Spatial ecological networks: planning for sustainability in the long-term

Abstract

Humans are producing complex and often undesirable social and ecological outcomes in many landscapes around the world. To sustain biodiversity and ecosystem services in fragmented landscapes conservation planning has turned to the identification and protection of large-scale spatial ecological networks (SEN). Now widely adopted, this approach typically focuses on static connectivity, and ignores the feedbacks between changes to the network's topology and the eco-evolutionary dynamics on the network. We review theory showing that diversity, stability, ecosystem functioning and evolutionary adaptation all vary nonlinearly with connectivity. Measuring and modelling an SEN's long-term dynamics is immensely challenging but necessary if our goal is sustainability. We show an example where the robustness of an SEN's ecological properties to node and link loss depends on the centrality of the nodes targeted. The design and protection of sustainable SENs requires scenarios of how landscape change affects network structure and the feedback this will have on dynamics. Once established, SEN must be monitored if their design is to be adapted to keep their dynamics within a safe and socially just operating space. When SEN are co-designed with a broad array of stakeholders and actors they can be a powerful means of creating a more positive relationship between people and nature.

Figure 1

(a) A graphical representation of a landscape showing a network of forest patches embedded in other networks of human land covers: landscapes are networks of networks (image from Encyclopedia Britannica 2013). (b) The network of the forest patches (nodes) is connected by weighted links defining flows of resources, energy, information or organisms among forest patches, but also to flows with other networks (e.g. fields, or rivers). Analysis can be used to assess the robustness of important variables, like total network flow, to node loss. Here the loss of a single node (dotted circle) leads to the fragmentation of the network. (c) Typically, a spatial ecological network (SEN, see main text for definition) is identified for protection based on a set of objective conservation criteria. Criteria for investment include the number, quality and configuration of forest nodes and the spatial extent of the network that best meets the criteria for sustainability of the species set and ecosystem services offered by the SEN (e.g. support for wild pollinators valuable to fruit production). An SEN can move from a sustainable to a non-sustainable region (arrow) of performance space through changes in key structural features, such as network area and connectivity. Many networks (black dots) could conceivably be sustainable and meet some or all of the conservation criteria. In most instances the spatial dynamics of the SEN network are not modelled or used to inform conservation planning. Without models of the eco-evolutionary dynamics of the SEN and how human intervention in the landscape mediates the feedback between network topology and dynamics it is impossible to assess the long-term sustainability of the network.

Figure 2

The spatial insurance hypothesis [see 37] connects the spatial structure of a network the dynamics of the communities embedded within the metacommunity and to the emergent diversity and ecosystem functioning at node and network levels. (a) Node and network diversity vary with dispersal rate: at very low dispersal rates (thin arrows) each habitat patch maintains a single species (colored circles correspond to the presence and abundance of different species) that is best adapted to the local conditions in each patch. At intermediate dispersal rates the number of species per patch is maximal because of source–sink effects. Note that each patch maintains several species but that only one species is dominant (large colored circle) whereas the others are of low abundance (small colored circle). Ecosystem productivity (b) is greatest, and (c) the variability of productivity, measured by the coefficient of variation, (CV) is lowest at intermediate rates of dispersal (d = 0.01) because of the insurance effects of biodiversity and the spatial-averaging of environmental heterogeneity allowed by dispersal (see text for explanation). At high dispersal rates only one species is present throughout the metacommunity. This species is the best competitor under the average conditions across all patches and excludes all other species. Biodiversity has been lost and ecosystem productivity and stability are maintained only by spatial-averaging. In general, we do not know where SEN lie on this spectrum of connectivity, so we do not know how changes to connectivity will alter the spatial insurance effects present in a landscape.

Figure 3

Impact of habitat loss on SEN diversity, ecosystem functioning and stability (adapted from [38]). A metacommunity (dispersal rate = 0.015) on an environmentally heterogeneous network (node color indicates environmental condition at one point in time). (a) An intact network, and a fragmented network after 14 patches have been removed based on three removal sequences: removing the patch with the minimum betweenness centrality (yellow triangle) — betweeness centrality is the value of an individual habitat patch in adding to the connectivity of the metacommunity by being a stepping-stone for dispersing individuals — removing a random patch and removing the patch with the maximum betweenness centrality (red triangle). The impact of each patch removal sequence on (b) mean local species richness, (c) mean local biomass and d) mean local biomass variability (CV = coefficient of variation). Lines are mean values from 100 replicate simulations and ribbons show the range between the 2.5th and 97.5th percentiles of the data. Metacommunities are not robust to the loss of habitat nodes of maximum betweeness centrality.

Figure 4

The safe operating space (SOS) for an SEN defined as a region of dynamic multivariate space that depends on the structural and functional properties of the network. Here we show the SOS for a single variable, community biomass. (a) The fluctuations of the SEN are initially bounded, but as nodes are lost the fluctuations increase in variance and leave the SOS. The dotted box indicates the period over which the SEN is within the SOS. At a critical level of node and link loss the variance increases through time, the network is no longer functionally connected, and so the SEN leaves the SOS and biomass collapses. The SOS can be defined non-arbitrarily based on the SEN structure that ensures long-term persistence. The degree of network erosion/protection will depend on how risk adverse society is with respect to the fluctuations, and the SEN’s robustness to node deletion. (b) and (c) show the biomass fluctuations as increasing orbits in phase space at the network-level and the node-level respectively. The dynamics are taken from the model described in Figure 3, where patches of min betweenness centrality are targeted for removal from the network.

Figure 5

The cycle of steps involved in identifying an SEN. This cycle of analysis expands on [69] by including the dynamics of the network’s properties as a criterion for node and link prioritization. For a given landscape six steps are structured within a loop to identify priorities for node and link protection: (1) identify focal criteria, in this case species with a range of life-history characteristics and habitat preferences; (2) identify habitat and dispersal networks of each species from monitoring data and expert opinion; (3) analyse the connectivity of species-specific habitat networks and quantify the resistance of the landscape and the contributions of each habitat pixel to short-range and long-range connectivity; (4) project the ecoevolutionary dynamics of the species occupying the SEN; (5) identify the spatial prioritization of habitat patches for conservation action based on the network criteria, such as short and long distance connectivity, that maintain the dynamics within the safe operating space. Additional criteria may include ecosystem services supplied by these species or the habitat nodes they occupy (e.g. carbon stored in forest); (6) establish the effectiveness of different prioritization schemes into the future based on climate change projections and spatially explicit dynamic land-use change simulations.