Study area
The MPAY is located in the subantarctic region (57.05ºS, 65.36ºW), south of Tierra del Fuego, in the geographical area of the Drake Passage. The MPA constitutes the Southwestern end of the Argentine Exclusive Economic Zone, as well as a strategic area of great bio-regional importance as an oceanic corridor between the Atlantic and the Pacific oceans (Tombesi et al., 2020) (Fig. 1). The MPAY covers a marine surface of 68834.31 km2, and spans an area from the continental slope (below the 500 m isobath) to deep oceanic basins (> 3000 m), capturing a diversity of reliefs such as slopes, canyons and seamounts (Palma et al., 2021).
The MPAY receives waters from different origins, creating a complex oceanographic environment. The Antarctic Circumpolar Current (ACC) impacts much of the southern and central sectors. This current flows eastwards and carries cold and nutrient-rich subantarctic waters from the Drake Passage along the upper portions of the western slope of the Argentine basin (Piola and Gordon, 1989; Peterson and Whitworth, 1989; Acha et al., 2004; Romero et al., 2006). The waters of the ACC mix at its northern boundary with those from the Cape Horn current which also flow Southwest-Northeast (Guihou et al., 2020). Moreover, the northern sector of the MPAY receives waters from the cold estuarial front of the Beagle Channel (Martín et al., 2023).
Construction of the food web
We constructed the network of predator-prey interactions for the MPAY using the species list published in the Argentinean Biodiversity Information System for the area as a starting point (https://sib.gob.ar/#!/area-protegida/area-marina-protegida-yaganes). Afterwards, we performed a literature search using the name of the species in question and the keywords ‘diet’, ‘prey’, ‘predator’, ‘feeding ecology’ and ‘trophic ecology’ in Google Scholar and Scopus. This process identified more than 50 references including peer-reviewed published articles, Ph.D. theses and public databases (Supplementary Material, Table S1). Because of the limited information available regarding predator-prey interactions among the species inhabiting and taking advantage of the MPA as a feeding ground, we proceeded with the assumption that if two species interact in an ecosystem similar to that of Yaganes and both coexist in the MPA, then such interaction was considered valid for the MPAY. Thus, we included predator-prey interactions mentioned by (Marina et al., 2024) for the MPAN-BB, and by (López-López et al., 2022) for the Scotia Sea, if both species coexist in the MPAY. Furthermore, in those cases in which two or more taxa presented the same set of predators and prey, the taxa were collapsed into a single trophic species (e.g. Decapoda, Copepoda, Isopoda). In all cases, we considered only adult diets for each species. As a way of validating the list of trophic interactions resulting from the mentioned methodology, we consulted with experts on the different functional groups and species inhabiting the area (refer to Acknowledgements for the experts we contacted). We decided to limit our scope to pelagic, bentho-pelagic and demersal species due to the lack of information for the benthic habitat, which is linked to the complex bathymetry of the MPA.
Once the compiled list of predator-prey interactions was validated from experts, we transformed this data into a graphical representation, with nodes representing trophic species, edges indicating predator-prey interactions, and arrows pointing from prey to predator. This directional setup mirrors the flow of matter and energy within the ecosystem. Each node represents either a biological species or a group of taxa sharing identical sets of prey and predators, collectively referred to as trophic species (Cohen and Briand, 1984).
The network of predator-prey interactions constructed represent potential trophic relationships among pelagic, benthopelagic and demersal species without considering spatial and temporal variations. The geographical location and extensive surface, together with the influence of various sea fronts, ocean currents and complex bathymetry make the MPA a highly diverse area, affected by different biological and physical processes (Guihou et al., 2020). Due to this, it is expected that the species diversity, as well as the interactions among them will vary spatially and temporally.
The method used to construct the food web was based on a bibliographic review, so the obtained product is not exhaustive, but represents a baseline and solid knowledge about potential predator-prey interactions and the importance of them for the integrity of the entire ecosystem. The MPAY is an incredibly vast, complex and diverse area of international interest due to its geographical location and the physical conditions that characterise it, so establishing a food web with predator-prey interactions with higher resolution, based on a more exhaustive bibliographic review combined with new information would be important. Furthermore, it would be also necessary to study the benthic community across all the MPA, and the way it interacts with the pelagic habitat, to better understand the ecological dynamics of such a complex ecosystem.
Data analysis
The food web was analysed at two levels: network-level and species-level. The former approach considers all species and interactions within the network, aiming to describe the complexity and structure of the entire food web. The latter analysis considers the interactions and species related to a particular species, with the objective of describing the role played by the individual species.
Network-level analysis: In order to characterise the food web’s complexity and structure, we calculated the following properties: number of species (S), number of interactions (L), density of interactions (L/S), connectance (C), percentage of basal, intermediate and top species, path length (PL), clustering coefficient (CC) and small world pattern (SW). In a network context, basal species represent the energy sources of the food web (e.g. phytoplankton, detritus), intermediate species present both prey and predators (i.e. primary and secondary consumers), and top species are consumer without predators (Cohen and Briand, 1984).
Connectance (C) is defined as the ratio between the number of established interactions (L) and the number of possible interactions in the food web (S2). This property is a key descriptor of the network structure (Martinez, 1992) and a good estimator of the community's sensitivity to disturbances (Dunne et al., 2002a; Montoya et al., 2006).
The path length (PL) is a global network property that measures the distance between species in the food web; here we consider the shortest average distance between species (Watts and Strogatz, 1998). The clustering coefficient (CC) is defined as the average fraction of node pairs connected to the same nodes, which are also connected to each other; provides information about the clustering of nodes within their immediate neighbourhood (Delmas et al., 2018). These two network properties are typically analysed in order to gain insight into the small world pattern (SW) (Watts and Strogatz, 1998). To determine whether the food web presented this pattern, we compared the empirical values of CC and PL with those resulting from 1000 randomly generated networks created using the Erdös-Rényi model (Erdös and Rényi, 1959), with the same number of species (S) and number of interactions (L), following the method proposed by Marina et al. (2018).
Additionally, we studied the degree distribution, which measures the probability that a species has k interactions within the network. This is calculated as P(k) = N(k)/S, with N(k) being the number of nodes with k interactions (Albert and Barabási, 2002). From the analysis of this property, important nodes of the network can be identified, such as potential generalist species (i.e., species with a diverse diet, that is, with a high number of interactions), and specialists (i.e., species with a narrow diet, that is, with a low number of interactions). We considered both in-degree (i.e., number of preys) and out-degree (i.e., number of predators) of each species.
Species-level analysis: The species-level approach focused on determining the individual role of species in the context of the food web, given that not all species in large communities fulfil the same ecological role, nor are they equally important for network processes and properties (Delmas et al., 2018). For this, we considered the following properties: a) degree, b) closeness, c) betweenness, and d) trophic level. The first three correspond to centrality properties, that is, referring to a particular node in the network and its relationship with the rest of the nodes, and quantify the influence of a species within the network (Delmas et al., 2018). The degree is the total number of interactions of a species, considering in- (number of predators) and out-interactions (number of prey) (Jordán, 2009); it quantifies the immediate influence between species (Delmas et al., 2018). Those species with a high degree are the most connected, so their influence on the robustness (i.e. fraction of primary species loss that induces at least 50% total species, Cordone et al., 2018) of the network is high. The extent of trophic connections a species holds within a food web, the greater its potential to affect community structure (Dunne et al., 2002b). We calculated the cumulative degree distribution for the food web, and fitted it to exponential, power law and truncated power law distribution models, using the AIC values as decision criterion.
Closeness measures the proximity of a species to all other species in the network. It is based on the shortest path between two species and therefore indicates how efficiently perturbations about a species are likely to influence the network (Delmas et al., 2018). Regarding betweenness, it is defined as the number of times that a species is found between a pair of other species in the network (Delmas et al., 2018), being an ideal measure to study the influence of species loss in fragmentation processes (Chadès et al., 2011; Earn et al., 2000; McDonald-Madden et al., 2016), or in the face of the dispersion of disturbances.
The trophic level measures the distance, in number of interactions, between a species and the basal species of the network (Lindeman, 1942). It represents the number of feeding links separating an organism from the base of production (Thompson et al., 2007), and indicates basal, intermediate and top species.
Centrality indices have been used to identify possible key species in ecological networks (Jordán and Scheuring, 2004; Martín González et al., 2010); key species being understood as those that interact strongly (Davic, 2003) and are disproportionately important in their community (Mouquet et al., 2013). Based on this, we created a Keystone Species Index (KSI). To do this, we ranked species according to each centrality property: the higher the degree, closeness and betweenness values, the higher the species rank. Then we averaged these three rankings to obtain the KSI for every species. In this way, the KSI takes values from 1 (species with the highest centrality indices, first position in each ranking) to the number of species in the network (species with the lowest centrality indices, last position in each ranking). When two or more trophic species have the same KSI value, they share the same position in the ranking. Furthermore, we explored the relationship between the KSI and the trophic level, with the aim of gaining insights into the trophic position of the relatively most important species in the network. Thus, we considered the trophic level as the dependent variable and the KSI as the independent variable. Fitting was done locally, meaning that for the fit at point x, the fit is made using points in a neighbourhood of x, weighted by their distance from x (Cleveland et al., 1992).
Analyses were carried out using R software (version 4.1.2, R Core Team 2023), particularly the packages igraph version 1.4.2 (Csárdi et al., 2023), multiweb version 0.6.8 (Saravia, 2023) and NetIndices version 1.4.4.1 (Kones et al., 2009). The source code and data are available at the following GitHub repository: https://github.com/melinascian/AMP-YAGANES.git.