Landscape configuration affects probability of apex predator presence and community structure in experimental metacommunities

Abstract

Biodiversity is declining at an unprecedented rate, highlighting the urgent requirement for well-designed protected areas. Design tactics previously proposed to promote biodiversity include enhancing the number, connectivity, and heterogeneity of reserve patches. However, how the importance of these features changes depending on what the conservation objective is remains poorly understood. Here we use experimental landscapes containing ciliate protozoa to investigate how the number and heterogeneity in size of habitat patches, rates of dispersal between neighbouring patches, and mortality risk of dispersal across the non-habitat 'matrix' interact to affect a number of diversity measures. We show that increasing the number of patches significantly increases γ diversity and reduces the overall number of extinctions, whilst landscapes with heterogeneous patch sizes have significantly higher γ diversity than those with homogeneous patch sizes. Furthermore, the responses of predators depended on their feeding specialism, with generalist predator presence being highest in a single large patch, whilst specialist predator presence was highest in several-small patches with matrix dispersal. Our evidence emphasises the importance of considering multiple diversity measures to disentangle community responses to patch configuration.

Keywords: Dispersal; Diversity; Heterogeneity; Protected area; SLOSS.

Fig. 1

Schematic representation of the five reserve configurations plus dispersal manipulations. Habitat patches (grey circles), their volume in mL (number within grey circles), plus potential local dispersal (dotted lines) and matrix dispersal (solid lines) are all shown. Dotted lines represent one local dispersal ‘event’, and solid lines represent one matrix dispersal ‘event’

Fig. 2

Effects of heterogeneity and number of patches in an ecosystem for each matrix dispersal regime on the probability of a landscape containing a generalist predators and b specialist predators. Lines are model-averaged binomial GLM outputs, shaded bands are 95% confidence intervals, and points are observed data points (N = 12 for each treatment except 4HoHM (N = 11), 6HoHM (N = 10), and 6HeN (N = 11) due to leaking microcosms)

Fig. 3

The effects of heterogeneity and number of patches for each dispersal regime on γ diversity. Columns are local dispersal quality level and rows are matrix dispersal quality level. Lines are model-averaged Gaussian GLM coefficients from the top models (∆AICc > 2), shaded polygons are 95% confidence intervals, and points are the observed data points (N = 4 for each treatment except 4HoHLHM, 6HoNLHM, 6HoHLHM, 6HeHLNM (all N = 3) due to leaking microcosms)

Fig. 4

Effect of heterogeneity and number of patches in an ecosystem on overall number of extinctions in a landscape. Top models (∆AICc < 2) did not include matrix or local dispersal as explanatory variables so they are not presented here. Lines are model-averaged Quasi-Poisson GLM model outputs, shaded polygons are 95% confidence intervals, and points are observed data points (N = 36 for each treatment except 4Ho (N = 35), 6Ho (N = 34), and 6He (N = 35) due to leaking microcosms)

Fig. 5

Probability of specialist predator presence as a function of probability of generalist predator presence. Fitted line is the logistic regression, shaded polygon shows 95% confidence intervals, and points are observed data points, jittered slightly for clarity. Relationship was significant (P = 0.0000023; N = 734)