SEARCH

SEARCH BY CITATION

Keywords:

  • Asynchrony;
  • energy channels;
  • food webs;
  • interaction strength;
  • metabolism;
  • resource compartments;
  • size;
  • space

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

Ecologists have long searched for structures and processes that impart stability in nature. In particular, food web ecology has held promise in tackling this issue. Empirical patterns in food webs have consistently shown that the distributions of species and interactions in nature are more likely to be stable than randomly constructed systems with the same number of species and interactions. Food web ecology still faces two fundamental challenges, however. First, the quantity and quality of food web data required to document both the species richness and the interaction strengths among all species within food webs is largely prohibitive. Second, where food webs have been well documented, spatial and temporal variation in food web structure has been ignored. Conversely, research that has addressed spatial and temporal variation in ecosystems has generally ignored the full complexity of food web architecture. Here, we incorporate empirical patterns, largely from macroecology and behavioural ecology, into a spatially implicit food web structure to construct a simple landscape theory of food web architecture. Such an approach both captures important architectural features of food webs and allows for an exploration of food web structure across a range of spatial scales. Finally, we demonstrated that food webs are hierarchically organized along the spatial and temporal niche axes of species and their utilization of food resources in ways that stabilize ecosystems.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

Food webs have been a central organizing concept in ecology for the better part of the last century (Elton 1927). Not only do they provide an appealing visualization of the feeding linkages among species within a community, but they also provide a natural structure within which various subdisciplines of ecology can operate. Areas as disparate as foraging behaviour and biogeochemical cycling have been incorporated into the food web framework, and this is perhaps its greatest strength. Yet as intuitively appealing (and promising) as food webs may be, there exist challenges in quantitatively documenting food webs in nature and harnessing the food web framework to generate meaningful predictions regarding the structures and processes that impart stability in nature.

A primary challenge lies in the inherent topological complexity of food webs, which are often comprised of scores of species and an overwhelming number of interactions. As enumerating all present species and quantifying all interactions within an ecosystem is an unfeasible task, the data for most food webs are incomplete. This paucity of data has the potential to result in misleading or spurious relationships (Moore et al. 1989; Winemiller 1989; Martinez 1991; Polis 1991). Even with the vast complexity and variability of the data, however, researchers have documented informative and repeatable patterns in nature. The search for such food web statistics and the quest to seek out empirical patterns and their consequences for stability have been active and turbulent areas of research for over 30 years (Cohen 1978; Yodzis 1981; Cohen & Briand 1984; Sugihara et al. 1989). Food web ecologists continue to catalogue an impressive number of food web patterns and a range of models capable of reproducing these patterns (Cohen et al. 1985; Williams & Martinez 2000; Dunne 2006; Montoya et al. 2006). Beyond the arrangement of species and linkages, recent work has incorporated body size and species abundance into food web patterns, highlighting informative patterns of energy flow and connectedness in nature (Cohen et al. 2003; Woodward et al. 2005a). Further, recent research on the diet breadth of organisms has also been used to explain how network properties of food webs can emerge from foraging behaviour (Beckerman et al. 2006).

As encouraging as these results have been, there are concerns that such empirical approaches lack explicit spatial and temporal components (Polis & Winemiller 1996). In fact, ecologists are beginning to find that the variability of food webs in space and time are critical to the stability and function of ecosystems (Tilman et al. 1998; de Ruiter et al. 2005; Neutel et al. 2007). While spatial patterns in ecological systems have been extensively scrutinized (Levin 1992), this focus has largely been on the distribution of populations or spatial patterns in competition and consumer-resource dynamics (Hastings 1980, 1990; de Roos et al. 1991; Fryxell et al. 2005). Integrating spatial structure into the larger food web framework has proved more challenging. As a first step, McCann et al. (2005) linked foraging behaviour, movement and trophic position within a food web framework to show that higher order consumers have the potential to stabilize large food webs by coupling lower level webs in space. Extending this approach to incorporate biomass flux in eight documented food webs, Rooney et al. (2006) implicitly incorporated a spatial context into their analysis. They showed that food webs are comprised of energy channels based on discrete basal resources which are coupled by top predators, a result that interestingly recapitulates an observation made by Lindeman (1942) some 64 years earlier. The resulting architecture, which also exhibits consistent differences in biomass turnover rates between the coupled energy channels, was shown to impart both equilibrium and non-equilibrium stability to food webs. Whereas, this work has identified key stabilizing structures and processes in well-documented food webs, it still leaves ecologists with the task of quantifying interactions within their webs of interest. Further, the work does not explain how the observed overarching architecture emerges within the food webs.

While there has been a longstanding tradition of looking at the role of size and metabolism on important ecological traits (Peters 1983; Brown et al. 2004), ecologists have also argued that body and metabolism ought to allow us an alternative viewpoint of food web structure (Warren & Lawton 1987; Cohen et al. 2003). More recently, Woodward et al. (2005b) began to outline how body size could be more formally incorporated into our view of food web structure and function. They argue that measuring body size promises a way to collapse sets of covarying traits into one variable, without necessarily having to observe the traits directly. Specifically, they examine relationships between body size, home range size, ingestion and production rates, numerical abundance and nutrient turnover. The authors end their article by asking whether metabolic theory could be applied to predict the structure and dynamics of complex ecological networks.

Here, we provide a framework for identifying key food web structures and processes at the landscape scale. It should be noted that the goal of this paper is not to present a comprehensive review of factors that influence food web stability. Rather, we endeavour to review and synthesize several previously disparate disciplines (food web ecology, macroecology and behavioural ecology) and present a framework that integrates them into a conceptual landscape theory for food web architecture. The relationships upon which this framework is based are general relationships, often spanning orders of magnitude of variation. Although there are interesting and informative exceptions to some of these relationships (some of which are discussed later in the paper), we will not focus on the exceptions in the hopes of presenting a general model. Thus, in what follows, we review well-developed concepts in food web ecology, macroecology and behavioural ecology and project whole food web consequences of some identified relationships. We then show how both the horizontal and vertical structure of food webs can be explained by these empirical relationships, and speculate about the utility of this framework across a variety of spatial scales. We find that there is an excellent match between this collective framework and patterns in real food webs at the landscape scale. Finally, we review some recent food web concepts that explore the implications of landscape scale food web structure for the stability of whole food webs in space and time.

A brief review of empirical patterns in food web structure

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

We will refer to two food webs used in the analysis of Rooney et al. (2006) to illustrate our points – the Cantabrian Sea Shelf marine food web (original data from Sanchez & Olaso 2004) and the Central Plains Experimental Range (CPER) shortgrass prairie soil food web. Trophic interactions within these food webs were measured as flux rates of g C m−2 year−1 between trophic groups. Primary producers and detritus were assigned trophic positions of 1 and, where not directly reported, higher order consumer trophic positions (TPC) were calculated as:

  • image(1)

where n is the number of resources consumed by the consumer, PC is the proportion of the consumers diet accounted for by a resource and TPR is the trophic position of the resource. In similar manner, the proportion of carbon derived from basal resources (% BRC) was calculated as:

  • image(2)

where n is the number of resources consumed by the consumer, PC is the proportion of the consumers diet accounted for by a resource and % BRR is the proportion of carbon derived from the basal resource in the resource being consumed. A full description of the methodologies for deriving food web characteristics can be found in Rooney et al. (2006).

The results of this analysis showed the food webs to be structured such that lower order consumers derive the bulk of their energy from individual basal resources (phytoplankton or detritus in the marine system and bacteria or fungus in the soil system), whereas higher order consumers tend to integrate across the energy channels (Fig. 1).

image

Figure 1.  Lower order consumers tend to specialize on specific basal resources, whereas higher order predators ultimately integrate energy from the distinct energy channels. As these energy channels show different biomass turnover rates (P : B ratios), the food webs are structured such that top predators couple energy channels (adapted from Rooney et al. 2006) that exhibit asymmetric energy fluxes. Figures show data from (a) The Cantabrian sea shelf where the two major basal resources are phytoplankton and detritus, and (b) the CPER shortgrass steppe ecosystem where bacteria and fungi serve as basal resources.

Download figure to PowerPoint

Further analysis of the process rates within these defined energy channels revealed consistencies in biomass turnover rates between coupled energy channels. Here, biomass turnover rates are represented by the production : biomass (P : B) ratio, defined as the summed annual biomass production (g C m−2 year−1) divided by the mean annual biomass (g C m−2) for a given trophic guild within an energy channel. Coupled energy channels were shown to have consistently asymmetric biomass turnover rates within food webs. Within marine food webs, phytoplankton energy channels were shown to have consistently higher biomass turnover rates compared detrital energy channels, and within soil food webs, bacterial based energy channels displayed consistently higher biomass turnover rates compared with fungal based energy channels. Thus the resulting architecture that emerged across all studied food webs was that of paired energy channels with asymmetric energy flux (Rooney et al. 2006). We will now go on to explore some possible mechanisms that may result in this emergent architecture using concepts borrowed from macroecology, behavioural ecology and food web theory.

Relationships and hypotheses from macroecology

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

Body size and metabolism have shown promise as major organizing axes for ecology (Peters 1983; Brown et al. 2004). Body size is tightly correlated with a number of fundamental ecological characteristics (see Box. 1). These patterns appear to be general, spanning enormously different evolutionary histories and body sizes. Metabolism forms one of these broad patterns with body-size. As the metabolic rate governs resource uptake and expenditure, metabolism can be used as a general framework for understanding patterns in biomass flux. Brown et al. (2004) have referred to this framework as the metabolic theory of ecology (MTE), and used this theory to explore numerous patterns in life history (e.g. mortality, development rate) and ecosystem processes (e.g. biomass production, biogeochemical cycles). This body of research provides compelling evidence for a slow-fast metabolic continuum across body-sizes. Specifically, we outline the following three size-based relationships that are directly related to biomass flux:

Table Box 1.   Allometry and metabolic theory
Allometry: the empirical study of power formulas involving body mass which can be expressed mathematically as: ln(Y) = a + b*ln(body Size); where Y is a specific attribute of interest, a is the intercept, and b is the slope of the empirical relationship
Metabolism-body size: Both historical and recent temperature-corrected estimates of slope suggest the slope is 0.75. Thus, the per unit weight metabolic cost scales with body size to the −0.25. Larger organisms, therefore, tend to have lower metabolic costs (Brown et al. 2004)
Consumption rate-body size: Both historical and recent temperature-corrected estimates of slope suggest the slope is 0.75. Thus, the per unit weight consumption rate scales with body size to the −0.25. Larger organisms, therefore, tend to have a lower consumption rates per unit weight (Brown et al. 2004)
Longevity-body size: Higher metabolism per unit weight implies higher biomass loss rates and mortality rates. Both historical and recent temperature-corrected estimates of the mortality body size slope suggest the slope is 0.75. Thus, larger organisms with lower metabolic costs tend to live longer and have lower mortality rates (Brown et al. 2004)
Brain size-body size: Brain size scales with body size such that larger organisms tend to have bigger brains, with the slope of the relationship equal to 0.75 (Harvey & Pagel 1988)
Turnover Rate: [production (P): biomass (B) ratio] If we assume that on long time scales the rate of biomass loss (flux out) must be equal to the rate of biomass gain (must be true for species that do not go extinct) then a slow metabolism implies slow biomass gain. This implies that the turnover (P : B) ought to have slope of approximately −0.25. The empirical relationship is consistent with this except within trophic levels appears to be slightly more negative (Kerr & Dickie 2001). Nonetheless, larger organisms tend to have lower turnover rates (Brown et al. 2004)
Movement and metabolism: lower cost of transportation means movement ought to be more fully adapted in larger organisms. Estimates of the movement body size slope suggest the slope is near 0.75 (Peters 1983)
Brain size and foraging behaviour: Larger more mobile organisms experience a wider variety of habitats. Further, increased brain size has been found to correlate with ability to make foraging decisions within a complex habitat (Eisenberg & Wilson 1978; Harvey et al. 1980; Martin 1981; Budeau & Verts 1986)

R.i(a) as body size increases, the metabolic rate per unit of biomass tends to decrease (Peters 1983; Brown et al. 2004).

R.i(b) as body size increases, the consumption rate per unit of biomass (consumption/weight) tends to decrease (Peters 1983; Brown et al. 2004).

R.i(c) as body size increases, the biomass turnover rate (biomass/weight) tends to decrease (Peters 1983; Brown et al. 2004).

Thus, the given differences in mean body size between distinct organisms in a food web then, these relationships immediately identify differences in biomass flux. Small organisms tend to have higher loss rates, higher growth rates and greater biomass turnover than larger organisms. These relationships are readily identifiable in both the Cantabrian Sea Shelf food web and the CPER soil food web, where production : biomass ratios decline with increasing body size (Fig. 2). A further empirical relationship not directly linked to biomass flux allows additional insight into how the body size influences other aspects of food web structure:

image

Figure 2.  Biomass turnover rates decrease with increasing body size. This pattern is consistent in both (a) The Cantabrian sea shelf (slope = −0.26, r2 = 0.78, P < 0.001), and (b) the CPER shortgrass steppe ecosystem (slope = −0.11, r2 = 0.67, P = 0.002).

Download figure to PowerPoint

R.ii as body size increases, trophic position tends to increase (e.g. Warren & Lawton 1987).

Although there are exceptions to this pattern to be found in nature (e.g. Layman et al. 2005), this pattern has been observed in many food webs (e.g. McCann et al. 2005; Jennings et al. 2007) and the pattern is consistent for the two food webs examined in this paper, although to a lesser extent within the soil food web (Fig. 3). This pattern will be explored more fully later in the paper.

image

Figure 3.  Larger organisms tend to have higher trophic positions in (a) The Cantabrian sea shelf (slope = −0.31, r2 = 0.59, P < 0.001), and to a lesser extent (b) the CPER shortgrass steppe ecosystem (slope = 0.13, r2 = 0.21, P = 0.11).

Download figure to PowerPoint

Two additional macroecological relationships have been documented that inform us to how food webs might operate in space and time.

R.iii as body size increases, the metabolic cost of transport per unit biomass tends to decrease and so larger organisms tend to be more mobile (Peters 1983).

R.iv as body size increases, total brain complexity increases (Harvey & Pagel 1988).

Thus, given the significant changes in mean body size between distinct organisms in a food web then, these three relationships (R.ii–iv) immediately identify gross differences, in trophic position, mobility and foraging ability. By piecing together these six broad relationships we can generate some elementary, but very useful, hypotheses about the architectural framework of food webs on the landscape. Below, we present hypothesized food web properties that emerge from the synthesis of the above empirical relationships (i.e. R.i–iv).

On the basis of relationships i and ii above, we can generate hypotheses as to the distribution of energy fluxes within the food web. Specifically:

H.i Biomass turnover rates (P : B ratios) of species should decrease with increased trophic position within food webs.

On the basis of the notion that the larger organisms are more mobile on the landscape (R.iii), and will move across a broader range of habitats than organisms at lower trophic levels, we can generate two separate, but linked hypotheses regarding the gross architecture of food webs.

H.ii(a) Higher trophic level organisms in the food web ought to increasingly couple the more spatially isolated lower trophic levels through consumption.

H.ii(b) Lower trophic levels should be more spatially isolated or compartmentalized compared with higher trophic levels.

Finally, there exists a body of research on the allometric relationship between brain and body size (Box 1). Before delving into this relationship too far, we make a distinction between the evolution of increased brain complexity and size observed across phyla within the animal kingdom, and the increase in the relative size of brains relative to body mass within and among vertebrates (e.g. birds or mammals). For both frames of comparative reference, empirical evidence points to the ability for taxa with larger and complex brains to process, integrate and store more information. Specifically, brain size relative to body mass has been shown to correlate to innovation (Lefebvre et al. 1997, 1998), cognition (Sol et al. 2005), learning (Bouchard et al. 2007), and the ability to store complex information from a heterogeneous environment (Budeau & Verts 1986). Further, increased relative brain size also maps to increased sensory perception (Hutcheona et al. 2002), no doubt essential for increased foraging efficiency. While much of this research has focused on relative brain size, a recent meta-analysis found that even within taxa, absolute brain size is the best predictor of cognitive ability (Deaner et al. 2007). This relationship is likely to be even stronger across taxa, within food webs, for example, from zooplankton to tuna, where the complexity of the brain (i.e. development) and brain function increases. Taken together, these findings suggest that organisms with larger and more complex brains (i.e. more highly developed) have the capacity for increased sensory perception and the ability to not only innovate but learn and store acquired information. These characteristics, combined with the increased scale of movement of larger bodied organisms, most certainly ought to translate into increased foraging abilities.

One can therefore generate a prediction more formally in terms of how this enhanced foraging ability ought to change attributes of the functional response. Given that the world is a heterogeneous collection of resources (i.e. there are high and low densities of resources on the landscape), then higher order organisms with greater brain complexity ought to more readily store information that enables them to leave low density patches for more profitable high density patches. From the perspective of any local patch then, such a large mobile consumer ought to display a type III functional response. Thus, our final hypothesis is:

H.iii Higher trophic level organisms ought to have more developed foraging abilities and therefore be more likely to exhibit type III functional responses on average than those that populate lower trophic positions.

Testing hypotheses at the landscape scale

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

To test hypothesis H.i, we examined the relationship between trophic position and P : B ratios in our representative food webs. We note here that our analysis of the food webs is not meant to be exhaustive, and the various correlations presented are not necessarily statistically independent since they share various variables in various permutations. The analyses, however, do provide a valuable tool for exploring the hypotheses presented in the preceding section. Figure 4 shows that for both the Cantabrian Sea food web and, to a lesser extent, the CPER soil food web, P : B ratios decrease with increased trophic position. In fact, for the eight food webs summarized in Rooney et al. (2006), in 20 of 24 cases increases in trophic level also are followed by decreases in P : B ratios or biomass turnover. Thus, biomass turnover does appear to decrease with increasing size and trophic level (P < 0.001). Interestingly, as biomass flux based estimates of interaction strength are subsumed by the definition of biomass turnover rate (Rooney et al. 2006), this relationship implies that interaction strengths should decrease as trophic level increases. As such, macroecological relationships within food webs unfold to result in ordered distributions of interaction strengths which may have profound implications for the stability of food webs. There will, of course, be exceptions to this general pattern. For example, Bascompte et al. (2005) found that a few shark species were involved in the bulk of strongly interacting tritrophic (predator–consumer–resource) food chains in a large Caribbean marine food web. Such strongly interacting top predators are key in the occurrence of trophic cascades. Interestingly, even when these instances occur, the general pattern remains. For each of their top 10 predatory species identified in Bascompte et al. (2005), mean predator–consumer interactions were always less than the mean consumer–resource interactions. On average, interaction strengths were six-fold higher between consumers and resources compared with predator–consumer interactions.

image

Figure 4.  Consistent with the earlier observed patterns, biomass turnover rates tend to be slower in higher order predators in (a) The Cantabrian sea shelf (slope = −0.59, r2 = 0.62, P < 0.001). (b) The relationship is not, however, significant in the CPER shortgrass steppe ecosystem (P = 0.23).

Download figure to PowerPoint

With respect to hypotheses H.ii(a) and H.ii(b), Fig. 1 shows that based on feeding linkages, higher trophic level organisms derive energy from different energy channels, suggesting increased movement among habitats compared with lower trophic level organisms. Further evidence for this hypothesis comes from the empirical relationship between body size, trophic position and movement (Fig. 5). Higher order predators tend to be larger and have greater scales of movement than their prey. While acknowledging that data are only available for two ecosystems, this relationship appears invariant to scale as we observe this across the expansive Atlantic Ocean food web (used here as an example of a marine food web, in place of the Cantabrian sea shelf where the data was unavailable) at the kilometre scale and within the CPER soil food web at the kilometre subcentimetre scale (Fig. 5a, b respectively). This and other similar hierarchical examples suggest that food webs should generally be compartmentalized at lower resource levels but increasingly coupled by higher order mobile predators (Moore & Hunt 1988). These results taken together point to an organizational and a functional compartmentation within food webs, which has historically been a source of controversy in food web ecology (Pimm & Lawton 1980). Earlier attempts to detect compartments were, however, hampered by the fact that the data often did not include measures of biomass flux or interaction strength. Recently, using a novel technique from the sociological literature, Krause et al. (2003) found evidence of compartmentation. Krause et al. (2003) noted that the ability to detect compartments appeared to dependent upon the detail of the food web data (both resolution and interaction strength) whereby more complete webs showed more evidence of compartmentation.

image

Figure 5.  The relationship between body size and movement is also linked to trophic position. Higher order predators tend to be larger and more mobile than their prey in both a) the North Sea Shelf (adapted from McCann et al. 2005) and b) the CPER shortgrass steppe.

Download figure to PowerPoint

It is difficult to assess foraging ability (i.e. hypothesis H.iii) using the food webs presented in this paper. Nonetheless, the literature on phylogeny and brain size and development have used novel studies to show that organisms with greater brain size and development also appear to display enhanced abilities to respond to complex foraging cues (Eisenberg & Wilson 1978; Harvey et al. 1980). It would be a mistake to interpret H.iii as saying that these organisms only populate upper trophic positions as they populate food webs as consumers along with those with lesser develop nervous systems and less complex brains throughout the food web. However, concerns of study-bias notwithstanding (Hassell et al. 1977), the upper trophic positions where higher order predators reside tend to be populated by taxa from the apexes of either the deuterostome or protostome lines within the animal kingdom (e.g. chordates, cephalopods, and arthropods); groups which have highly developed nervous systems and brains, and which also tend to have more sigmoidal functional responses than lower trophic level organisms (Holling 1965).

Patterns within food webs

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

While the established empirical relationships have identified some gross properties of food web architecture, it would be interesting to see whether these relationships can help explain some more specific food web characteristics when taken in the context of coupled energy channels. We now explore the connection between our identified food web relationships and specific attributes of the energy channels identified within food webs. Since we have shown that food webs couple many very different habitats from above, it behoves us to question whether different habitats have different food web traits.

If we take three relationships together (R.i–iii) that body size should increase with trophic position, that biomass turnover should decrease with increasing body size, and that fast and slow energy channels are ultimately coupled by higher order consumers, a number of predictions ensue with respect to the internal architecture of food webs. First, for given trophic levels, fast channels should, on average, contain smaller organisms compared with slower channels. The result should be of different slopes and intercepts for the relationships between trophic position and body size for fast and slow energy channels. Further, given that higher order predators couple the energy channels, the differences should diminish as one ascends through the food web, and this is in fact observed in Fig. 6a,b. This pattern suggests that taking into account energy channels within food webs may help explain the absence of this relationship in food webs that are based on numerous basal resources (Layman et al. 2005). Examining the partitioned food webs also provides further insights into another established allometric relationship, that between trophic position and P : B ratio. When we partition the relationship between trophic position and P : B ratio between energy channels, a pattern emerges that helps to explain some of the earlier observed variation in the relationship at the whole food web scale. That is to say that organisms in slow energy channels have, for a given trophic position, lower P : B ratios (Figs 6c,d). These relationships also converge at higher trophic levels, reflecting once again the spatial coupling of energy channels by higher order predators.

image

Figure 6.  Basic metabolic relationships differ between groups of organisms that occur in fast and slow energy channels. The relationship between body size and trophic position differs such that for a given trophic position, organisms within the slower energy channels are larger. This trend holds true for (a) Cantabrian Sea Shelf, where the slope of the relationship between body size and trophic position within the phytoplankton channel (slope = 2.77, r2 = 0.79, P < 0.001) is significantly higher (ancovaF = 12.31, d.f. = 23, P = 0.002 for interaction effect) than the slope for the same relationship within the detrital channel (slope = 0.74, r2 = 0.28, P = 0.03). A similar pattern holds for (b) the CPER shortgrass steppe ecosystem where the relationship between body size and trophic position within the bacterial channel (slope = 3.1, r2 = 0.75, P = 0.005) is significantly higher (ancovaF = 10.07, d.f. = 9, P = 0.02 for interaction effect) than the slope for the same relationship within the fungal channel, where the slope is not significantly different from 0 (P = 0.55). The energy channels also exhibit consistent and highly suggestive differences in the relationship between trophic position and biomass turnover in (c) Cantabrian Sea Shelf, where the slope of the relationship between body size and log (P : B ratio) within the phytoplankton channel (slope = −0.7, r2 = 0.61, P = 0.003) appears to be more negative (ancova, F = 2.76, d.f. = 23, P = 0.11 for interaction effect) than the slope for the same relationship within the detrital channel (slope = −0.38, r2 = 0.71, P = 0.001). Note that this relationship does not include detritus (P : B ratio of 0) as a trophic level of 1, which would make the slope of the relationship shallower in the detrital channel. The difference in slopes does hold for the CPER shortgrass steppe ecosystem where the slope between body size and log(P : B ratio) within the bacterial channel (slope = −0.45 r2 = 0.86, P = 0.003) is significantly more negative (ancovaF = 6.43, d.f. = 9, P = 0.03 for interaction effect) than the slope for the same relationship within the fungal channel, where the slope is not significantly different from 0 (P = 0.73).

Download figure to PowerPoint

Since biomass flux based estimates of interaction strength are subsumed by the definition of biomass turnover rate (see Rooney et al. 2006), relationship (R.iii) can be re-interpreted in terms of interaction strength (per unit of predator biomass). Thus, interaction strength ought to decrease with increasing body size, suggesting differential energy fluxes and interaction strengths between energy channels. This exact pattern of asynchronous biomass turnover was pointed out by the analysis of Rooney et al. (2006). In 11 of 11 possible cases, pelagic biomass turnover at a given trophic level was greater than benthic biomass turnover (P < 0.0001). For soil food webs the analyses yield a similar trend with the smaller-sized bacterial channel organisms tending to biomass turnover at a greater rate than the fungal channel organisms, but the result is only moderately significant (= 0.07). More precise measures of interaction strength will clearly depend on the ratio of predator : prey biomass ratios and other factors (de Ruiter et al. 1995; Emmerson & Raffaelli 2004). However, these results may also yield insights in light of the results of Bascompte et al. (2005), who found a non-random distribution of interaction strengths in a marine food web. Specifically, they found that the occurrence of strong interactions on two consecutive levels of food chains occurs less frequently than expected by chance. Further, when they do occur, they are accompanied by strong omnivorous links more often than expected by chance. Given this relationship, one might predict strong interactions within fast channels to be more associated with ominvory, and the presence of larger bodied, lower trophic level species in adjacent slow channels might provide energetically feasible omnivorous alternatives.

Food web architecture across spatial scales

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

It would be interesting to know if these energy channels also occur at smaller and larger spatial scales as suggested by both the Cantabrian Sea Shelf marine food web and the CPER shortgrass prairie soil food web. There are some reasons to expect that they may. At smaller spatial scales within aquatic ecosystems, for example, the pelagic or open water zone has zooplankton and pelagic planktivores that couple the classic planktonic grazing chain (based on inorganic nutrients) to the microbial loop (based on organic carbon). Interestingly, the classic pelagic grazing chain also tends to have larger organisms (e.g. crustaceans) than the microbial loop (e.g. protists, Fig. 7a). The summed production and trophic interactions within the pelagic zone are often collapsed into one energy channel by limnologists who study benthic–pelagic coupling in lakes. The resulting food web then incorporates both benthic and pelagic players, and this architecture also appears to exhibit similar body size relationships between channels (Fig. 7b). Further, at very large scales, birds and mammals couple aquatic and terrestrial ecosystems (Fig. 7c), which again display differential body size spectra for given trophic levels. Interestingly, this pattern has also been suggested to account for differences in trophic cascades between terrestrial and aquatic habitats (Shurin & Seabloom 2005; Shurin et al. 2006), which could be thought of as a signature of energy channel strength. The underlying suggestion here is that the pattern of distinct channels, differential size within channels, and consumer coupling may exist across spatial scales. All appear to show some degree of resource compartmentation (as evidenced by different basal resources) as well as coupling by higher order predators in the food webs. Further work is required but the adaptation of organisms to different aspects of the ecosystem (e.g. pelagic vs. benthic, dry vs. wet soil, grazing vs. detrital, above-ground vs. below-ground, terrestrial vs. aquatic) may tend to produce resource compartments defined by asymmetric body size relationships. If so, then we would also expect a tendency for asymmetric rates of biomass turnover in these different compartments.

image

Figure 7.  Spatial scaling of the coupled energy channel architecture. The horizontal lines represent the range in body size observed within energy channels and trophic guilds. Green and red boxes around the lines correspond to fast (small) and slow (large) energy channels respectively. Based on body size, it appears that the coupling of energy channels by higher order consumers operates at spatial scales varying from (a) within habitat (classic chain–microbial loop coupling), through (b) between habitats (benthic-pelagic coupling) to (c) between ecosystems (aquatic-terrestrial coupling).

Download figure to PowerPoint

Recent research has proposed a ‘growth hypothesis’ that potentially ties into the above landscape theory (Elser et al. 1996). Specifically, this research argues that body size ought to correlate with N : P ratios such that larger organisms have lower P levels and higher N : P ratios. Further, low N : P ratios tend to imply more rapid growth rates since organisms laden with phosphorous heavy RNA will tend to produce organisms capable of rapidly synthesizing proteins (Sterner 1995; Elser et al. 1996). Thus, low relative N : P ratios tend to produce high relative growth rates. If this pattern holds across body size then their arguments are entirely consistent with the food web predictions above. Additionally, their lower level mechanism offers insight into how these different resource compartments ought to recycle nutrients. Benthic pathways, for example, ought to have lower phosphorous demands than the faster growing pelagic pathway. Similarly, fungal pathways ought to have lower phosphorous demands than bacterial pathways. We leave this interesting intersection of ideas for future research.

Exceptions and limitations

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

As with all such general patterns, there exist exceptions and limitations to the general model. For example, within some food webs, consumers can be more mobile compared with their predators despite their similar size (Hebblewhite & Merrill 2007). Further, some of the relationships outlined above do not apply to arthropods, where smaller organisms often have higher trophic positions compared with their larger prey (e.g. parasitoids and biting insects). Such smaller bodied organisms may in fact have increased dispersal rates, and thus may functionally couple otherwise ecologically isolated habitats. It is interesting to note, however, that these two very attributes can lead to a similar pattern of coupled habitats outlined in this paper. Eveleigh et al. (2007) document that spruce budworm outbreaks in a balsam fir and mixed deciduous forest stands result in a higher diversity of primary parasitoids and higher-order generalist parasitoids (hyperparasitoids) and a greater percentage of species at higher trophic levels. In this case, even though the higher order predators are smaller, they respond to ecological cues at the landscape scale to suppress the outbreak of a resource within a defined habitat.

Some emerging theoretical implications

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

Although it has been long known that the population dynamics of species within ecological systems and the dynamics of systems as a whole are variable, ecologists are only now really beginning to fully embrace this fact (DeAngelis & Waterhouse 1987; Chesson & Huntley 1997; Tilman et al. 1998). Recent theory has argued that variability in resources in both space and time create a complex biological canvas that consumers can react to by decoupling from some consumer–resource interactions or re-initiating other consumer–resource interactions (Holt 2002; McCann et al. 2005). If this decoupling occurs when densities in the resources are low, on average, then such biological structure can drive persistent food webs by allowing the subsystem a reprieve from consumptive pressures exactly when it needs it – when the subsystem is experiencing low average densities. Similarly, if coupling occurs when resource densities are high, then top-down pressure is engaged just when a resource is growing excessively.

The food web architecture outlined in the preceding sections lends itself to these theoretical arguments (Box 2). In a sense, the food web is organized such that the dynamics are driven by a combination of both bottom–up and top–down forces that inspire and mute variability. Landscape scale food webs imply that the basal resources are ultimately separated by large ecological distances in space. As such, there is a tremendous amount of variation in space in the productivity and density of these lower level organisms. The mobile higher order predators, on the other hand, respond to this lower level variance. Given their ability to store complex information, then they can rapidly respond to this variation. This rapid behavioural response is critical in that it allows large mobile organisms the ability to respond at time scales much faster than population dynamics. Finally, the intermediate consumers in this architecture are defined by series of trophic levels with very different rates of production. As such, they beget energy pathways or channels that supply energy to the higher levels at different rates. These differential pathways, or asymmetrical production, tend to drive the asynchronous responses of the different pathways discussed above (Rooney et al. 2006). Thus, the identified landscape scale architecture has the key ingredients for maintaining persistent non-equilibrium ecosystems. Unfortunately, anthropogenic modifications may be threatening these key ingredients (Box 3).

Table Box 2.   Bottom–up and top–down synergy: a potent stabilizing mechanism
The food web architecture discussed above provides large scale mechanisms that enable a system to respond to a variable world (see Fig. B.2.i below) in a manner that promotes the persistence of complex webs. Specifically, we highlight three integral aspects of the structure that recent theory has used to argue promotes persistent complex webs
Lower trophic level organisms (the variation): the basal resources spread out across the landscape to provide a heterogeneous landscape. Due to the different life histories, different habitat characteristics, etc. resources can be expected to readily show some degree of asynchronous production in space and time. In other words, these lower level organisms produce variation across the landscape
Higher level organisms (the couplers): The higher order consumers are mobile and capable of responding to complex environmental information. Given resources of high and low density on the landscape such a coupler will tend to move in space to reduce the high density resource and therefore simultaneously release the low density resource from predation pressure. Note, that if space is limited or homogeneous then the stabilizing influence of higher order predators will be compromised. In such a case, they can drive strong top-down suppression. Thus, in spatially constrained ecosystems, mobile predators can be destabilizing (McCann et al. 2005)
Intermediate pathways or energy channels (the asymmetry): Food webs show pathways with different abilities to propagate energy. This serves two aspects of regulating the dynamics of food webs. One, such an architecture allows the ability for a food web to respond rapidly to a large perturbation (i.e. the fast pathway or channel). Two, as the fast channel responds the combination of slow and fast channels acts to drive compensatory dynamics of the different pathways (Rooney et al. 2006). In other words, the internal asymmetry of energy flow readily promotes asynchronous dynamics. This asynchrony thus helps maintain the precise conditions required for the couplers to mute lower level variability (see Couplers above) inline image
Table Box 3.   Human influences on the landscape
Research suggests that human influences have homogenized space and production on the landscape. Further, human harvesting and habitat fragmentation has eliminated numerous mobile higher order predators (Fig. B.3.i below). Thus, the human influence appears to be targeted directly at eliminating the ‘variability’ and the ‘couplers’ that underlie landscape level food webs and their stability
Resource homogenization (loss of resource variation): Elevated inputs of phosphorous and nitrogen runoff from surrounding landforms, often disproportionately affects a single compartment (e.g. pelagic or bacterial); thus converting a spatially structured and compartmented ecosystem to one that is much more homogeneous in its basal production. This is common to both aquatic and agricultural ecosystems (Hendrix et al. 1986; Scheffer et al. 1997). Invasive species have also dramatically influenced the structure of ecosystems. For example, the zebra mussel, has completely shunted the pelagic production in lakes to the nearshore (Hecky et al. 2004). Hecky et al. (2004) referred to this as the Great lakes benthic energy shunt, and again human impact has strongly skewed or homogenized basal production
Large predator removal (loss of couplers): Human modification has ubiquitously harmed large mobile predators. Fisheries researchers have now shown that culling has led to the fishing down of the food web, a progressive decline in top predators (Pauly et al. 1998; Jackson et al. 2001). Terrestrial ecosystems also have lost large mobile organisms due to habitat fragmentation and culling (Terborgh et al. 2001). There has been documented losses in diversity, stability and function accompanying such changes (Jackson et al. 2001; Terborgh et al. 2001; Worm et al. 2006). In a very recent survey of 48 quantitative parasitoid-pollinator food webs across a gradient in human modification, researchers found that highly modified webs were dominated by one or a few pathways (Tylianakis et al. 2007). Here, parasitoid : host ratios were inflated and parasitism rates elevated which influences critical ecosystem services like pollination (Tylianakis et al. 2007) inline image

Along these lines, recent spatially motivated food web theory has found that ecosystem size, or the homogenization of habitats, ought to seriously threaten nature’s balance (McCann et al. 2005). In small ecosystems, the mobile predators’ stabilizing influence can be reversed as heightened mobility in a spatially limited ecosystem can effectively synchronize lower level resources. This synchrony reduces habitat variability and unites runaway dynamics. The homogenization of habitats has similar outcomes. In both cases, the loss of lower level variability in space compromises the ability for the mobile generalist foragers to mute runaway dynamics (McCann et al. 2005). This theory depends on asynchronous dynamics in space. There exists a burgeoning literature on the synchronizing influence of large scale abiotic drivers on population dynamics (the Moran effect). This literature is largely focused on population level phenomena and theoretically and empirically tends to show diminished synchrony with increasing spatial scale (Lande et al. 1999). In a very simplified sense then this research suggests that nearby organisms ought to be more correlated, and so the stabilizing role of predators coupling in space may therefore rely on coupling of food webs at large spatial scales. Given that different species have different traits, then the synchronizing effect of large scale environmental drivers on different species may be weaker than that found within a species (Engen & Saether 2005). These ideas beg for an empirical analysis of the covariance dynamics between different species and subsystems across spatial scales. Such empirical findings may aid our understanding of what inspires unified subsystem dynamics (i.e. resources increase or decrease together) as well as what inspires asynchrony. It may also help us identify critical levels of aggregation for food web dynamics. In other words, asynchronous trophic aggregates may identify important functional groupings for a given spatial and temporal scale.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

Over 30 years ago, Schoener (1974) reviewed the available literature and concluded that food, habitat, and time were the three principal resources or niche axes of the niche hypervolume for individual species. Taken within this context, our work places the relationships among species along these very axes. We show that the trophic interactions among species form the foundation of a hierarchical organization of communities that lends itself to dynamically stable architectures. Here, we have used body size and metabolism to reveal some general properties of food webs that are matched by available data. Specifically, food webs are large comprehensive entities whereby localized resources (food) and habitats are progressively coupled by more mobile higher order predators or consumers. We then review theory and relate it to the landscape architecture of food webs. We argue that the large organisms promote the balance and maintenance of a diverse and variable assemblage of organisms while the diverse lower level organisms, in turn, form a complex system of species capable of differentially responding to an ever changing world. Interestingly, variation in body size also maps to horizontal variation in food web characteristics. Specifically, differences in body size explain variation in energy flux rates within coupled energy channels, which has been shown to be an important stabilizing architectural component of food webs (Rooney et al. 2006). In this sense, the landscape architecture discussed here implies that nature is an intriguing balance of bottom–up (habitat heterogeneity) and top–down (predator) forces. Current human influences tend to homogenize resources and remove higher order couplers thus seriously threatening the non-equilibrium balancing act of complex food webs.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References

We thank J. Newman, and A. de Bruyn for helpful comments. This work was supported by grants from NSERC to K.S.M. and the US National Science Foundation to J.C.M. Support was also provided by a National Center for Ecological Analysis and Synthesis (NCEAS) grant to the ‘Detritus Working Group’.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. A brief review of empirical patterns in food web structure
  5. Relationships and hypotheses from macroecology
  6. Testing hypotheses at the landscape scale
  7. Patterns within food webs
  8. Food web architecture across spatial scales
  9. Exceptions and limitations
  10. Some emerging theoretical implications
  11. Conclusion
  12. Acknowledgements
  13. References