alters expected impacts of climate change, Proceedings of the Royal Society B: Biological Sciences, 282, 1813, (20150837), (2015).
    • , Linking Biodiversity, Ecosystem Functioning and Services, and Ecological Resilience, Ecosystem Services - From Biodiversity to Society, Part 1, 10.1016/bs.aecr.2015.09.004, (55-96), (2015).
    • , Local extinctions of obligate frugivores and patch size reduction disrupt the structure of seed dispersal networks, Ecography, , (2018).
    • , Tropical Forest Fragmentation Affects Floral Visitors but Not the Structure of Individual-Based Palm-Pollinator Networks, PLOS ONE, 10.1371/journal.pone.0121275, 10, 3, (e0121275), (2015).
    • , Projected climate change threatens pollinators and crop production in Brazil, PLOS ONE, 10.1371/journal.pone.0182274, 12, 8, (e0182274), (2017).
    • , Regeneration of a keystone semiarid shrub over its range in Spain: habitat degradation overrides the positive effects of plant–animal mutualisms, Plant Biology, , (2018).
    • , Overlooked coral predators suppress foundation species as reefs degrade, Ecological Applications, , (2018).