Introduction
A shift in focus from species to ecological networks of interactions has recently been proposed as a necessary step in the adaptation of conservation goals to the maintenance of ecosystem integrity and the services ecosystems underpin (Scotti & Jordán 2010; Harvey et al. 2017; Pecl et al. 2017). Ecological network enables to evaluate the vulnerability of ecosystems to a perturbation through the study of the changes in the structure of the network (Tylianakiset al. 2007; Stouffer & Bascompte 2011; Hattab et al.2016; Robinson & Strauss 2020). Its use to assess ecosystem state has increased in the recent years as it allows considering in a single framework the effects of fluctuation in species’ abundance and their preys and predators, but also on indirectly linked species and the whole network itself (Jordán et al. 2006; Wallach et al. 2017).
The risk of network collapse can be characterized by its robustness to species extinction, namely its capacity to withstand the cascading perturbation generated by the removal of species (Dunne et al.2004; Dunne & Williams 2009; Jonsson et al. 2015). Human impact on ecosystems is intensifying and has already caused numerous collapses of species (Duarte et al. 2020). As ecosystems are being degraded at an unprecedented rate, the need to understand which perturbation sequences are expected to be more devastating than others has become pressing (Jonsson et al. 2015). The collapse of some well-connected species is expected to have a disproportionate impact on their ecosystems relatively to their biomass, and these species have to be identified (Jordán 2009; Klemm et al. 2012; Worm & Paine 2016). Central species with a large number of interactions, are likely to influence numerous species and thus have been defined as network hubs that should be prioritized for conservation (Curtsdotter et al.2011). To identify these central species, mesoscale measures (i.e. intermediate level between the species (local) and the entire network (global)), such as eigenvector centrality, betweenness centrality or closeness centrality have proven to be particularly suited for the assessment of species importance in the spread of a perturbation across the network (Estrada 2007; Jordán 2009).
Sensitivity and vulnerability assessment of species (defined as sensitivity added to exposure (IPCC 2001)) have largely been conducted based on species traits (Tillin et al. 2010; Le Quesne & Jennings 2012; van Treeck et al. 2020). However, to evaluate properly the potential of a species to spread perturbation across the network, its centrality should be assessed alongside its sensitivity and exposure to a given pressure. Indeed, a species can be sensitive and vulnerable but not central, or it may be central but not sensitive nor vulnerable. In both cases, the species would not spread a perturbation across the network. In that sense, one can ask whether the robustness of the trophic network to the spread of a perturbation can be inferred from the sensitivity of its components.
To answer this question and shift the sensitivity assessment focus from species to ecological network, we propose a framework that examines network robustness to a given perturbation at the local (species), mesoscale and global (network) level, based on species traits and on the topology of the network. Specifically, we investigated the impact on the robustness of the network of the loss of the sensitive species, the exposed species and the central species.
We applied this framework to fishing pressure robustness of a trophic network from a historically exploited fishing ground, the Celtic Sea. Fishing impacts on the ecosystems are numerous, from decreased species abundance and depletion of higher trophic levels (Pauly & Palomares 2005) to altered trophic networks as fishing pressure increases (Gilarranz et al. 2016). At the species level, life history traits (e.g. maximum length, longevity or age at maturity) are good proxies of species’ demographic characteristics and enable to characterize their sensitivity to fishing by ranking them along a “slow-fast” continuum of life history strategies (Le Quesne & Jennings 2012). Fishing tends to favour small-sized, short-lived species that mature early and have a better capacity to recover after a fishing perturbation (Jennings et al. 1998; Le Quesne & Jennings 2012; Wiedmann et al. 2014). In the Celtic Sea, the intense development of fishing in the area during the second half of the 20th century until its climax in the late 1980’s, deeply altered the ecosystem structure through the depletion of large demersal predators, i.e. cod, whiting, hake and sole (Guénette & Gascuel 2012; Hernvann & Gascuel 2020). Based on literature and isotopic measurements for 69 taxa (including fishes, elasmobranches, cephalopods, bivalves, and crustaceans), we assessed the vulnerability to fishing of both taxa and the trophic network structure. Specifically, we investigate whether (i) the most sensitive and the most exposed taxa to fishing are the most central; and whether (ii) the loss of sensitive, exposed or central taxa is the most detrimental for the network robustness. For a better understanding of the processes at play, we tested the robustness of the network against different species ‘removal sequences, and notably assessed the importance of the number of predators and preys.