The biggest losers: habitat isolation deconstructs complex food webs from top to bottom

Abstract

Habitat fragmentation threatens global biodiversity. To date, there is only limited understanding of how the different aspects of habitat fragmentation (habitat loss, number of fragments and isolation) affect species diversity within complex ecological networks such as food webs. Here, we present a dynamic and spatially explicit food web model which integrates complex food web dynamics at the local scale and species-specific dispersal dynamics at the landscape scale, allowing us to study the interplay of local and spatial processes in metacommunities. We here explore how the number of habitat patches, i.e. the number of fragments, and an increase of habitat isolation affect the species diversity patterns of complex food webs (α-, β-, γ-diversities). We specifically test whether there is a trophic dependency in the effect of these two factors on species diversity. In our model, habitat isolation is the main driver causing species loss and diversity decline. Our results emphasize that large-bodied consumer species at high trophic positions go extinct faster than smaller species at lower trophic levels, despite being superior dispersers that connect fragmented landscapes better. We attribute the loss of top species to a combined effect of higher biomass loss during dispersal with increasing habitat isolation in general, and the associated energy limitation in highly fragmented landscapes, preventing higher trophic levels to persist. To maintain trophic-complex and species-rich communities calls for effective conservation planning which considers the interdependence of trophic and spatial dynamics as well as the spatial context of a landscape and its energy availability.

Keywords: allometry; bioenergetic model; dispersal mortality; food webs; landscape structure; metacommunity dynamics.

Figure 1.

Conceptual illustration of our modelling framework. In our meta-food-web model (b), we link local food web dynamics at the patch level (a) through dynamic and species-specific dispersal at the landscape scale (d). We consider landscapes with identical but randomly distributed habitat patches, i.e. all patches have the same abiotic conditions, and each patch can potentially harbour the full food web. We model fragmented landscapes which differ in the number of habitat patches and the mean distance between patches (c).

Figure 2.

Heatmaps visualizing $\overline{\alpha }$-, β- and γ-diversity (colour-coded; z-axis) in response to habitat isolation, i.e. the mean patch distance ($\overline{\tau }$, log₁₀-transformed; x-axis) and the number of habitat patches (Z; y-axis), respectively. We generated the heatmaps based on the statistical model predictions (see the electronic supplementary material).

Figure 3.

Top row: Mean biomass densities [log₁₀(biomass density + 1)] of animal consumer species (a) and basal plant species (b) over all food webs (B_i, log₁₀-transformed; y-axis) in response to habitat isolation, i.e. the mean patch distance ($\overline{\tau }$, log₁₀-transformed; x-axis). Each colour depicts the biomass density of species i averaged over all food webs: (a) colour gradient where orange represents the smallest, red the intermediate and blue the largest consumer species; (b) colour gradient where light green represents the smallest and dark green the largest plant species. Bottom row: Mean species-specific landscape connectance (ρ_i; y-axis) for consumer (c) and plant species (d) over all food webs as a function of the mean patch distance ($\overline{\tau }$, log₁₀-transformed; x-axis). See the electronic supplementary material, figure S9 for standard errors in biomass densities for four exemplary species.