The niche concept is fundamental to ecology. The idea of the ecological niche has been central to most theory explaining how biotic and abiotic factors affect the abundance and distribution of species, and how species utilize, modify and compete for resources in the environment. However, the ecological niche has been defined multiple times through the history of ecology (Leibold, 1995; Chase and Leibold, 2003). Grinnell (1917) defined a species' niche in terms of the match between the species' traits and its habitat, whereas Elton (1927) considered the niche to be a qualitative description of the resource requirements for species persistence, as well as the impacts of species on their environment. By contrast, the Hutchinsonian niche (Hutchinson, 1957) represents a more quantitative description of all biotic resources and abiotic conditions (i.e., multiple n-dimensions) that affect the fitness of a species. More recently, trait-based approaches have merged the Hutchinson n-dimensional niche perspective with the Eltonian niche to quantify a species' niche based on multiple functional traits. The main premise of trait-based ecology is that the niche can be characterized by the phenotypic space occupied by a group of organisms. Quantification of this functional trait space has provided insights into understanding the ecological and evolutionary processes that structure functional diversity, and the relationship between functional diversity and ecosystem processes (Carmona et al., 2016).

An alternative to trait based or Hutchinsonian niches are niches based on the chemical composition of organisms. As energy and material acquisition, storage and exchange are essential to life, and because the chemical composition of an organism has been shown to shape its ecology, it has been argued that biochemistry represents a key aspect of a species' niche (Sterner and Elser, 2002; Elser et al., 2003; Newsome et al., 2007; Carnicer et al., 2015). The chemical composition of an organism (isotopic, elemental, and molecular) is determined by the resources it consumed and the environmental conditions it experienced, as constrained by its nutritional and metabolic requirements. Importantly, these nutritional and metabolic requirements can often be expressed in only a few well-established dimensions, potentially resolving some of the ambiguity in selecting traits. In particular, isotopic ecology (Layman et al., 2007; Newsome et al., 2007; Jackson et al., 2011; Cucherousset and Villéger, 2015; Swanson et al., 2015), the geometric framework of nutrition (Simpson and Raubenheimer, 1993; Raubenheimer and Simpson, 2004; Raubenheimer et al., 2009, 2015; Kearney et al., 2010; Raubenheimer, 2011), and ecological stoichiometry (Elser et al., 2000; Peñuelas et al., 2008, 2010) have all represented the ecological niche using the chemistry of living organisms.

A number of metrics originally developed by functional ecologists to quantify trait functional diversity have been repurposed and applied in stable isotope ecology research (Rigolet et al., 2015; Comte et al., 2016). Like traditional traits, isotopic traits can be fit with a convex hull, the size of which has been extensively used as a proxy of the size of the trophic niche occupied by a group of individuals (Layman et al., 2007, 2012; Cucherousset and Villéger, 2015). Similarly, the geometric framework of nutrition (NGF) has been used to build multidimensional models of animal macronutrient budgets (i.e., multidimensional nutritional niche) in which information on food macronutrient contents (i.e., proteins, carbohydrates, lipids), animal macronutrient requirements, and animal nutritional processes such as macronutrient intake, growth and macronutrient use, are integrated, modeled and visualized as three macronutrient axes (Raubenheimer et al., 2015; Machovsky-Capuska et al., 2016. Although this approach has been extremely useful to improve our understanding of the foraging behavior, post-ingestion allocation of macronutrients, and the dietary niche breath of a large diversity of animals, it is a data hungry approach and some of the required nutritional descriptors are restricted to specific taxonomic groups—such as animals but not plants. Further, it does not allow for a formal empirical or theoretical application of the laws of conservation of matter to link the chemical niche of individual organisms to ecosystem-level processes (Sperfeld et al., 2016, 2017). This imposes some limitations on the use of NGF to integrate and describe entire ecological communities, and to identify links between chemical phenotypes and ecosystem processes (Sperfeld et al., 2016). Here we propose a new way to chemically describe the niche—the multidimensional stoichiometric niche, based on ecological stoichiometry theory—which may be even more useful.

All living organisms share biochemistry based on carbon (C), nitrogen (N), phosphorus (P), and other chemical elements (Elser et al., 2000; Sterner and Elser, 2002). However, organisms differ widely in the proportion of these elements in their biomass, both across taxonomic groups and trophic groups (Elser et al., 2000; Sterner and Elser, 2002). Ecological stoichiometry (henceforth ES: Sterner and Elser, 2002) studies the balance of energy and multiple chemical elements in ecological interactions, and assumes that the elemental composition of living organisms reflects their demands for chemical elements and the degree of elemental homeostasis. The demands of chemical elements largely depend on the individual investment in structural resources such as C-rich cellulose and lignin by plants, C-and N-rich chitin and muscle by arthropods, and P-rich bones by vertebrates, but are also influenced by physiological processes such as growth and reproduction (Sterner and Elser, 2002; Elser et al., 2003; Leal et al., 2017). Similarly, the different degrees of elemental homeostasis among living organisms (e.g., plants and animals) can determine important variation in organismal stoichiometry (Sterner and Elser, 2002; Persson et al., 2010). Therefore, stoichiometric traits are defined here as the composition, assimilation, allocation, or excretion of elements at the individual level, quantities that are usually assessed by the elemental content or elemental ratios of living organisms (Leal et al., 2017). Overall, stoichiometric trait variation, just like variation in any other trait, can be scaled-up from individuals to entire communities (Violle et al., 2007).

As with other quantitative traits, elemental composition and variation can be represented in a multivariate space (Sterner and Elser, 2002). Thus, each axis can represent the quantity (e.g., proportion) of a chemical element in the body of an organism, which allows a hypervolumetric visualization and analysis of the niche and trait distribution at multiple scales (groups of individuals, entire communities or regional pools of species; see Villéger et al., 2008; Cucherousset and Villéger, 2015; Carmona et al., 2016). This elemental view of a living organism has been called the elemental phenotype, which could include the ~25 elements composing the biomass of living organisms (Jeyasingh et al., 2014; Leal et al., 2017). Each of these chemical elements play key roles in the chemistry of life ranging from biomolecules and organelle structure to its life history (Sterner and Elser, 2002), and consequently the elemental phenotype is thought to be shaped by the variation in classic traits (e.g., growth rates, body size) related to the structural and functional demands of individuals (Sterner and Elser, 2002; Hessen et al., 2013; Jeyasingh et al., 2014). Although the elemental phenotype approach goes beyond the mere quantification of the elemental composition of an organism, it relates to the niche by considering an integrated and multidimensional view of the elemental make-up of life. This multiscale approach—from individuals within a species to entire communities—allows ecologists to determine how much different levels of biological organization contribute to overall trophic and resource diversity.

The ES focus on chemical elements has significant advantages over the nutritional niche framework for analyses of ecological niches. The ES perspective provides a common currency that facilitates comparisons across diverse taxonomic groups, ecological levels of organization (individuals, populations, communities or regional pools of species), and habitats (e.g., terrestrial, freshwater), and links the elemental phenotype to community and ecosystem-level processes. The elemental phenotype view is not new; in fact plant ecologists previously described it as the “biogeochemical niche” (Peñuelas et al., 2008, 2010). The biogeochemical niche represents a species position in multivariate space as defined by elemental content (sensu Peñuelas et al., 2008). Much of the progress done by plant ecologists on this topic has been made possible because elemental traits in plants show strong responses to environmental gradients, with consequences for plant fitness. As with isotopic niches, biogeochemical niches have been quantified for individual species, based on the mean value of plant elemental traits (Peñuelas et al., 2008, 2010; Violle and Jiang, 2009), and scaled-up to describe entire plant communities (Kerkhoff and Enquist, 2006). Although the elemental composition of living organisms (plants, animals and microorganisms) is known to shape their ecology and influence the structure and functioning of ecological systems (Hall, 2009; Hawlena et al., 2012), similar analyses of the elemental niche of entire communities, including their size and overlap, remain largely unexplored. Therefore, further progress can be made by the development of an approach that allows the description and comparison of the biogeochemical niche across all taxa (e.g., bacteria, fungi, plants, and animals) and habitats (e.g., aquatic and terrestrial), as well as across observational gradients or experimental manipulation conditions.

Borrowing some ideas from the “biogeochemical niche” as defined by Peñuelas et al. (2008) and Sardans and Peñuelas (2013), and adapting metrics developed for the analysis of stable isotope data and functional trait diversity (Layman et al., 2007; Cucherousset and Villéger, 2015), we develop here the multidimensional stoichiometric niche approach for ecological stoichiometric research. Our approach expands from Peñuelas et al. (2008) who described and analyzed the “biogeochemical niche” for plants entirely based on the position of species in multivariate space using principal component analysis (PCA). We define the stoichiometric niche as the region of multivariate niche space occupied by a group of individuals where the axes represent their elemental content, but the stoichiometric niche could also be extended to incorporate nutrient recycling rates (Table 1). Incorporating element fluxes and transformation rates as additional or alternative axes describing the multidimensional stoichiometric niche, can provide a more complete picture of the relative importance of species redundancy in elemental cycling at the ecosystem level (Table 1).

We propose that the multidimensional stoichiometric niche offers a powerful and unified framework to perform analyses of elemental composition data across taxa and habitats in ecological stoichiometry. Importantly, this concept allows researchers to examine how different levels of biological organization affect the stoichiometric trait space occupied by entire food webs, how organisms differ in their stoichiometric niches based on ecological role (functional group, trophic level), and how entire food webs are affected by habitat and exposure to stressors (see section Discussion and Perspectives, Table 1). Here we convert some metrics developed earlier for stable isotope and functional trait data (mainly based on traditional traits such as growth, survival, photosynthetic rate; Violle et al., 2007) into a set of stoichiometric metrics based on the elemental content of individuals or groups of individuals within communities, which together describe the multidimensional stoichiometric niche. These metrics are as follows:

Stoichiometric niche volume: Also known as convex hull volume, this metric measures the amount of elemental space filled by a group of organisms. This space represents the stoichiometric diversity of a group of organisms within a population, functional group or whole community. For example, a small stoichiometric niche volume would indicate that individuals of a particular group (e.g., species, trophic group) exhibit low intra-group variation regarding their elemental traits.

To illustrate the applicability of the multidimensional stoichiometric niche approach, we used data on C, N, and P content of several terrestrial and freshwater communities composed by multiple trophic groups (from primary producers and detritivores to carnivores) and species. We use these data to answer Questions 1, 7, 8, and 9 in Table 1. First, we calculated the niche volumes in stoichiometric trait space occupied by the aquatic and terrestrial food webs, trophic groups, and individual morphospecies (hereafter “species”). Then we evaluated several aspects of the stoichiometric niche space (i.e., niche size, niche overlap and nestedness), which provide information of the stoichiometric complementarity within and among these groups. Finally, we identify key limitations of this approach, and discuss perspectives that may facilitate the development of quantitative intra and interspecific comparisons, and food-web approaches using stoichiometric data.

As an example of the multidimensional stoichiometric niche, we analyzed entire terrestrial and aquatic food webs. The whole data set includes 73 species (35 families and 18 orders) belonging to multiple trophic positions along three axes: C, N, and P content. The terrestrial and aquatic food webs include, respectively, 33 and 40 species, 20 and15 families, and 13 and 7 orders.

Terrestrial invertebrates and vertebrates associated with Tillandsia bromeliads were collected from four sites in the coastal zone of the Atacama Desert (Chile; 20°13′S, 70°8′W) in 2007 and 2009 (González et al., 2011a). Low mean annual rainfall characterizes this area, as the only source of moisture is associated with fog events (Cereceda et al., 2008). The increased air moisture and nutrient deposition in the fog zone permits the development of isolated vegetation “islands.” These “islands” are dominated by several terrestrial bromeliad species of the genus Tillandsia, which depend exclusively on fog inputs as their primary water source (Pinto et al., 2006), and sustain a diverse community of terrestrial invertebrates (e.g., Coleoptera, Araneae, Solifugae) and vertebrates (e.g., Squamata). At each of the four sites, we collected arthropods and lizards, which compose the terrestrial food webs associated with Tillandsia (see González et al., 2011a for sampling details).

Aquatic invertebrates were sampled from tank bromeliads from one site in each of two countries: Estacion Biologica Pitilla (Costa Rica; 10°59′N, 85°26′W), and Cardoso Island (Brazil; 25°03′S, 47°53′W) in 2012 and 2013. Costa Rica is dominated by primary and secondary tropical rain forests with sparse open clearings whereas Cardoso Island corresponds to a “resting” ecoregion (i.e., coastal vegetation located on nutrient-poor sand deposits; Magnago et al., 2012). Tank bromeliads are flowering plants that accumulate rainwater in their leaf rosettes creating an aquatic habitat—from a few milliliters up to several liters of water per plant—for a diverse community of macroinvertebrates (e.g., Diptera, Coleoptera, Haplotaxida, Ostracoda; Dézerald et al., 2013). Macroinvertebrates were sampled from tank bromeliads in each study site either by dissection of the plant or with a large-mouthed pipette.

The elemental composition of living organisms is of fundamental importance for population dynamics, consumer-resource interactions, food web structure, and nutrient cycling (i.e., effect traits) (Sterner and Elser, 2002; Hall, 2009; Yamamichi et al., 2015), and can mediate eco-evolutionary responses to environmental change (i.e., response traits) (Jeyasingh et al., 2014; Leal et al., 2017). Stoichiometric traits, have been measured for a large number of very different species: primary producers (Reich and Oleksyn, 2004; Peñuelas et al., 2008, 2010; Reich, 2014), phytoplankton (Klausmeier et al., 2004; Litchman and Klausmeier, 2008; Quigg et al., 2011), microorganisms (Mouginot et al., 2014; Godwin and Cotner, 2015), invertebrates (Fagan et al., 2002; Woods et al., 2004; Hambäck et al., 2009; González et al., 2011a; Wiesenborn, 2011, 2013; Lemoine et al., 2014), and vertebrates (Torres and Vanni, 2007; González et al., 2011a), allowing across-species comparisons. For plants, studies on functional traits have revealed the existence of adaptive trait continua, which describes the phenotypic space of trait variation produced by evolutionary processes (Donovan et al., 2011; Carnicer et al., 2015; Diaz et al., 2016). The observed inter-specific variation in stoichiometric traits (i.e., C, N, and P contents) along these trait continua has shown strong trait co-variation between N and P (Carnicer et al., 2015). However, lacking in this framework are studies of whole communities focused on the relationships between major axes of variation in stoichiometric traits.

The multidimensional stoichiometric niche approach proposed here extends Hutchinson's (1957) concept of niches as n-dimensional hypervolumes in order to describe and compare the stoichiometry of groups of organisms at different scales using data from observational or experimental approaches. Overall, stoichiometric traits provide a common currency to estimate the dimensionality of trophic niche space, and help reduce the number of axis required to characterize community structures. All organisms interacting in an ecological community are composed by the same chemical elements, and these common currencies enable us to estimate trait similarities or differences across taxonomic groups. Several lines of evidence support the idea that differences in the elemental composition of living organisms reflect the evolution of internal and external structures (e.g., muscular, skeletal), as well as differences in organism life histories (e.g., Reiners, 1986; Sterner and Elser, 2002), and these stoichiometric differences exert strong influences on individual fitness, ecological interactions and ecosystem functioning (Hall, 2009; Leal et al., 2017). For example, the fitness of consumers can be directly related to the mismatch between their body's own elemental composition and that of their resources (Sterner and Elser, 2002). This mismatch can be elegantly encapsulated as the distance and consequent low overlap between the stoichiometric niches of consumers and resources (Q5 in Table 1). The multidimensional stoichiometric niche also provides a means of defining the elemental composition of an entire food web in more than one dimension, thereby capturing both the independent and interactive effects of nutrients on living organisms. This allows us to compare food webs in different habitats, locations, or points in time, in terms of their stoichiometric niches, specifically niche position, size and overlap (Q3, 4, 7, in Table 1). For example, Cross et al. (2007) reported the effects of experimental nutrient enrichment on biomass production and pathways of C, N, and P in a detritus-based stream food web. The multidimensional stoichiometric niche approach developed here could be used to re-analyze these and other data from experimental manipulations and evaluate the extent at which a particular or multiple stressors impact the stoichiometric niche (i.e., niche size, shape and position) of the food web in treated and reference systems. Importantly, this approach provides a novel way to quantitatively assess changes an ecosystem has experienced as a function of global change drivers, and to link such changes to shifts in the biogeochemical processes carried out by organisms (Q2 and Q6 in Table 1).

The idea of a multivariate description of the relative elemental content of organisms has been previously suggested and applied in plant ecology (Peñuelas et al., 2008, 2010), providing insights into plant trait differentiation and climate change responses. Here, we expand on those original ideas to describe the stoichiometric niche of entire communities from contrasting ecosystem types (i.e., aquatic vs. terrestrial) and trophic groups. Further, our approach includes novel ways to estimate the stoichiometric niche by borrowing and adapting metrics developed by isotopic ecologists to measure trophic niche (Layman et al., 2007; Jackson et al., 2011; Cucherousset and Villéger, 2015). These metrics have enabled us to compare the size, shape and position of the stoichiometric niche at different ecological levels, from a species niche to communities. We propose that this direct quantification of a multidimensional niche across ecological scales is possible from a set of elemental measurements of the individuals composing ecological communities, regardless of their taxonomic identity. Further, as stoichiometric differences within and between species are widely recognized, this approach enables us to quantify the stoichiometric niche of a single species up to that of entire communities.

Although the primary purpose of the case study was to demonstrate the use of the multidimensional stoichiometric niche approach, it also provides some insights into the concept of stoichiometric niche as a tool to describe and compare the stoichiometry of organisms at different scales. The questions we have addressed with this approach include those at the scales of individual morphospecies (intraspecific variation), trophic groups and food-web types, as well as questions that cross scales. Our analyses showed that species-level stoichiometric niches are relatively small and do not overlap much with each other. This suggests that species within terrestrial and aquatic systems exhibit large complementary in their stoichiometric niches. In addition, our analyses at the intraspecific level revealed similar patterns of small niche sizes, low overlap, and different niche shapes among geographically distinct populations of a predatory species (Ammotrechidae sp.). The observed variability in the elemental composition of those populations could be related to the fog nutrient supply that broadly varies across the coastal desert sites, thus affecting primary producers in those areas (González et al., 2011b), and potentially influencing the elemental composition of consumer populations such as Ammotrechidae sp. Finally, the niche size of Ammotrechidae sp. was significantly smaller than expected by chance, suggesting that strong deterministic factors may constrain the elemental composition of those organisms (including biological, environmental, and phylogenetic factors; Elser et al., 2000).

Our analyses also revealed that, although the terrestrial and aquatic food webs differ in both habitat and geographic location, the two food webs shared more than one third of their total niche volume. Thus, as with other quantitative traits, there are both gaps and crowded regions in the elemental space. The gaps suggest that there are fundamental constraints to stoichiometric evolution (Kay et al., 2005; Jeyasingh, 2007); macroscopic life (or at least that subset we sampled) may have only evolved a restricted set of morphological, physiological and cellular strategies for exploiting elemental resources. By contrast, the packed regions of phenotypic space suggest that certain trait combinations may have evolved either repeatedly or deep within phylogenetic trees (Pigliucci, 2007; Donovan et al., 2011). Interestingly, our findings are in agreement with recent worldwide studies on vascular plants, in which the trait hypervolume occupied by these plants is highly constrained within a relatively small plane, representing a small set of successful trait combinations (Diaz et al., 2016). We also found that the size of the terrestrial and aquatic niche volumes are driven by contrasting stoichiometric traits, this is, by covariance in N and P for terrestrial food webs and opposing effects of C and N for aquatic food webs. Although previous analyses have shown differences in elemental compositions between aquatic and terrestrial primary producers (Elser et al., 2000; Cebrian and Lartigue, 2004), our results are remarkable in that aquatic invertebrate food webs in bromeliads are largely fueled by allochthonous inputs of terrestrial leaf litter (Farjalla et al., 2016). Therefore, despite both depending on terrestrial plants, aquatic and terrestrial invertebrates in bromeliads still show large differences in their stoichiometric niches. However, we caution that the terrestrial and aquatic food webs we examined also differ in terms of biome (desert vs. forest), ontogenetic stage (mostly larvae for aquatic and mostly adults for terrestrial) and geographic location (Pacific vs. Atlantic side of the Americas); we hope that the method we demonstrate here will inspire more detailed studies that compare terrestrial and aquatic food webs in the same location. Our analyses also showed that terrestrial herbivores represent a unique group in terms of their stoichiometric niche, distinct from animals that consume detritus or other animals. The differences in stoichiometric niche between herbivores and detritivores are curious, given that both groups consume vascular plants, but may reflect the importance of detrital-based bacteria and fungi in the diet of detritivores (terrestrial and aquatic), plus opportunistic consumption of dead or live animals by detritivores. Potential animal matter consumption by detritivores may explain the large stoichiometric overlap between them and predators.

Our findings also demonstrate that subsampling abundant species may substantially underestimate the actual (population-level) stoichiometric niche occupied by such species in food webs. We thus strongly recommend integrating the natural interspecific variability in abundances rather than trying to standardize the number of individuals per species, as the latter could lead to erroneous estimates of population-level niches and thus misleading estimates of species contributions to the niche of the entire community. This suggestion is contingent on researchers sampling species proportionally to their natural abundance. However, once species differences in stoichiometric niches are established, the role of intraspecific variation in determining niche differences could then be assessed by rarefying samples.

In our case study, we focused on the three elements (C, N, P) that have consistently been shown to constrain species interactions and limit production in ecological communities (Sterner and Elser, 2002). However, if other elements (e.g., Ca, K, S, Fe, Mg) or other physiological (e.g., growth and reproductive rates) and behavioral (e.g., microhabitat preferences) traits are found to further define particular ecological communities, then other dimensions would need to be considered. Hypervolumes that appear to overlap in low dimensions (e.g., C, N, P) may not overlap if more dimensions are included, and conversely, with the addition of extra redundant dimensions, estimates of overlap may be falsely inflated. We could thus expand our current method and include more elements by calculating our metrics in the PCA space. With this approach we can reduce any set of n elements to only three principal components, which usually explain most of the variance in the data (i.e., the first three PCA axes) and would allow the calculation of our metrics. The resulting metrics could, then, be compared across studies and ecosystem types if the same set of elements is used to construct the reduced PCA space. Therefore, our focus on C, N and P dimensions provides a useful starting point for broad comparisons.

We think these insights show only a fraction of the potential of possible applications of the stoichiometric niche in ecological stoichiometric research, and that the multidimensional stoichiometric niche can be a general approach for assessing and comparing stoichiometric variation within and among species, populations and communities (Table 1). There are numerous potential applications of stoichiometric data to understand trophic structures especially when used in combination with other trophic-related functional traits, such as body size or with community-level properties such as species numerical abundances. This may offer a powerful and unified framework to assess niche partitioning and food web structures across contrasting ecosystem types and broad biogeographic scales. Such approaches could also be applied to explore individual specialization (Bolnick et al., 2002; Araujo et al., 2011), and to assess how populations and different genotypes vary in their fitness within a multidimensional stoichiometric space (Leal et al., 2017).

Recent research in ecological stoichiometry has highlighted the large intraspecific variation in elemental composition of plants and animals (e.g., González et al., 2011a; El-Sabaawi et al., 2012; Borer et al., 2013; Ebel et al., 2015). Within-species variation in stoichiometric traits can be driven by differences in morphological, physiological, behavioral, and life history traits, as well as predation pressure and spatial-temporal environmental heterogeneity in the quality of resources (Jeyasingh et al., 2014; Leal et al., 2017). The magnitude and origin of stoichiometric trait variation (i.e., phenotypic plasticity or genetic) within populations can be used to link the fitness of any population to changes in resource quality (Leal et al., 2017). For example, if directional selection is acting on stoichiometric traits to reduce elemental mismatches between a population and its resources, we should expect a high overlap between the stoichiometric niches of a consumer and its resources. In contrast, if fluctuating selection is acting upon stoichiometric traits along spatial or temporal variation in resource quality, then the stoichiometric niches of a population should reflect this variation by displaying a larger niche volume. Following Leal et al. (2017), we envision the assessment of how different genotypes vary in their fitness within a multidimensional space, in which stoichiometric trait axes determine the functional space.

Another exciting area of research in niche ecology is the analysis and mapping of stoichiometric niche-environment (i.e., elemental availability of resources) relationships, which can help understand and predict how well the stoichiometric niches of individuals are predicted by environmental gradients. The integration of stoichiometric models and outcomes from the multidimensional stoichiometric niche could expand and strengthen predictions derived from traditional approaches linking stoichiometry traits to fitness consequences to trophic interactions. Further, as nutrient dynamics are tightly linked to the stoichiometry of living organisms, analyses of niche overlap in the multidimensional space can provide insights into species functional redundancy or complementarity in ecosystem function (e.g., nutrient recycling rates). The stoichiometric niche is therefore a concept with the potential to unify physiological, ecological and ecosystem approaches to understanding the biogeochemical role of life.

Procedures involving animals were approved by the ethics committee of the Pontificia Universidad Católica de Chile and by the Chilean Agriculture and Livestock Service (SAG).

AG developed the concept, designed of the manuscript, and wrote the manuscript. OD contributed on the development of the ideas presented here, analyzed the data, and helped writing the manuscript. PM, GR, and DS contributed to the writing of the manuscript. All authors made substantial contributions to the development of the ideas presented here and commented critically on drafts of the manuscript. This is a contribution of the Bromeliad Working Group.