The strength of the biodiversity-ecosystem function relationship depends on spatial scale

Abstract

Our understanding of the relationship between biodiversity and ecosystem functioning (BEF) applies mainly to fine spatial scales. New research is required if we are to extend this knowledge to broader spatial scales that are relevant for conservation decisions. Here, we use simulations to examine conditions that generate scale dependence of the BEF relationship. We study scale by assessing how the BEF relationship (slope and R²) changes when habitat patches are spatially aggregated. We find three ways for the BEF relationship to be scale-dependent: (i) variation among local patches in local (α) diversity, (ii) spatial variation in the local BEF relationship and (iii) incomplete compositional turnover in species composition among patches. The first two cause the slope of the BEF relationship to increase moderately with spatial scale, reflecting nonlinear averaging of spatial variation in diversity or the BEF relationship. The third mechanism results in much stronger scale dependence, with the BEF relationship increasing in the rising portion of the species area relationship, but then decreasing as it saturates. An analysis of data from the Cedar Creek grassland BEF experiment revealed a positive but saturating slope of the relationship with scale. Overall, our findings suggest that the BEF relationship is likely to be scale dependent.

Keywords: Jensen's inequality; ecosystem functioning; nonlinear averaging; spatial scale; species richness; β-diversity.

Figure 1.

Illustration of the shape of the BEF relationship with different values of b. Higher values of b are shown in panel (a), lower values of b are shown in panel (b). This separation was done for clarity to show the curvature of the BEF relationship for low values of b.

Figure 2.

Illustration of how the BEF relationship changes with spatial scale in cases I–IV. Panel (a) shows how the BEF slope, b_A changes relative to the mean local slope , at different spatial scales (see electronic supplementary material, figure S1 for raw values of b_A). Panel (b) shows the R² of this relationship. The four cases are shown in the different panels—no variation in local species richness or local b_i (case I), variation in local a_i (case II), variation in local S (case III) or variation in local b_i (case IV). The solid line shows the median across 100 replicate simulations each consisting of 2000 replicate regions at each scale. Inter-run variability omitted for clarity in panel (a), but shown in electronic supplementary material, figure S1. In panel (b), the interquartile range is smaller than the width of the lines and so is not shown. (Online version in colour.)

Figure 3.

The strength of biodiversity effects, b_A (panel (a)) and R² (panel (b)), at different spatial scales when there is incomplete compositional turnover across local patches. Different degrees of compositional turnover are indicated by colour (low values of B₁ correspond to low turnover, B₁ = 0 indicates complete turnover). The solid line indicates the median across 100 replicate simulations each consisting of 2000 replicate regions at each scale. The bands show the interquartile range. (Online version in colour.)

Figure 4.

The strength of biodiversity effects, b_A (panel (a)) and R² (panel (b)), estimated for each year at each spatial scale in the Cedar Creek simulations. Values of b_A were estimated at each spatial scale by drawing 5000 replicate combinations of local plots and estimating the slope of the relationship between their aggregated biomass and their unique species in log–log space. Local scale plots were 81 m² and the max spatial scale corresponds to 24 plots aggregated together. The dark grey shows the results when the species identities from the experimental plots were used to estimate γ-diversity, the light grey shows the results when all plots were assumed to consist of unique species. The solid line indicates the mean across all 14 years in the dataset, the bands show the interquartile range.