Incorporating extinction risk and realistic biodiversity futures: implementation of trait-based extinction scenarios

Martin Solan, Finlay Scott, Nicholas K. Dulvy, Jasmin A. Godbold, and Ruth Parker

Introduction

Rates of species extinction have intensified over the last century (Barnosky et al. 2011), largely as a product of human activity (Sala et al. 2000; Halpern et al.

2008), climate forcing (Thomas et al. 2004), and their interactions (Brook et al. 2008), and it is presently anticipated that extinction rates will continue to increase well into the next century (Pereira et al. 2010). Concerns over the potentially important consequences that this may have for the stability and functioning of ecosystems has provided impetus for research, and an extensive body of literature now exists that explicitly focuses on the ecological and environmental consequences of biodiversity loss (reviewed in Balvanera et al. 2006; Cardinale et al. 2006; Schmid et al. 2009; Solan et al. 2009). Collectively, this knowledge base provides unambiguous evidence that, irrespective of the system and the cause of expiration, declining species richness tends to disrupt processes that maintain ecosystem integrity. This occurs because, as species are removed, their absence—or reduction in abundance and/or biomass—directly causes: (i) a reorganization of sampling/selection effects (Aarssen 1997; Loreau and Hector 2001); (ii) a reduction in the level of interspecific resource partitioning; and, (iii) a declining probability that non-additive interspecific interactions—facilitative or inhibitive——will take place. In addition, a range of functional effects may occur in response to the indirect effects of species loss on specific surviving species through various mechanisms, including release from competition (Godbold et al. 2009) or predation (Burkepile and Hay 2007) , breakdown of mutualistic associations ( Harrison 2000 ; Hughes et al. 2009 ) or the cascading effects of co-extinction (Moir et al. 2009, Fowler 2010). Whilst it is likely that a subset of both direct and indirect mechanisms will simultaneously operate in natural systems subject to extinction forcing, their relative role in altering ecosystem properties is only just emerging; random assembly experiments where researchers have manipulated species diversity have, more often than not, predominantly attributed the functional consequences of species loss to a combination of the loss of the single most productive species and/or a reduction in species complementarity (Cardinale et al. 2006 ).

Accepting the mechanistic processes that may underpin the ecological consequences of biodiversity loss, a general difficulty of small-scale experiments is that it is not tractable to work with the full complement of species that represent a natural system, or to explore fully the range of likely—or anticipated——alternative extinction scenarios that such assemblages may encounter (Naeem 2008). A number of notable contributions have attempted to address these logistic difficulties by using information from naturally assembled communities to refine species selection and limit the number of

Marine Biodiversity and Ecosystem Functioning. First Edition. Edited by Martin Solan, Rebecca J. Aspden, and David M. Paterson. © Oxford University Press 2012. Published 2012 by Oxford University Press.

experimental assemblage combinations (Zavaleta and Hulvey 2004; Srinivasan et al. 2007; Bracken et al. 2008; de Visser et al. 2011), but the majority of experiments examining how altered levels of biodiversity affect the functioning of ecosystems have incorporated communities that do not exist in natural systems, and they have generally assumed that species loss is random (Naeem 2008). These simplifying assumptions have received substantial criticism—summarized in Naeem et al. 2002, Solan et al. 2009—because, far from being a random process, extinction of species tends to follow an ordered sequence that reflects the relative distribution of extinction risk amongst species within an assemblage. Hence the ecological significance of extinction will be dependent on the sequential order of species loss and whether the extinction risk of each species correlates with the life-history traits that are important in mediating ecosystem functioning (Solan et al. 2004). If they do, on average the ecological outcome is likely to be exacerbated relative to that observed following a random extinction where extinction takes place irrespective of the distribution of species traits (Gross and Cardinale 2005) . Hence there is a compelling need to determine the species and traits—effect and response traits (Hooper et al. 2002)—that are important for specific ecosystem functions, and assess the fate of species endangered by multiple drivers of change across a range of systems.

Identifying the threat status of species is difficult, partly because the necessary information—geographic distribution, temporal changes in population structure, changes to ecological function—is difficult to obtain or not available, but also because there is a lack of generally accepted criteria for identifying when species become impoverished or extinct (Nicholson et al. 2009). The International Union for Conservation of Nature (IUCN) Red List, a global index of the state of degeneration of biodiversity, is based on a set of transparent and quantitative criteria—largely based on geographical distribution and population status (IUCN 2001, Mace et al. 2008)—that assess the threat status of a species—i.e. least concern, near threatened, vulnerable, critically endangered, or endangered. These categorizations are useful in other contexts, but are of limited value when parameterizing models with sequential species losses as the categorizations are very broad in coverage. Furthermore, there are difficulties in determining whether species are genuinely at risk at local scales (Dulvy et al. 2003). There are a variety of alternative ways, however, to estimate the relative extinction vulnerability of a species including, for example, the use of local knowledge and the grey literature (Castellanos- Galindo et al. 2011), the examination of the distribution of species along gradients of disturbance (Pearson and Rosenberg; 1978, Vitousek et al. 1994; Isbell et al. 2008 ; Hall-Spencer et al. 2008 ), and the collation of information on the physiological tolerance (Labrune et al. 2006; Srinivasan et al. 2007), or behavioural sensitivity (Liow et al. 2009 ) of species to specific agents of perturbation. In addition, a suite of metrics and protocols (reviewed in Schlapfer et al. 2005) are available that have been developed to determine objectively the conservation priority status of various groups of species, most notably species with high conservation status (fish, Dulvy et al. 2004; mangroves, Polidoro et al. 2010; seagrass, Short et al. 2011), but also for less charismatic but functionally important groups (benthic invertebrates, Freeman et al. 2010). These range from simple metrics based on specific traits, such as body size (Olden et al. 2007), to more sophisticated multivariate frameworks that encompass multiple traits (Bremner et al. 2003) and species-environment interactions (Branco et al. 2008 ; Graham et al. 2011 ). The resulting ranked inventories have the advantage that they provide continuous data on the relative vulnerability of species, allowing subtle differences in risk to be accounted for, and the arbitrary division of species into specific risk groupings of equal ranking to be avoided. Notwithstanding the need to test rigorously the sensitivity and appropriateness of such metrics, it is clear that most of these methodologies are readily transferrable to most marine species and habitats, thereby offering a means to generate credible sequences of extinction risk for specific populations and context.

For a predictive understanding of the ecological consequences of extinction, inventories of species vulnerabilities to extinction must be coupled with information on the magnitude, functional role, and extent to which species perform unique contributions—i.e. the degree of trait overlap within an assemblage (Naeem et al. 2002; Micheli and Halpern 2005; Zhang and Zhang 2007. Fortunately, there is a long tradition in ecology of documenting the functional effects of species (additive) and assemblages (non-additive) on a variety of ecosystem properties in the presence and absence of a wide range of abiotic stressors. From this repository of information, we know that the importance of individual species will be context-dependent, and even subtle changes in species behaviour in response to a range of abiotic (e.g. temperature, Beveridge et al. 2010; food availability, Nogaro et al. 2008; elevated CO₂, Bulling et al. 2010; habitat properties, Larsen et al. 2005, Godbold et al. 2011) and biotic (presence ofpredator, Maire et al. 2010) drivers of change can have dramatic effects on ecosystem properties. These changes in community dynamics and species interactions can frequently alter, or even reverse, species vulnerabilities to environmental forcing (Griffen and Drake 2008 ; Ball et al. 2008), and may be a product of multiple interacting components of an ecosystem (Edgar et al. 2010). Despite such complexities, however, species within an assemblage can be readily grouped into cohorts that share common traits based on the way they modify a given ecosystem process (Hooper et al. 2002), making it feasible to assemble inventories of species that are likely to respond to particular perturbations in a similar manner. Indeed, many readily discernable species traits, such as body size and trophic position, are known to form intimate linkages with many ecosystem properties (Bremner et al. 2006; Fisher et al. 2010; Laughlin 2011), such that a priori predictions can be made that the preferential loss or reduction of these traits within an assemblage may have disproportionate effects on ecosystem functioning (Hillebrand and Matthiessen

2009). For species where there are information gaps or incomplete knowledge about species traits, iterative hierarchal schemes for assigning functional group classifications can be adopted (see Figure 17.2 in Hooper et al. 2002; e.g. benthic macrofauna, Swift et al. 1993; Solan et al. 2004; fish, McIntyre et al. 2007) that allow functional groups to be assigned based on secondary information—i.e. possession of shared or similar traits and/or expert opinion.

Such eechniques allow information on the role of individual species to be extended to entire assemblages and, when coupled with large datasets on extinction vulnerability, provide sufficient information to examine alternative ecologically realistic extinction scenarios (Coreau et al. 2009 ).

One way in which this can be achieved is through the use of trait-based extinction scenarios, where the sequence of species extinction can be ordered in accordance with specific extinction mechanisms, and the functional consequences of each scenario can be estimated. The consequences of possible biodiversity-environment futures for ecosystem functioning have been explored in this way across a range of freshwater, terrestrial, and marine habitats, and for a variety of ecosystem functions, including productivity (Smith and Knapp 2003; Schlapfer et al.

• 2005), resistance to invasive species (Zavaleta and Hulvey 2004) , nutrient cycling (Solan e t al. 2004 , McIntyre et al. 2007 ), carbon storage ( Bunker et al.
• 2005), and decomposition (Ball et al. 2008). The flexibility and adaptability of this approach is yet to be fully exploited, but marine ecologists are particularly well placed to take advantage of the technique as the data required to generate probabilistic numerical simulations is routinely collected as a matter of course. In this chapter, we provide the relevant model code for a range of extinction scenarios and demonstrate how this modelling framework can be applied at local and regional scales. Our objective is to encourage widespread application of modelled predictions and foster dialogue on how such methodologies can be expanded and improved.