Richness – heterogeneity relationships: positive, unimodal, both, or neither?

Submitted by editor on 3 June 2014.
A Black-throated Green Warbler, one of the bird species found in the study area.


Avi Bar-Massada and Eric M. Wood

One of the most common notions in ecology is that species richness has a positive relationship with habitat heterogeneity. This idea, naturally, stems from the niche-based view of ecological communities. Heterogeneous habitats are expected to consist of more niches than homogeneous ones, and consequently a larger number of species may be able to establish in those habitats. Despite the straightforward logic behind this notion, empirical studies have found not only positive richness-heterogeneity relationships, but also negative and unimodal ones, as well as no significant relationships at all.

A recent study (Allouche et al. 2012) suggested that the fundamental type of the richness-heterogeneity relationship should in fact be unimodal. The reasoning behind this theory, termed the area-heterogeneity trade-off, is that increasing habitat heterogeneity in finite space results in an area comprising increasingly smaller sub-habitats of different types. Eventually, these sub-habitats become too small to support persistent populations of different species, and subsequently the rate of local stochastic extinction increases, and overall species richness decreases. Therefore, the unimodal richness-heterogeneity relationship is an outcome of environmental filtering at low heterogeneity levels (which prevents many species from establishing in the first place), and increased stochastic extinctions at high heterogeneity levels (which prevents small populations from persisting).

As expected given that it challenged a cornerstone of ecological theory, the area-heterogeneity trade-off has received some criticism (see comments to Allouche et al. (2012) in PNAS). While we find the new theory to be robust, we also think that the current empirical support for it is somewhat incomplete as it was based on a rather simplistic measure of heterogeneity, namely elevation range across a broad spatial extent. While bird species (the study organisms of the original paper) are likely to respond to elevation range across broad biogeographic scales, we were curious to see if the theory will hold true at much finer spatial scales, in which individuals respond to their habitats according to the presence (among others) of nesting sites, forage, and cover from predators. Furthermore, we wondered whether different heterogeneity measures may have different relationships with species richness. Finally, we wanted to test if the richness-heterogeneity relationship differs among habitat types, compared to its type at the landscape scale (where the landscape comprises multiple habitats).

To answer these questions, we analyzed an extensive dataset of bird species that was collected between 2007 and 2009 in Ft. McCoy, Wisconsin, USA. The landscape consists of three major habitat types: grasslands, black oak (Quercus velutina) savannas, and woodlands, and as such was an ideal study site for our purpose. The dataset included a wealth of information on species and habitat characteristics in 254 sample plots, and it enabled us to quantify in great detail the heterogeneity in each sample plot. We used two measures of habitat heterogeneity in each plot: foliage height diversity (a measure of vertical habitat heterogeneity), and cover type diversity (a measure of horizontal habitat heterogeneity). Based on these two metrics, we quantified the richness-heterogeneity relationships at the landscape scale and for each habitat type separately.

An oak savanna in Ft. McCoy, Wisconsin, USA.

When we analyzed the relationships, we were surprised to find that the choice of heterogeneity metric dictated the type of the richness-heterogeneity curve. At the landscape scale, species richness had a unimodal relationship with foliage height diversity. In contrast, it had a positive (and linear) relationship with cover type diversity. Moreover, when we analyzed the relationships within single habitat types, we found that the unimodal relationship between richness and foliage height diversity was mostly gone (except for in savannas), and replaced by a positive linear relationship in grasslands, and a negative linear relationship in woodlands. The positive relationship between richness and cover type diversity, in contrast, was perfectly retained in grasslands and woodlands (but not in savannas, where we found no significant relationship). These results suggest that the hierarchical level of the analysis, i.e., within habitats or at the landscape scale, may have profound implications on the type of richness-heterogeneity to be found.

Sampling bird species in the grasslands of Ft. McCoy, WI.

An intriguing question that arises from these results is why did these two heterogeneity metrics have such different relationships with species richness? A clue about the possible answer is the rather large difference in the range of heterogeneity values between the metrics. Foliage height diversity (which had a unimodal relationship with richness at the landscape scale) had a much longer range of heterogeneity values compared to cover type diversity (which exhibited a positive relationship with richness). Could it be that cover type diversity simply did not have a sufficiently long range of values, so that the relative cover of individual sub-habitat types was never too small to support viable populations of bird species? We think that this is a plausible answer. Yet it raises another interesting question about the nature of richness-heterogeneity studies in general: is it possible that many studies that did not find unimodal relationships simply did not cover a sufficient range of heterogeneity values?

The results of our study suggest that our ability to identify a fundamental ecological relationship such as the richness-heterogeneity heavily depends on our methodological approach, in this case the choice of heterogeneity metric, coupled with the characteristics of the habitats in the study area (as well as the niche characteristics of the species of interest, according to Allouche et al. (2012)). To mitigate this problem, we suggest that studies of the richness-heterogeneity relationship use multiple measures of heterogeneity, ideally corresponding with non-correlated habitat characteristics (in our case, vertical and horizontal habitat). Furthermore, the relative prevalence of positive richness-heterogeneity relationships reported in the literature may be an outcome of the relative short range of heterogeneity gradients in the study systems. This is especially true in studies conducted in single habitat types (and possibly finer spatial scales), where the range of heterogeneity values may be simply too short to allow for the area limiting factor from the area-heterogeneity trade-off to “kick in”.

The introduction of the area-heterogeneity trade-off provides us with a wealth of exciting opportunities to shed new light on a well-worn ecological question, and, more importantly, a possibly better understanding of a fundamental aspect of community ecology. Possible future research directions include meta-analysis of existing richness-heterogeneity studies in an attempt to find first principles that drive the type of these relationships, as well as field, theoretical and modeling studies that attempt to find additional processes that may affect the characteristics of this fundamental ecological relationship across different scales, habitats, and taxa.


Allouche, O. et al. 2012. Area–heterogeneity tradeoff and the diversity of ecological communities. Proc. Natl Acad. Sci. 109: 17485–17500.