Portal de Periódicos da CAPES

Implications for biodiversity conservation of the lack of consensus regarding the humped‐back model of species richness and biomass production

2013; Wiley; Volume: 28; Issue: 1 Linguagem: Inglês

10.1111/1365-2435.12147

1365-2435

Simon Pierce,

Species Distribution and Climate Change

Lay Summary Podcast The humped-back model of species richness and biomass production (Grime 1973, 2001; Al-Mufti et al. 1977) predicts that taxonomic richness may be greatest at intermediate biomass production and at intermediate intensities of factors that limit production, such as disturbance or stress. This model has, over the last four decades, become widely supported as observational evidence has accumulated from a range of plant, animal and microbial communities in terrestrial and aquatic habitats world-wide (reviewed by Grime & Pierce 2012). The humped-back curve is not always evident (Mackey & Currie 2001; Mittelbach et al. 2001; Hughes et al. 2007), but several authors (Moore & Keddy 1989; Guo & Berry 1998; Keddy 2005) have noted that this is symptomatic of studies that investigate a restricted biomass range, such as within communities rather than across highly diverse habitat types, and these studies potentially bias meta-analyses. Furthermore, observations from natural habitats are sometimes specifically excluded from meta-analyses in favour of culture experiments (Cardinale et al. 2011) which, it has been argued (Grime & Pierce 2012), defeats the aim of understanding phenomena occurring in nature. In order to provide a large-scale observation of the productivity–richness relationship in nature, Adler et al. (2011) measured species richness and biomass production in situ for herbaceous vegetation across five continents. They concluded that productivity–richness relationships are 'weak and variable', at both local and global scales. Adler et al. (2011) plotted their entire global data set as a single scatterplot, with regressions then fitted to within-site data (this is reproduced in Fig. 1a). When I produced the same scatterplot of species richness and biomass from the data (available as an appendix to the original publication), a cloud of data points was evident that, as Adler et al. (2011) state, was not well fitted by a humped-shape curve (Fig. 1b). However, the humped-back model does not state that the productivity–richness relationship is a simple curve. Rather, it describes an upper limit to the potential diversities that may develop along the biomass gradient (Grime 2001; Grime & Pierce 2012), or in other words, a boundary that separates productivity–richness relationships that are possible (below the curve) from those that are not (above). This is something of which Adler et al. (2011) were aware: 'an alternative hypothesis states that productivity sets the upper limit on richness, with stochastic forces such as disturbance causing deviations below this limit'. Grace (2001), one of the co-authors of Adler et al. (2011), had previously noted 'weak regression relationships between biomass and richness despite strong constraints on boundary relationships' which he described as a 'unimodal envelope'. Indeed, a humped-back relationship may be lacking when measurements are taken in years that differ from the typical local climate, but may reappear when conditions return to those more typical and favourable for the establishment and growth of local ecotypes (Laughlin & Moore 2009). Grime (2001) also noted that 'where vegetation is subject to such vicissitudes, it is not to be expected that the maximum in biomass and litter measured in any one year will be predictably related to species richness'. Additionally, attainment of the maximum potential species richness depends on the size of the local species pool (Grime 2001). For example, Foster (2001) found that experimental addition of seed of 34 species to a prairie increased species richness markedly at lower productivities, but much less so at higher productivities. He concluded that potential species richness is constrained by competition at higher productivities but is sensitive to the size of the local species pool towards lower productivities. Houseman & Gross (2006) similarly augmented the local species pool available to a mid-successional old-field site that included a range of microsites of varying productivity. The productivity–richness relationship was negative and linear when the species pool was restricted, but a larger species pool allowed richness to increase only at intermediate productivities, resulting in a humped-back curve. They termed the curve 'saturated' to denote its status as an upper potential limit to richness. Olde Venterink et al. (2001) had previously labelled the curve as 'filled' for similar reasons, reflecting Grace's (2001) 'unimodal envelope'. Similarly, Grime (2001) suggests that the particular species richness of each plant community represents 'an equilibrium between the habitat condition prevailing at each site and the reservoir of species in its vicinity' and presents an example of a filled humped-back curve measured from herbaceous plants in environments experiencing fluctuating patterns of vegetation management (Fig. 91, p. 266 of Grime 2001). Thus, the existence of communities of lower species richness that fall underneath the curve is an effect of environmental instability and limited local species pools. I hypothesized that a humped-back curve would be evident with analysis of only the upper limit of the Adler et al. (2011) data. Following the example of Adler et al. (2011)'s Fig. 2, I also started by plotting the entire data set. However, whereas Adler et al. (2011) used quantile regression to investigate the hypothesis of an upper envelope in their analyses, I used upper boundary regression (also known as factor-ceiling analysis or upper boundary analysis), an established method to fit a regression to the upper boundary of the data set, which is extensively used in ecological studies (Blackburn, Lawton & Perry 1992; Pakeman et al. 1999; Krause-Jensen et al. 2000; Lessin, Dyer & Goldberg 2001; Liira et al. 2002; Peat, Clarke & Convey 2007; Clarke & Rothery 2008; Henriksen 2009; Price & Whalen 2009; Werner et al. 2009; Bettridge, Lehmann & Dunbar 2010; Gogina, Glockzin & Zettler 2010; Weiss, Kiefer & Kipper 2012). This technique divides the continuous data on the x-axis into classes of equal range (known as bins), and a regression is fitted to the highest y-values within each bin. I took into consideration the highest species richness values within each 100 g m−2 interval of the biomass range (i.e. within bins of 0–99, 100–199, 200–299 g m−2, etc.), in order to examine only the greatest species richness values at each point along the biomass gradient and thus the upper limit of measured biodiversity. The 20 highest values within each bin were used in order to reduce any possible effects of outliers (Adler et al. (2011) removed outliers prior to analysis). The curve that most closely fitted the upper boundary of the Adler et al. (2011) data set using upper boundary regression was a unimodel (humped-shape) curve: a close fit (R2 = 0·810) that was highly significant (anova: F = 369·4, P < 0·0001; Fig. 1c). I found that fitting a Gaussian curve using upper boundary regression also resulted in a highly significant unimodal relationship (R2 = 0·754, anova: F = 264·5, P < 0·0001; not shown), although a Lorentzian curve was found to be the curve of best fit. Adler et al. (2011) also investigated the relationship between mean biomass and mean species richness at each site (Fig. 2a), as measurements from the same site are not necessarily independent. When I also analysed site means using upper boundary regression (Fig. 2b), a highly significant unimodal curve was evident (R2 = 0·981, anova: F = 102·9, P = 0·0004). Again, a Gaussian curve also fitted the data well and highly significantly (R2 = 0·935, anova: F = 28·6, P = 0·0043; not shown). This analysis indicates that the humped-back model, as a unimodal envelope or upper boundary to potential species richness along the productivity gradient at the global scale, is extremely well supported by the Adler et al. (2011) data set. The implications of this statement are profound. Because the data set represents a global observation across a range of habitat types, it is more comprehensive and therefore more important than the local, within-habitat observations or artificial experimental cultures that potentially colour the interpretation of meta-analyses (Mackey & Currie 2001; Mittelbach et al. 2001; Hughes et al. 2007; Cardinale et al. 2011). Fox (2013) points out that if a humped diversity curve only emerges in special circumstances (e.g. as a maximum potential curve), then it is only one of a range of possible outcomes and is therefore not very informative. However, as the preservation of high species richness is a desired outcome (out of a range of possible outcomes) of biodiversity conservation, understanding what these special circumstances are and in which situations they can, or cannot, occur is fundamental to successful conservation. The preservation of species richness maintains ecosystem stability and function and thus ecosystem services (Isbell et al. 2011). Previously, Mittelbach et al. (2001) had already determined that productivity–richness relationships are highly variable at the local scale, so this particular finding of Adler et al. (2011) is not novel or informative. The question of interest is: what species richness pattern is evident across broad ranges of biomass production, particularly at the global scale? Crucially, Adler et al. (2011) did actually produce a statistically significant unimodal (humped) regression at the global scale, although this result was downplayed in their discussion: 'At the global extent, the quadratic effect of productivity on richness [i.e. a unimodal curve] was significant (t = –2·39, P = 0·021)', and it was only 'when we removed nine sites of anthropogenic origin and the one salt marsh, the quadratic effect was no longer significant (t = –1·36, P = 0·18)'. In other words, they found a global-scale humped-back curve, but by then excluding certain habitats from the final analysis, the curve was no longer apparent. This part of their analysis is reproduced here in Fig. 2a. Fridley et al. (2012) made a specific point of criticizing the decision to exclude sites, to which Grace et al. (2012) replied simply that they were 'anticipating that some might object to inclusion of highly altered sites'. However, the fact that the relationship shifted from a significant unimodal regression to nonsignificance when these sites were excluded (Fig. 2a) demonstrates how crucial the sites were to the relationship, and their omission should be extremely well justified. Neither Adler et al. (2011) nor Grace et al. (2012) attempt to provide an explanation as to why the salt marsh in particular was removed. Nor do they explain the criteria used to distinguish 'highly altered' or 'anthropogenic' sites, how habitats can be categorized as entirely natural or not (rather than falling at some point along a spectrum of intensity of human impact), nor why they expect grassland grazed by domestic herbivores to exhibit a different productivity–diversity relationship than grassland grazed by wild animals. The excluded habitats are amongst the most crucial that exist in the context of the productivity–richness relationship. Salt marshes and other wetland herbaceous communities are characterized by higher productivities and small numbers of species (Grace et al. 2007), helping to delimit the right-hand side of the humped-back curve. At intermediate productivities, 'anthropogenic' habitats may be amongst the most species rich and help define the middle and hump of the curve: one square metre of meadow or pasture may include up to 89 flowering plant species, and by definition, these habitats are disturbed, respectively, by human-imposed mowing and grazing (Wilson et al. 2012). Furthermore, two of the sites excluded by Adler et al. (2011) were European (a pasture at 995 m a.s.l. in Switzerland and also the most productive site in the study, and an old-field site at sea-level in Germany), leaving only two European sites in the analysis. Europe has experienced intensive farming over many millennia and exhibits a range of semi-natural (anthropogenic) habitats such as montane pasture and lowland old fields and is thus particularly relevant to global studies of biodiversity, conservation and habitat restoration. Many of the habitats protected by the European Union Habitats Directive are threatened by the abandonment of traditional management regimes and the advancement of woodland, and many are mountain pastures or meadows [e.g. mountain hay meadows (habitat number 6520), species-rich Nardus grasslands on silicious substrates in mountain areas (6230*), semi-natural dry grasslands and scrubland facies on calcareous substrates (Festuco-Brometalia) (6210*)]. The specific association of moderate to high biodiversities with anthropogenic habitats such as these is thus a fact of modern biodiversity conservation. Habitat conservation requires a fine degree of understanding of the links between productivity, local species pools and biodiversity – understanding that will not be attained by expunging these habitats from the record solely because they are perceived to be in some undefined way unnatural or ignoble. Halting mowing or grazing coupled with fertilizer addition consistently allows taller species to dominate, as large species and the litter they produce limit the establishment of smaller species (Wheeler & Giller 1982; Wheeler & Shaw 1991; Kirkham, Mountford & Wilkins 1996) and effectively isolate the local patch from the species pool (Stevens et al. 2004). The resulting plant communities are not as rich as the meadows and pastures they succeed. Atmospheric inputs of excess nitrogen can also drive dominance and reduce species richness. For example, the grass Brachypodium pinnatum has increased in frequency in European dry calcareous grasslands due to the direct effects of atmospheric nitrogen deposition, to the detriment of species richness (reviewed by Bobbink, Hornung & Roelofs 1998). Bobbink, Hornung & Roelofs (1998) also reviewed the effects of nitrogen deposition on a range of plant communities, and their main conclusion can be readily interpreted in terms of the humped-back model: 'the availability of nitrogen in ecosystems has gradually increased, leading to competitive exclusion of characteristic species by more nitrophilic plants, especially under oligo- to mesotrophic soil conditions. On very nutrient-poor soils, diversity has sometimes increased as species that were not tolerant of the original conditions have been able to invade, but the native flora has still disappeared'. Thus, nitrogen addition to mid- and high-productivity sites reduces species richness via competitive effects (the right-hand flank of the curve), whereas nitrogen addition at very low-productivity sites may increase species richness as constraints to resource availability are relaxed (the left-hand flank). The highest plant species richness does not occur at the highest productivity [this is clearly evident from the Adler et al. (2011) data set] and productivity must be zero when there are zero species, meaning that the diversity–productivity relationship inevitably passes through the origin. Species-rich habitats can only exist at an intermediate point between these extremes of low richness [indeed, communities exhibiting species richness of over 40 species m−2 were evident only at intermediate biomasses of 400–500 g m−2 in the Adler et al. (2011) data set; Fig. 1a]. This agrees with the assertion that high species richness in herbaceous communities is associated with unfertilised soils and disturbances such as grazing, mowing or burning (Grime 2001; Grime & Pierce 2012; Wilson et al. 2012). In sum, the humped-back model is strongly supported by the present analysis of Adler et al. (2011)'s large and authoritative data set. Indeed, the data set itself is a tour de force of international collaboration and individual toil. The analysis, however, does not do the data set the justice it deserves and supports both a conclusion that is untenable and a worldview in which human environmental impacts on biodiversity can be ignored. However, factors such as land-use change, climatic change and nitrogen deposition are the most prominent drivers of global biodiversity loss (Sala et al. 2000), and so an analysis in which anthropogenic situations are deliberately disregarded is both naïve and potentially dangerously counterproductive. The Convention on Biological Diversity has largely failed to halt biodiversity losses (Butchart et al. 2010), and recognition and understanding of the interactions between humans and ecosystems are essential if this goal is to be realized before the new 2020 deadline. If Adler et al. (2011)'s conclusions were to be believed, and the humped-back model abandoned, ecologists would effectively be endorsing the message that management regimes and human impacts are not important for habitat conservation. This would undoubtedly cause a massive loss of credibility on the part of people tasked with managing wildlife and ultimately the dislocation of the science of Ecology from the practice of biodiversity conservation. Three reviewers, James B. Grace (US Geological Survey, USA), Gary G. Mittelbach (Michigan State University, USA) and Simon M. Smart (Centre for Ecology and Hydrology, Lancaster, UK), provided contrasting points of view that all helped improve the manuscript during the revision process. I thank Bruno Cerabolini (University of Insubria, Varese) and Ilda Vagge (University of Milan) for comments on the manuscript. I would like to thank Mauro Villa, Roberta M. Ceriani and the staff of Parco Monte Barro, Lecco, Italy, and of the Parco Delle Orobie Bergamasche, Bergamo, Italy, and the smallholders of Pescate, Lecco, for allowing me to gain personal experience of orchid-rich meadow habitats within the parks and the management of these habitats over the last decade. I hereby declare no commercial conflict of interest. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.