- Top of page
The nature of the relationship between resource availability and competitive intensity has been the subject of intense debate (Thompson 1987; Tilman 1987a, 1987b; Thompson & Grime 1988) and numerous empirical tests that yielded contradictory or inconsistent results (Wilson & Lee 2000). We propose that these inconsistencies and controversies result from methodological issues.
One argument states that competition occurs in the most productive and benign habitats (e.g. temperate rainforests), and also in habitats that are subject to high levels of abiotic stress (e.g. artic tundra and deserts), but it is argued that there are qualitative differences (Tilman 1988; also see Newman 1973). Competition for soil resources may predominate in harsh abiotic environments, whereas competition for light is more important in productive habitats. This trade-off is assumed to exist because biomass allocated to root systems cannot be simultaneously allocated to tissues used to acquire light (Tilman 1988). The alternative argument is that competitive intensity reaches a maximum in the most benign habitats with the ability to capture above- and below-ground resources linked by positive feedback (Donald 1958, cited in Grime et al. 1997): access to mineral nutrients results in the ability to build photosynthetic enzymes, which produce energy, which promotes the uptake of mineral nutrients, and so on. Intensity of competition (both root and shoot combined) is therefore expected to increase as a function of resource availability.
The comprehensive review by Aerts & Chapin (2000) on the effects of resource availability on plant structure and function supports some elements of both arguments. An alternative approach is to use removal experiments in which intensity of competition is measured as a function of resource availability, using either natural (standing crop) gradients or artificial resource gradients, but the results are equally inconsistent. One possible reason for these inconsistencies is variation in scale, or range, of resource availability investigated. Belcher et al. (1995), Bonser & Reader (1995) and Foster (1999) all found a logarithmic relationship between standing crop and the relative intensity of competition (RCI), which would not be apparent in studies using a narrow range of resource availability, or only the upper end of the resource gradient (where the log function flattens out). Disturbance and fertility can, however, be confounded on gradients of standing crop (Wilson & Tilman 1991, 1993; Peltzer et al. 1998), which questions the validity of the correlation with RCI. Glasshouse studies of the relationship between RCI and fertilization, such as we present here, avoid such confusion. With the exception of the work by Miller (1996), we are unaware of other studies that have explicitly considered this question.
Removal experiments may also confound resource competition with indirect effects (Tilman 1987b). Citing such studies in support of Grime (1977) implies the acceptance of a ‘non-mechanistic, Lokta-Volterra-based, phenomenological definition of competition’ (Tilman 1987b). According to the best such definition (Shipley et al. 1991), competition occurs when there is a ‘decrease in the fitness of a plant … due to the presence of another plant, without any necessity that the decrease in fitness be due to differential consumption of a limiting resource’. However, indirect effects, such as apparent competition (Connell 1990), can then be confused with resource competition (Tilman 1987a), as when grasses suppress the growth of cacti by attracting/housing invertebrate herbivores (Burger & Louda 1994). Strong correlations are often observed between the biomass of the plant community and the abundance of invertebrate herbivores (Southwood et al. 1988; Bonser & Reader 1995; Strong et al. 2000) and the effects of resource competition and herbivory may therefore be confounded (Reader 1992).
Experimental tests of these arguments have been conducted with a fairly limited range of habitats and life-forms. Examples include aquatic reeds along swamp margins (Wilson & Keddy 1986; Shipley et al. 1991), herb-fields (Belcher et al. 1995; Reader 1990), old-fields (Wilson & Tilman 1991, 1993; Reader et al. 1994) and grasslands (Peltzer et al. 1998). We only found a single comparative study of the relationship between resource availability and competition in prairie and forest (Wilson 1993). Plant architecture affects the invertebrate community (Lawton 1983) and the effects of ‘competition’ from trees/shrubs and grasses/herbs may be quantitatively and qualitatively different along gradients of natural productivity, where the spectrum of life-forms often changes dramatically.
We have three main objectives: (a) to assess whether there is a logarithmic relationship between competitive intensity and resource availability; (b) to ascertain if there is any potential for confounding the effects of competition and herbivory in a field experiment; and (c) to determine whether changing the identity (life-form) of competing vegetation (shrubs vs. grasses) could affect the relationship between experimentally manipulated resource availability and competitive intensity (phenomenologically defined).
- Top of page
In contrast to many studies that have used artificial resource gradients (Goldberg & Barton 1992), our results are consistent with the unified concept of competition, closely mirroring those of Reader (1990). Without additional resources (i.e. in unfertilized pots and unfertilized/unwatered plots), the absence of competitors did not benefit tree seedlings. Thus competition was constrained by low nutrient supply both in the glasshouse (where we were able to measure both above- and below-ground biomass of eucalypt seedlings) and in the field (where we were restricted to measuring aerial biomass). Thus, whilst our results cannot falsify the argument that there is a trade-off between root and shoot competition, they provide strong support for the argument that the overall intensity of competitive interactions reaches a maximum in fertile habitats.
However, the variable used to assess competitive effects influences the outcome of the analyses: when survival (rather than biomass) data are used the relationship between resource availability and competitive intensity appears to be neutral (zero slope), as predicted by Newman (1973) and Tilman (1988). Reader (1990), Berkowitz et al. (1995) and Sammul et al. (2000) also found that the choice of dependent variable has significant implications. This problem could be resolved by using a mathematical function that includes more than one demographic parameter, as developed for instance in McPeek & Peckarsky (1998), although their model is not easy to generalize because it must be tailored to the life history of the species under study. However, because of the modular construction of vegetation, biomass is often an excellent indicator of plant fitness (e.g. Molofsky et al. 2000) and multiple parameter models may not be necessary. Some of the surviving E. camaldulensis seedlings in the plots with T. triandra had only one or two leaves left, and invertebrates had eaten 75% of each remaining leaf: their poor condition is apparent from their biomass, but not from survival data. Similarly, comparable numbers of tree seedlings survived in plots with Acacia pycnantha and no vegetation, and fertility had no effect upon survival, although E. camaldulensis seedlings in plots with no vegetation and additional resources were, on average, four times as large as those in unfertilized plots with no vegetation, and 12 times as large as in plots with A. pycnantha. We therefore conclude that biomass data, at least in this instance, are more meaningful and that increased resource availability resulted in more intense competition. This conclusion is supported by the correlation between seedling size and probability of survivorship reported by Sarukhán et al. (1984).
Previous field studies demonstrated a logarithmic relationship between standing crop and RCI (Reader et al. 1994; Belcher et al. 1995; Bonser & Reader 1995; Foster 1999). However, the argument that failure to detect a positive correlation between fertility and RCI may result from the use of a narrow range of fertility levels, and/or the use of fertility levels in the upper regions of the relevant gradient, where the log function flattens out, is seriously weakened because the effects of disturbance and resource availability can be confounded on gradients of standing crop (Wilson & Tilman 1991, 1993; Peltzer et al. 1998). As our glasshouse experiment used an artificial resource availability gradient, and there was no disturbance, we can assert with confidence that RCI increased substantially between the lowest and the intermediate resource level but barely, if at all, between the intermediate and the highest resource availability. This is consistent with a logarithmic response of RCI to resource availability.
Our results suggest that the effects of resource competition and herbivory can be heavily confounded, as previously reported (e.g. Southwood et al. 1988; Reader 1992; Burger & Louda 1994; Cebrián & Duarte 1994; Berkowitz et al. 1995; Bonser & Reader 1995; Strong et al. 2000; Groner & Ayal 2001; Scheidel & Bruelheide 2001). Although the leaf damage data suggest that the main mode of suppression may have been herbivory rather than competition, tree seedlings in the plots with Acacia pycnantha were only lightly grazed, and hence their reduced growth (an effect that increased with resource availability) was probably caused by resource competition. Regardless of the mechanism (invertebrate herbivory or resource competition) that limited the establishment of tree seedlings, the same underlying outcome was always apparent, i.e. additional growth was not possible without additional resources.
Because the effects of herbivores and resource competition in the field were correlated, the unified concept of competition was best supported when competition was defined phenomenologically. In the original C-S-R model (Grime 1977), herbivory was considered to be a form of disturbance, as was any factor that caused the partial or total destruction of biomass. Because the effects of resource competition and invertebrate herbivory are so closely associated, and are often confounded in experiments, it may be more appropriate to adopt a phenomenological definition of the interaction, and to group these processes under a single banner (e.g. interference), as suggested in the Southwood-Greenslade habitat templet (Southwood 1988). Studies that have assessed the relationship between RCI and resource availability using removal experiments on natural (standing crop) gradients may have measured the combined effects of invertebrate herbivory and competition, rather than resource competition alone. This may in part explain the observation by Goldberg & Barton (1992) that studies with natural gradients generally support the UCC, whereas studies with artificial resource gradients generally refute it. It also validates one of Tilman's (1987a) main objections to the C-S-R model: indirect effects are extremely easy to confuse with resource competition in removal type experiments. This observation highlights the value of continually revising and modifying the general framework (see Southwood 1988; Pickett et al. 1994).
Our results also indicate an important effect of biotic neighbourhood, or ‘plant architecture’ (Lawton 1983). The four neighbourhoods created harboured vastly different assemblages of invertebrates, resulting from different physical microenvironments or different food availability. The fact that plants can modify the environment is well established (Jones et al. 1994). The effects of competition from A. pycnantha were minimal in comparison with the negative effects of the grasses, and the correlation of the distribution and abundance of invertebrates with this pattern further supports the argument that competition should be defined, or at least considered, in a phenomenological sense for the purposes of the habitat templet/C-S-R model.
Many studies have tested whether competitive intensity and resource availability are correlated. Here the intensity of competition could be related to the effects of fertility, or to the size and abundance (biomass) of the neighbouring vegetation. Although we did not harvest the competitors at the end of the experiment, it was clear that biomass of the four A. pycnantha individuals, which were over 2 m tall by the end of the experiment, dwarfed that of both T. triandra and weeds (Fig. 4). We therefore conclude that the extent to which neighbouring vegetation inhibited the growth of tree seedlings was related to architecture of the neighbouring vegetation rather than the quantity of biomass.
Figure 4. Photographs of field plots showing how biomass of Acacia pycnantha (top photograph, centre) dwarfs that of exotic pasture grasses (top photograph, right end) and Themeda triandra (bottom photograph) planted as biological neighbourhood. The plants had been established 2 years prior to the introduction of the target species.
Download figure to PowerPoint
Our results are consistent with a recent meta-analysis (Langellotto & Denno 2004) suggesting that reductions in habitat complexity may cause a large decrease in natural enemy abundance. While litter has been implicated as one of the principal causes of this pattern, we observed very little litter accumulation. The high abundance of invertebrate herbivores in the plots with exotic grasses and T. triandra may have been attributable to several factors, including better protection against harsh, desiccating physical conditions, reduced abundances of invertebrate predators (e.g. ants and spiders) and/or a reduction in the foraging efficiency of predators such as birds (see Groner & Ayal 2001). It should also be noted that there was a low abundance of collembolans in the plots with exotic grasses and T. triandra. Collembola may be an important component in the diet of ants (Wilson 1959). Their relatively high abundance in the plots with A. pycnantha and no vegetation may have sustained the large number of predatory ants found there. It is also possible that higher C : N ratios in grasses (Elser et al. 2000) affected the growth efficiency of herbivores, resulting in low levels of predation.
The argument that habitats with a high abundance of resources should sustain high levels of competition was supported both in the glasshouse, where we were able to study competitive intensity as a function of resource availability without the natural variability characteristic of ecological systems, and in the field, although the effects of resource competition and invertebrate herbivory were then heavily confounded. We therefore conclude that a phenomenological definition of competition is most tractable and also suggest that habitat complexity should be the focus of future research, as it may provide insight into the durational stability (disturbance) component of the habitat templet/C-S-R model.