Dynamic trajectories of growth and nitrogen capture by competing plants


Author for correspondence:
David Robinson
Tel: +44 1224 273662
Email: david.robinson@abdn.ac.uk


  • Although dynamic, plant competition is usually estimated as biomass differences at a single, arbitrary time; resource capture is rarely measured. This restricted approach perpetuates uncertainty. To address this problem, we characterized the competitive dynamics of Dactylis glomerata and Plantago lanceolata as continuous trajectories of biomass production and nitrogen (N) capture.
  • Plants were grown together or in isolation. Biomass and N content were measured at 17 harvests up to 76 d after sowing. Data were fitted to logistic models to derive instantaneous growth and N capture rates.
  • Plantago lanceolata was initially more competitive in terms of cumulative growth and N capture, but D. glomerata was eventually superior. Neighbours reduced maximum biomass, but influenced both maximum N capture and its rate constant. Timings of maximal instantaneous growth and N capture rates were similar between species when they were isolated, but separated by 16 d when they were competing, corresponding to a temporal convergence in maximum growth and N capture rates in each species. Plants processed N and produced biomass differently when they competed.
  • Biomass and N capture trajectories demonstrated that competitive outcomes depend crucially on when and how ‘competition’ is measured. This potentially compromises the interpretation of conventional competition experiments.


The prominence of competition as a central concept in plant ecology, and the unresolved debates that hinge on it, argue the need for a particularly deep understanding of how it works. But despite decades of research, that understanding remains elusive. One reason for this is that, although plants’ competitive interactions are dynamic, as evidenced by the vast literature on density-dependent self-thinning trajectories in communities (Harper, 1977), they are usually studied experimentally as if they are not. This perpetuates uncertainty and confusion about the process of competition and its ecological role (Grime, 1979, 2001; Tilman, 1982; Goldberg, 1990; Craine, 2005).

Competition is a multi-faceted concept, but this is not necessarily problematic provided that the facet of competition being examined is clearly defined for a particular study. Here, ‘competition’ is defined strictly as ‘the tendency of neighbouring plants to utilize the same quantum of light, ion of mineral nutrient, molecule of water, or volume of space’ (Grime, 1979, p. 8). This definition restricts itself to resource capture by two or more neighbours. It deliberately avoids saying anything about possible demographic consequences of competition. The influence of competition sensu Grime on population dynamics and community structure cannot yet be tested unambiguously. At that scale, ‘competition’ can mean anything from resource capture by neighbouring individuals, as above, through to the eventual extinction of one population by another. Here we do not use ‘competition’ to mean anything other than resource capture by neighbouring plants.

On that basis, biomass production by neighbouring plants is strictly an outcome of competition (and other processes), but one that obviously has a close functional relationship with resource capture. Put simply, bigger plants tend to capture more resources (i.e. are potentially stronger competitors) than smaller ones (Berntson & Wayne, 2000). We argue that, to fully understand plant competition as a process, it is necessary to perform experiments that measure both biomass production and resource capture by the competitors, and their functional interactions. Such experiments are conspicuous by their scarcity in the plant competition literature.

Other than those on self-thinning, plant competition experiments typically involve measuring biomass differences between neighbours at a single harvest (Connolly et al., 1990; Gibson et al., 1999; Damgaard et al., 2002). Explaining how and to what extent neighbours influence those differences is what most plant competition experiments aim to do. But such experiments are notoriously bad at doing this. This is because the biomass of a plant measured at one harvest is an outcome of the physiological and developmental processes that occurred throughout its life up to that point. Consequently, biomass differences between neighbours measured at one time cannot be explained fully without knowledge of the earlier, but unmeasured, processes that caused them. Explicitly dynamic perspectives on plant competition have been advocated by many modellers (e.g. Mutsaers, 1991; Damgaard et al., 2002; Dormann & Roxburgh, 2005), but investigated by relatively few experimentalists (Ross & Harper, 1970; Connolly et al., 1990; Robinson et al., 1999, 2010; Andersen et al., 2007; Aikio et al., 2009). Over a decade ago, the exclusive reliance on ‘final yield’ experiments to characterize plants’ competitive interactions was described by Gibson et al. (1999) as ‘…possibly the single most neglected and important issue in current practice’ and that criticism remains a valid one.

In the context of our definition of ‘competition’, it is clearly necessary to know how well an individual and its neighbours captured resources, how effectively those resources were converted into biomass, and how that biomass was then used to capture more resources, and so on over a defined period of the plant’s life. Such information would reflect the functional interdependence between growth and resource capture, the temporal dynamics of which are intrinsic to how these processes operate and which might contribute to eventual demographic success. In practice, characterizing the dynamics of plants’ competitive interactions demands measuring simultaneous changes in both resource capture and biomass production by competing individuals, preferably before and during a period when they would be likely to interact with one another, not just at one point in time. Surprisingly, this seems never to have been done in a way that allows those dynamics to be characterized reliably, which is why this was the first aim of this study.

Just as the biomass measured at one harvest is the legacy of processes that preceded the harvest, so a plant’s resource capture is the accumulation of the resources captured up to the time of measurement. Net carbon and water capture could, in principle, be estimated nondestructively using continuous gas exchange systems, although partitioning of the total fluxes among competing individuals is technically difficult. Similarly, partitioning radiation interception among the canopies of competitors can be done, but with some difficulty (Beyschlag et al., 1990). However, for nutrients, net capture can be approximated most easily by determining the nutrient content of harvested plants, ignoring the relatively small losses (in young healthy plants, at least) via root efflux, losses of volatiles or soluble fractions from leaves, and herbivory. Given the relatively tight stoichiometric relations between elements in vegetation (Elser et al., 2010), it might be expected that temporal changes in resource accumulation would simply track those of biomass production, and that competition would have only a minor effect on those relationships, but that expectation has yet to be tested. Here we carry out such a test for the competitive capture of one essential resource, nitrogen (N), and that was the second aim of this study.

Rates of nutrient and water uptake and of photosynthesis change continually in response to environmental cues and internal signals (Robinson, 2005). The value of measuring instantaneous rates of resource capture among competitors, as opposed to cumulative resource capture, is that instantaneous rates reflect how plants are interacting with each other at the time when the measurement is made. Instantaneous rates reveal the simultaneous resource fluxes into each competitor. Such fluxes, if measured, would provide the closest approximations to and most unambiguous information about the actual competitive process. Instantaneous rates can also indicate periods when competitive interactions are at their most intense and potentially most decisive in influencing the competitors’ success. Cumulative biomass production and resource capture are the integrals of many instantaneous rates. Estimating those instantaneous rates repeatedly would reveal the temporal progression of those interactions as they unfold, providing a direct link between the process of competition and the physiological activities of the competing individuals. We argue that the most appropriate way to explore the dynamics of plants’ competitive interactions is in terms of measuring resource capture and growth by neighbouring individuals both cumulatively and instantaneously. Some information exists about how the growth rates of competing plants co-vary (Connolly et al., 1990). However, the temporal resolution of such data is poor because experiments are restricted at best to only a few harvests widely spaced in time (e.g. four harvests at intervals of 13–20 d up to 69 d after sowing in the experiment reported by Connolly et al. (1990); five harvests at intervals of 9–35 d up to 112 d after sowing in that by Andersen et al. (2007)) or many harvests over a very short period (e.g. eight harvests over only 21 d: Ross & Harper (1970)). Even that level of information appears to be lacking for competitive resource capture. By characterizing trajectories of biomass production and resource capture by competing plants we aimed, finally, to assess if interpretations of traditional, single-harvest competition experiments are potentially compromised if temporal dynamics are ignored.

We achieved these aims by statistically fitting an easily parameterized phenomenological model to measurements of cumulative biomass production and net N capture by competing and isolated plants (Robinson et al., 2010). From that model, we derived instantaneous rates of those processes at daily temporal resolution. This approach contrasts with the use of mechanistic models of plant competition (Baldwin, 1976; Huston & DeAngelis, 1994; Thornley et al., 1995; Schwinning & Parsons, 1996; Biondini, 2001; Raynaud & Leadley, 2004). Such models require the estimation of numerous parameters that cannot be measured routinely in any one experiment and are often assembled piecemeal from many experiments, some of which might involve no plant competition. This limits the extent to which mechanistic modelling is practical for the routine analysis of competitive interactions (Mutsaers, 1991), for which simple phenomenological models are an ideal alternative.

Materials and Methods


Full details are given by Robinson et al. (2010). Briefly, Dactylis glomerata L. (Poaceae) and Plantago lanceolata L. (Plantaginaceae) were grown in a glasshouse, in pots (10 cm diameter) containing low-fertility agricultural soil (sandy loam, pH 5.5) using supplementary lighting to provide photosynthetically active radiation (PAR) at 27 W m−2 at soil level over a 12-h day. Plants either grew in isolation (one plant per pot) or with an individual of the other species (two plants per pot). Pots were watered regularly with a general fertilizer solution. There were 20 destructive harvests, taken every 3 or 4 d from 10 to 76 d after sowing. Each plant combination was replicated three times. The relatively low replication at each harvest was a consequence of the need in this experiment to accommodate an unusually large number of harvests which allowed information to be obtained at a temporal resolution sufficiently fine for the reliable estimation of instantaneous rates. At each harvest, shoots were removed by cutting at the soil surface. Harvested material was oven-dried at 80°C for 48 h before weighing. Harvested material was ground to a powder before determination of total N content by dry combustion to N2 (NA 1500 NCS Analyzer, Fisons, UK). There was sufficient shoot material for N analysis from the fourth harvest onwards. N data span 17 harvests, from 20 to 76 d after sowing. Root material was also sampled and its biomass determined and partitioned among competitors by statistical modelling, as described elsewhere (Robinson et al., 2010). Data reported here refer exclusively to aboveground parts.

This experiment was not designed to simulate the conditions under which D. glomerata and P. lanceolata would necessarily compete and coexist in the field, but to allow the initial competitive dynamics of these plants to be characterized under defined conditions. Nor was the duration of the experiment meant to reflect lifetime competitive dynamics. Seventy-six days is a short time relative to the potential lifespans of these perennial species, but this brevity is by no means exceptional by the standards of many published competition experiments involving glasshouse-grown herbaceous plants. The important point is that this experiment was long enough, and the harvests sufficiently frequent, to allow competitive dynamics to be visualized in detail. Larger and longer experiments are needed before the full range of those dynamics, and their ecological relevance, can be understood.

Data analysis

The biomass (Y; mg) of a herbaceous plant grown from seed for several weeks is typically a sigmoid function of time, t (Damgaard et al., 2002; Yin et al., 2003). For each daily interval, Δt, Y was computed using a discrete two-parameter form of the logistic equation (Robinson, 2001):


(r, the rate constant (d−1) for per capita biomass production; Ymax, a maximum to which Y tends asymptotically.) Models other than a logistic can be used to suit particular circumstances (Hunt, 1982) provided that they generate time-specific estimates of Y from which instantaneous per capita rates can be derived. An advantage of Eqn 1 over a cubic polynomial, for example, is that its parameters have obvious physiological interpretations.

Eqn 1 was also used to describe temporal changes in shoot N content as a proxy for net N capture, r and Ymax then representing the rate constant for per capita N capture and the maximum shoot N content, respectively. For direct comparability with the vast majority of plant competition experiments, only shoot data were used here. Almost all plant competition experiments are similarly restricted to using only aboveground data even though it is recognized that important interactions that influence competition do occur belowground. That is a further reason for estimating the capture of soil resources in competition experiments in addition to biomass, even if capture is inevitably underestimated. Ideally, N capture would be estimated using both shoot and root N contents. The problems of isolating the intermingled roots of competitors to provide material for reliable chemical analysis remain formidable (Cahill, 2002; Zobel & Zobel, 2002), although the root masses of competitors can be partitioned using biochemical or statistical approaches (Robinson et al., 2010). N capture would also, ideally, be measured as gross rather than net N fluxes from soil to plant. This can be done using 15N-pool-dilution approaches which we report elsewhere for a limited temporal window of competition between D. glomerata and P. lanceolata (Trinder et al., 2012). The logistical demands required to use 15N-pool dilution frequently enough to show detailed temporal trajectories of gross N capture by competing plants render that technique unsuitable for this purpose. The approach we describe here is the best compromise between what is ideal and what is practicable in order to reveal competitive dynamics, which was our primary aim.

Eqn 1 was fitted separately to data for shoot biomass and N content of each replicate in each treatment by simultaneously adjusting r and Ymax to maximize R2 (Brown, 2001). R2 for shoot biomass ranged from 0.817 to 0.981 and that for N content from 0.764 to 0.912 (see Supporting Information Figs S1, S2). The parameter estimates obtained for each replicate were used in linear models (R Development Core Team, 2008) to test for effects of species (i.e. D. glomerata and P. lanceolata) and competition (i.e. isolated vs competing plants) on r and Ymax for cumulative biomass production or N content. Mean parameter estimates (Table 1) were used to construct logistic curves to describe temporal changes in mean cumulative biomass or N content (Fig. 1).

Table 1.   Mean values of the parameters in Eqn 1 fitted to temporal changes in per capita cumulative biomass production and cumulative net nitrogen (N) capture by Dactylis glomerata and Plantago lanceolata growing in isolation or when competing (Fig. 1)
ParameterSpeciesLinear models
Dactylis glomerataPlantago lanceolataSpeciesCompetitionSpecies × competitionR2
  1. A logistic model (Eqn 1) was fitted separately to each replicate by simultaneously adjusting r and Ymax to maximize R2 between data and model. Logistic models accounted for 82–98% of the variation in biomass data and for 76–91% of the variation in N capture data: see Figs S1 and S2. r, rate constant of per capita biomass production or net N capture; Ymax, maximum attainable per capita biomass production or net N capture. Parameter estimates are means ± SE (= 3). Linear models show the effects on each parameter of species (i.e. D. glomerata vs P. lanceolata) and competition (i.e. isolated vs competing plants) in terms of t-statistics and probabilities. Tests for which  0.05 are shown in bold. R2 indicates the fraction of overall variance accounted for by each linear model. Where no interactions are shown, > 0.05, and these were removed from the final linear models.

Biomassr (d−1)0.165 ± 0.010.150 ± 0.010.138 ± 0.010.138 ± 0.01−  0.38
Ymax (mg)4 653 ± 9433 360 ± 97.23 519 ± 5441 497 ± 149− 2.550.032.830.02  0.61
Nr (d−1)0.175 ± 0.010.097 ± 0.0050.152 ± 0.020.104 ± 0.01− 0.760.475.81< 0.001  0.79
Ymax (mg)64.9 ± 4.8293.1 ± 8.3155.3 ± 4.6347.8 ± 3.70− 5.68< 0.001− 3.530.0083.160.0130.82
Figure 1.

Cumulative per capita shoot biomass production by (a) Dactylis glomerata and (b) Plantago lanceolata. Open circles and dashed curves, isolated plants; closed symbols and continuous curves, competing plants. Each symbol represents a single harvest and is the mean of three replicates. Curves are Eqn 1 derived using mean parameter estimates (Table 1, which also shows statistical analyses of those estimates). (c, d) Corresponding data for cumulative nitrogen (N) capture.

It is not usually possible to measure instantaneous per capita rates of biomass production or N capture directly at such a high frequency and with sufficient replication to ensure that the data will faithfully reflect the underlying temporal trends (Hunt, 1982, p. 53). For that reason, instantaneous rates were derived as the slopes of the logistic models fitted to the cumulative biomass or N content data rather than directly from data. This is legitimate if Eqn 1 is a reasonable description of the trends in the data to which it is fitted, as judged by a high R2 value, and if successive harvests are closely spaced in time (Hughes & Freeman, 1967); both conditions were met in this experiment. This was done separately for each replicate. If cumulative processes are sigmoid functions of time, unimodal trajectories of the corresponding instantaneous rates follow automatically (Hunt, 1982, p. 127; Yin et al., 2003). Statistical comparisons were made using estimates for each replicate of the times at which the maximum instantaneous rate was attained (tmax) and the values of those maxima (Imax; Table 2). Mean values of r (Table 1) were used to construct curves to describe temporal changes in mean instantaneous rates of biomass production or N capture (Fig. 2).

Table 2.   Times (tmax) and rates (Imax) of maximum per capita instantaneous biomass production and nitrogen (N) capture by Dactylis glomerata and Plantago lanceolata growing in isolation or when competing (Fig. 2)
ParameterSpeciesLinear models
Dactylis glomerataPlantago lanceolataSpeciesCompetitionSpecies × competitionR2
  1. Values are means ± SE (= 3) derived from the logistic models fitted to cumulative biomass and N capture data (Fig. 1; Table 1). Linear models are as described for Table 1.

Biomasstmax (d)60.7 ± 2.1967.7 ± 1.7657.0 ± 2.0852.3 ± 1.20− 5.87< 0.001− 2.680.033.160.0140.75
Imax (mg d−1)173.3 ± 0.04120.8 ± 0.02114.2 ± 0.0160.4 ± 0.01− 1801< 0.0011567< 0.00126.5< 0.0011.00
Ntmax (d)49.7 ± 1.2065.0 ± 2.0042.3 ± 0.6748.0 ± 0.58− 9.64< 0.001− 8.69< 0.0013.880.0050.94
Imax (mg d−1)2.80 ± 0.0012.00 ± 0.0012.00 ± 0.0011.21 ± 0.001− 643< 0.001659< 0.001− 3.390.0101.00
Figure 2.

(a) Mean instantaneous per capita rates of shoot biomass production in Dactylis glomerata (red) and Plantago lanceolata (black) derived as the slopes of the logistic models fitted to cumulative data, derived using mean values of the rate constant, r (Table 1). Dashed lines, isolated plants; continuous lines, competing plants. (b) Corresponding data for instantaneous nitrogen (N) capture. Statistical analyses of maximum instantaneous rates and their timings are presented in Table 2.


Cumulative biomass production and N capture

The presence of a neighbour had no effect on the rate constant (r) for per capita biomass accumulation in D. glomerata or P. lanceolata, although there was a significant species difference in this parameter (Table 1). By contrast, maximum biomass (Ymax) was influenced strongly by the presence of a neighbour and it, too, differed between the species. Ymax in isolated D. glomerata was one-third larger than in isolated P. lanceolata. In competing plants, the maximum biomass of D. glomerata was more than double that of P. lanceolata by the end of the experiment. Competition reduced the maximum biomass of D. glomerata to c. 70% of that attained when growing alone. The corresponding reduction in competing P. lanceolata was more drastic, its maximum biomass being limited to less than half that when growing in isolation. Divergences in the biomass trajectories between competing and isolated plants were detectable in both species after c. 40 d, but indistinct before that time, especially in P. lanceolata (Fig. 1a,b).

In contrast to biomass production, the rate constant for cumulative N capture by both D. glomerata and P. lanceolata was reduced strongly by the presence of a neighbour (Table 1). Per capita N accumulation was 55% and 68% slower in competing D. glomerata and P. lanceolata, respectively, compared with isolated plants. There was no species difference in r for cumulative N capture. Maximum N capture (Ymax) was influenced by both species and competition. Ymax in D. glomerata increased in the presence of a neighbour, by almost 50%, whereas that in P. lanceolata decreased slightly, explaining the significant statistical interaction. Dactylis glomerata ultimately captured more N than P. lanceolata, whether growing together or alone. Again, the trajectories of cumulative N capture by competitors diverged gradually from those of isolated plants, becoming obvious for both species after c. 40 d (Fig. 1c,d). All subsequent analyses are derived from the logistic models fitted to the data and which are shown in Fig. 1.

Instantaneous biomass production and N capture

The presence of a neighbour delayed by 7 d the time (tmax) at which D. glomerata attained its maximum instantaneous per capita rate of biomass production (Imax) compared with isolated plants (Table 2; Fig. 2a). In P. lanceolata, competing plants attained their maximum rate 5 d before their isolated counterparts. By the time D. glomerata reached its peak biomass production rate of 121 mg d−1, that of its competitor had collapsed towards zero. Dactylis glomerata reached a maximum biomass production rate double that of its competitor, albeit 15 d after P. lanceolata attained its maximum. Neighbours reduced the maximum instantaneous rates of biomass production in both species, to 70% and 55% of the maxima achieved by isolated D. glomerata and P. lanceolata, respectively.

Neighbours delayed the time of maximum per capita N capture rate by 15 d in D. glomerata and 6 d in P. lanceolata compared with isolated plants (Table 2; Fig. 2b). The maximum N capture rates attained by both species were reduced by the presence of a neighbour, to 60–70% of those of isolated plants. As with biomass production, maximum instantaneous N capture in D. glomerata exceeded that in P. lanceolata, although the latter species attained its maximum rate 17 d before D. glomerata.

An unexpected effect of competition was to closely align the trajectories of biomass production and N capture in each species. This is illustrated by the times at which maximum instantaneous rates occurred compared with the isolated plants. In isolated D. glomerata, tmax for instantaneous biomass production and N capture were separated by 11 d; the corresponding separation in isolated P. lanceolata was 15 d (Table 2). But when these species competed, the temporal separation between tmax for instantaneous growth and that for N capture was reduced to only 2–4 d in both species.

Dynamic trajectories of biomass production and N capture

Cumulative biomass production and N capture trajectories were derived simply by plotting the fitted logistic models in Fig. 1 for the competing plants against one another. Initially, the cumulative production of biomass and capture of N in P. lanceolata were both greater than in D. glomerata in both isolated and competing plants (Fig. 3a,b). In competing plants, N accumulation by P. lanceolata exceeded that of D. glomerata for the first 62 d of the experiment. Thereafter, D. glomerata accumulated more N than P. lanceolata. Plantago lanceolata’s superiority in biomass accumulation lasted 4 d longer than it did for N accumulation, for 66 d. The general pattern of P. lanceolata’s early superiority being overtaken by D. glomerata occurred also in the isolated plants, although that point was reached 7 d earlier than when the plants competed.

Figure 3.

Trajectories of (a) cumulative biomass production and (b) cumulative nitrogen (N) capture by isolated and competing Dactylis glomerata and Plantago lanceolata. Open circles and dashed curves, isolated plants; closed symbols and continuous curves, competing plants. Curves are those in Fig. 1 for D. glomerata and P. lanceolata plotted against one another. (c, d) Corresponding trajectories of instantaneous rates of biomass production and N capture. Curves are those in Fig. 2 for D. glomerata and P. lanceolata plotted against one another. Diagonal lines show the 1 : 1 relationships between D. glomerata and P. lanceolata. Symbols represent the times of harvests (times (d) shown next to selected symbols for reference).

The initial superiority of P. lanceolata was also evident in the trajectories of their instantaneous rates of biomass production and N capture (Fig. 3c,d). Until 54 d, P. lanceolata produced biomass faster than the D. glomerata plants with which it competed, although by then the instantaneous production rate of P. lanceolata had peaked and was in decline (Fig. 2). That of D. glomerata was still increasing and continued to do so until 68 d (Table 2), after which it too declined and was heading towards zero when the experiment ended. The isolated-plant trajectories were initially similar to those of the competing plants, as might be expected, although the former attained faster maximum rates than did the competitors.

The corresponding trajectory of competitive instantaneous N capture was similar to that for biomass, except that the point at which N capture rate in D. glomerata exceeded that in P. lanceolata occurred at 50 d, 4 d earlier than for biomass production. The competitive trajectory was enclosed almost entirely by that of the isolated plants. From 59 d onwards, however, competitors captured N faster than the isolated plants whose N capture rates declined towards zero by the end of the experiment.

The co-trajectories of cumulative biomass production and N capture in isolated plants of both species were consistent with a simple dilution of N as biomass accumulated (Fig. 4a). Initially, N accumulation was relatively rapid when plants were small and N supply plentiful. That produced tissue N concentrations c. 5% of dry weight, as can be inferred from the slopes of the curves in Fig. 4(a) for isolated plants smaller than c. 500 mg. Thereafter, growth and N accumulation followed gradually saturating co-trajectories, and tissue N concentrations fell to c. 1–2% when plants reached 3000 mg in dry weight. By contrast, the co-trajectories of N capture and biomass production in competing plants were not consistent with a dilution process. Cumulative biomass production and N capture in competitors followed inverse sigmoid co-trajectories throughout.

Figure 4.

(a) Co-trajectories of per capita cumulative nitrogen (N) capture and cumulative biomass production in competing Dactylis glomerata (red) and Plantago lanceolata (black). Curves are those in Fig. 1. Arrows indicate the directions of the trajectories. (b) Corresponding co-trajectories of per capita instantaneous rates of N capture and biomass production from Fig. 2.

Instantaneous rates of biomass production and N capture in competing D. glomerata and P. lanceolata were more tightly coupled than when plants grew alone (Fig. 4b). The co-trajectories of instantaneous production and N capture rates were relatively loose in both species. The curves looped widely through the N-biomass space, and their maxima on the N and biomass axes did not coincide. Those of competing plants were more tightly coupled, occupying a much narrower fraction of the N-biomass space. The maxima of the curves coincided more closely than did those for the isolated plants, and of course reflect the temporal coincidence seen in Fig. 2. But expressing the curves in Fig. 2 as co-trajectories emphasizes the strong effects that neighbours had on the dynamics of resource capture and growth compared with the corresponding dynamics in isolated plants.


This simple experiment has yielded new information about the dynamics of plant resource competition. This was achieved by: (1) measuring biomass and N contents of shoot tissues of competing and noncompeting plants at many frequent harvests; (2) fitting to the data a general phenomenological model to describe cumulative resource capture and biomass production as a function of time; (3) deriving from the model instantaneous per capita rates of net N capture and biomass production; (4) producing temporal trajectories of cumulative and instantaneous biomass production and N capture to determine how these processes co-varied temporally in competing individuals.

Using this approach, we were able to visualize how the competitive interaction between neighbouring plants unfolded (Fig. 3d). The trajectories of instantaneous N capture rate by competing plants were largely distinct from those of the isolated plants, indicating a strong neighbour effect on instantaneous N capture at all but the earliest harvests. This effect was predictable given the time needed for leaves of one plant to over-top those of a neighbour and for concentration depletion zones around the roots of adjacent plants to overlap as root densities in soil increase (Newman & Andrews, 1973; Baldwin, 1975; Newman, 1983). The contest for N between P. lanceolata and D. glomerata shifted gradually in favour of the latter species. Robinson et al. (2010) attributed the ultimately superior growth of D. glomerata in this experiment to a relentless increase in its root : shoot ratio while that of the P. lanceolata with which it grew remained relatively fixed. The strong allocation response by D. glomerata to P. lanceolata would be circumstantial evidence for a superior competitive capture of N and other nutrients by D. glomerata (Freckleton & Watkinson, 2001). That suspicion was confirmed by the direct evidence of greater N capture by D. glomerata (Fig. 3). Progressively larger investment in roots by D. glomerata would have allowed it to gradually attain faster rates of N capture than P. lanceolata, and that is indeed what happened. Greater N capture would have resulted in more biomass production by D. glomerata, and this again is supported by the data.

It is notable that the point at which D. glomerata overtook P. lanceolata in terms of cumulative or instantaneous N capture preceded by 3–4 d the point at which it gained the advantage in terms of biomass production (Fig. 3). That is consistent with faster N capture by D. glomerata leading to greater biomass production when growing with P. lanceolata. That simple cause-and-effect is unlikely to be the full story, however. The competitive capture of resources that were not measured (light, water and nutrients other than N) would also have influenced the capacity of D. glomerata to eventually grow better than its neighbour, but those influences cannot be evaluated using data from this experiment. Although we focused on only one of the resources for which plants can compete, the competitive dynamics of N capture nevertheless illustrate how that process is linked to those of biomass production, and ultimately how one individual could gain a competitive advantage over a neighbour.

Instantaneous N capture by isolated D. glomerata and P. lanceolata followed a similar, but wider and more rapid, trajectory to that of the competitors, D. glomerata eventually overtaking P. lanceolata. This indicates that the N capture trajectory of the competing plants was determined not only by the nature of their interaction, but also by external conditions such as soil N availability, which would have affected the time-courses of growth and N capture by isolated plants as much as those of the competitors, and by phenological differences between D. glomerata and P. lanceolata. In the small pots in which the plants were grown, eventual exhaustion of available N was possible, even with regular fertilizer additions during the experiment. That would have limited per capita rates of N capture. This is supported by the near-zero instantaneous N capture rates of the isolated plants of both species at the later harvests compared with their rapid rates at c. 40–50 d, a pattern typical of plants that are depleting their available N supply (Van Vuuren et al., 1996). Therefore, the characteristic ‘footprint’ of competitive dynamics is not solely the trajectory of instantaneous resource capture by neighbouring plants, but the comparison of that trajectory with that of isolated plants, as in Fig. 3(d).

It would be instructive to test the extent to which resource capture trajectories such as those in Fig. 3(d) can be modified by intermittent additions of N and other resources during competition. Campbell & Grime (1989) demonstrated the potential importance of the frequency, duration and predictability of nutrient pulses to the growth rates of two perennial grasses, Festuca ovina and Arrhenatherum elatius, with contrasting ecological distributions. In the British Isles, F. ovina is a slow-growing dominant of low-productivity, nutrient-poor grasslands, and A. elatius a large, fast-growing species characteristic of productive, nutrient-rich habitats (Grime et al., 1987). F. ovina grew faster than A. elatius when nutrient pulses were short, infrequent and unpredictable, matching, presumably, the nutrient-supply characteristics of undisturbed, infertile soils. The position was reversed when pulses were long and predictable, as would be expected of fertile soils. The implications of this for understanding how neighbours respond to resource supplies in different habitats are obvious, but direct tests of how temporal or spatial heterogeneity in resource supply influences competitive dynamics among co-existing plants have yet to be performed.

If it were possible to combine Campbell & Grime’s (1989) approach with that used here, it is easy to envisage that the resulting trajectories of instantaneous N capture could be very different from those in Fig. 3(d). We can predict that if a pulse of N were to become available when the plants’ capacities to take up N were constrained by the initial supply being close to exhaustion, the N capture trajectories would again increase, instead of heading towards zero as happened towards the end of this experiment. The resulting effects of such a response of instantaneous N capture on N accumulation and biomass production would depend on the relative and absolute sizes of the competitors at the time when the N pulse occurred, and on their capacities to use newly available N to produce new biomass. The competitive trajectories presented in Fig. 3(d) can reflect only a small range of the rich and varied dynamics that neighbouring plants in real communities might express. Because plant size changes with time, and per capita resource capture is size-dependent (Berntson & Wayne, 2000), the competitive dynamics of plants are necessarily size- as well as time-dependent. Size, not time, is probably the more important determinant of plants’ short-term competitive dynamics, but one that is itself dependent on those dynamics, a co-dependence illustrated by the trajectories in Fig. 4.

The implications of Fig. 3 for traditional plant competition experiments are serious. Comparing biomass produced or N accumulated by competitors and isolated plants at only a single point in time is clearly inadequate to fully explore the scale and nature of neighbours’ interactions. Whether comparing cumulative biomass production or N capture, or their instantaneous equivalents, different quantitative and qualitative outcomes would be obtained depending on when the comparison was made; this is also clear from the arable crop data of Andersen et al. (2007) and even from the 21-d pot experiment on D. glomerata reported by Ross & Harper (1970). Any arbitrarily chosen harvest along the trajectories shown in Fig. 3 would generate estimates of ‘competitive effect’ (Weigelt & Jolliffe, 2003) or ‘competitive strength’ (Andersen et al., 2007) very different from those produced at earlier or later harvests; even determining which competitor ‘wins’ is strongly time-dependent, as Fig. 3 demonstrates. It is impossible to estimate how many traditional, single-harvest, plant competition experiments have been influenced by such an artefact of sampling, but given their ubiquity and the temporally dynamic nature of resource capture and biomass production, the most likely answer is ‘all of them’. Single-harvest comparisons provide not merely snapshots of plants’ dynamic interactions, but potentially misleading ones. This has been acknowledged for a long time (Newman, 1983; Connolly et al., 1990, 2001; Gibson et al., 1999), but the scale of the problem has remained unknown. It is perhaps unsurprising that different single-harvest experimental studies of plant competition can give contradictory or inexplicable results when plants are sampled at arbitrary points along unknown dynamic trajectories. Small differences in initial conditions, such as in seedling size, from one competition experiment to another can also generate unexpected outcomes at subsequent harvests (Andersen et al., 2007).

Fig. 3(d) also reveals the striking possibility of an experiment falsely providing evidence for facilitation (Brooker et al., 2008), rather than competition. Had instantaneous N capture been measured only at 59 d or later, rates of N capture by competitors would have exceeded those by isolated plants. Such an effect, if genuine, would be difficult to explain mechanistically, but we can see that it was an artefact of the different temporal trajectories of competitors and isolated plants. Because the trajectories of isolated plants proceeded faster than those of the competitors, so that the former probably exhausted available N before the latter, more N would have been available to the competitors in the later stages of the experiment, allowing those plants to sustain faster N capture rates at those times. But that explanation can be suggested only because the trajectories were known. This problem of ‘pseudo-facilitation’ would not have arisen in this experiment had the comparison been based on biomass, but there is no certainty that that would always be the case.

The most surprising information to emerge from this experiment was the extent to which the dynamics of N capture and biomass production were modified by neighbours, compared with the corresponding dynamics of isolated plants. The presence of a neighbour induced a wider separation of the times of maximum instantaneous biomass production and N capture rates compared with those in isolated plants (Fig. 2). This could be interpreted as an example of temporal niche separation (Bretagnolle & Thompson, 1996), by which the activities of co-existing individuals in exploiting available resources are adjusted to occur partly at different times, with the potential effect of minimizing competition between neighbours and, hence, the potentially negative effects of one individual on another. But it is more probable that this effect was the secondary consequence of the closer temporal coupling between maximum biomass production and N capture rates, albeit at different times in each species: 48–52 d in P. lanceolata; 65–68 d in D. glomerata.

The mechanism by which that effect arose is unknown, but presumably it would have involved neighbour-detection processes (Karban, 2008; Cahill & McNickle, 2011). These include red:far-red radiation reflectance from, and transmission through, the leaves of a neighbour (Smith, 2000) and rhizosphere signalling to detect the presence of roots of another individual (Bais et al., 2006). Irrespective of the detailed signal reception and transduction pathways involved, the gross effect in this experiment was to more closely align the dynamics of instantaneous rates of biomass production and N capture (Fig. 4b). The corresponding trajectories of cumulative N capture and biomass production (Fig. 4a) were not those expected of a progressive dilution of captured N by the subsequent production of new biomass, as was seen in the isolated plants. Therefore, it seems likely that, when D. glomerata and P. lanceolata competed in this experiment, the plants used captured N differently from when they grew in isolation. One possibility is that less N was diverted to internal storage pools in competitors than in isolated plants and was instead used immediately by the competitors to produce more biomass (Lemaire & Millard, 1999), as if there was a greater imperative for immediate N use when that N was captured competitively rather than when it was acquired free of that constraint. In the case of D. glomerata, new biomass produced from the competitive capture of N and other resources was increasingly allocated to roots (Robinson et al., 2010). This would have facilitated the capture of more N and other nutrients. This suggests a positive feedback between resource capture and growth by D. glomerata that was decisive in determining its eventual dominance over P. lanceolata in the experiment. That such a response was confined to competing plants indicates the physiological and developmental flexibility with which plants can respond to the presence of a neighbour.

A second unexpected result was the different statistical effect of neighbours on cumulative biomass production and N capture (Table 1). Neighbours influenced only the ultimate amount of biomass produced (Ymax), not its rate constant (r), a result confirming that found by Damgaard & Weiner (2008) for the growth of Chenopodium album individuals in dense monospecific stands. By contrast, neighbours influenced both Ymax and r for cumulative N capture by both D. glomerata and P. lanceolata. Testing the generality of this result depends on further experiments on more species that compete under a wider range of conditions. Meanwhile, we note that measurements of biomass production and N capture can provide similar, but not identical, information about plants’ competitive interactions, as indicated by the analyses in Tables 1 and 2. As argued in the Introduction, experimental investigations of plant competition should measure resource capture by neighbours, biomass production being an outcome of competitive resource capture and of many other processes. But it is obviously vital to have information about both resource capture and biomass production as these processes interact so intricately, and iteratively, when individuals compete. Combining such information with measurements of resource availabilities (in the case of N, variations in soluble N production rates in the rooting volume) would allow direct tests to be made of Goldberg’s (1990) ideas about how plants change the availability of resources for which they compete and the resulting responses of plants to those changes (respectively, ‘competitive effects’ and ‘competitive responses’ in Goldberg’s terminology). Doing so would further improve our understanding of the competitive process.

There is a need to extend the approach described here to explore competitive dynamics over a larger fraction of plants’ lifespans, especially across successive growing seasons to account for phenological responses (Bretagnolle & Thompson, 1996) and regenerative processes (Grubb, 1977) under field conditions. There is evidence that the long-term population dynamics of some plant communities can be dominated by intraspecific, not interspecific, interactions between individuals (Rees et al., 1996). Therefore, a necessary development of the approach described here is to characterize the dynamics of intra- vs interspecific competition; this has recently been done for D. glomerata (M. C. Woo et al., unpublished data). It would also be valuable to know if, for example, seed production by competitors correlates with their resource capture during vegetative growth, and how competitive resource capture trajectories respond to local stochastic events (defoliation, disturbance, spatio-temporal fluctuations in resource availability, and climate), to temporal variations in population density, and to large disparities in neighbour size. Such information would allow plant community dynamics to be modelled in terms of local competitive processes (Travis et al., 2006; Berger et al., 2008). If competition does influence the structure and long-term dynamics of plant communities, then that influence will originate from the kinds of short-term competitive dynamics between individuals that are reported here.


This work was funded by the Natural Environment Research Council (NE/F004591/1), the James Hutton Institute and the Macaulay Development Trust. We thank the reviewers who made valuable suggestions that improved this paper.