Temperature-dependent interactions explain unexpected responses to environmental warming in communities of competitors


  • Lin Jiang,

    Corresponding author
    1. Department of Ecology, Evolution, and Natural Resources, Cook College, Rutgers University, 14 College Farm Road, New Brunswick, NJ 08901, USA
      and present address: Lin Jiang, Institute of Marine and Coastal Sciences, Rutgers University, 71 Dudley Road, New Brunswick, NJ 08901, USA. Tel: + 1 732 932 3209; Fax: + 1 732 932 8746; E-mail: ljiang@imcs.rutgers.edu
    Search for more papers by this author
  • Peter J. Morin

    1. Department of Ecology, Evolution, and Natural Resources, Cook College, Rutgers University, 14 College Farm Road, New Brunswick, NJ 08901, USA
    Search for more papers by this author

and present address: Lin Jiang, Institute of Marine and Coastal Sciences, Rutgers University, 71 Dudley Road, New Brunswick, NJ 08901, USA. Tel: + 1 732 932 3209; Fax: + 1 732 932 8746; E-mail: ljiang@imcs.rutgers.edu


  • 1Predictions about how species will respond to climate warming are based commonly on eco-physiological models or niche models. One critique of this practice is that it ignores interspecific interactions.
  • 2We examined how resource competition affects the responses of two ciliate species, Colpidium striatum Stein and Paramecium tetraurelia Sonneborn, to warming in laboratory microcosms.
  • 3We found that warming had a negative effect on Colpidium abundance and a negligible effect on Paramecium abundance in the absence of interspecific competition. When Colpidium and Paramecium competed, however, Paramecium coexisted with Colpidium at low and high temperatures, and was competitively excluded at intermediate temperatures.
  • 4Temperature-dependent strength of the interaction between Colpidium and Paramecium may explain the unexpected responses of Paramecium, which were inconsistent with its response to temperature in the absence of Colpidium.
  • 5Our results provide further support for the important role of species interactions in understanding how species respond to environmental warming, and suggest that the complex interplay between species interactions and climate may make it extremely difficult to predict accurately how species embedded in complex communities will respond to climate warming.


Increasing evidence suggests that recent climate warming has already affected various aspects of ecological communities, including organism phenology, species abundance and distribution, population dynamics and community- and ecosystem-level properties (reviewed in Hughes 2000; McCarty 2001; Stenseth et al. 2002; Walther et al. 2002; meta-analyses in Parmesan & Yohe 2003; Root et al. 2003). A persistent warming trend, driven largely by anthropogenic production of greenhouse gases, is projected to cause the global surface temperature to rise between 1·4 and 5·8 °C by the end of the 21st century (compared with 0·6 °C in the 20th century, IPCC 2001). How to predict accurately the responses of species and communities to rapid climate warming in the 21st century thus emerges as an important question.

Ecologists have commonly used climate mapping to predict how species and communities would respond to climate warming (e.g. Sutherst, Maywald & Skarratt 1995; Eaton & Scheller 1996; Iverson & Prasad 1998; Peterson & Vieglais 2001; Peterson 2001; Peterson et al. 2002). Climate mapping typically comprises two steps. First, species current abundances and distributions are correlated with climate factors across a geographical range. Secondly, under certain climate change scenarios, future abundances and distributions are projected based on the constructed relationship between climate and current abundance and distribution. The key assumption of climate mapping is therefore that physiology or, more directly, the fundamental niche, essentially determines species responses to climate change.

Climate mapping has been criticized for failing to incorporate the potentially important role of factors other than climate, particularly interactions within and among species, in determining species’ abundance and distribution (e.g. Davis et al. 1998a,b). Both theoretical (Ives & Gilchrist 1993; Ives 1995; Abrams 2002) and empirical studies (Brown, Valone & Curtain 1997; Davis et al. 1998a,b; Post et al. 1999) have suggested that interspecific interactions could alter species’ responses to climate change substantially. Models (Ives 1995), manipulative experiments (Fox & Morin 2001) and time-series analyses of natural population dynamics (reviewed in Stenseth et al. 2002) all indicate that intraspecific interactions could also influence species responses to climate change.

Another important aspect ignored by climate mapping is that species interactions may themselves be affected by climate change. Both the sign and intensity of interactions may be influenced by climate. For example, climate warming can affect the strength of intraspecific density dependence (Coulson et al. 2001; Stenseth et al. 2002) and the strength of predator–prey interactions (Post et al. 1999; Sanford 1999). Climate variation can also alter the nature of species interactions: neighbouring plants in New England salt marshes often showed negative competitive interactions in cold climates and positive facilitative interactions in warm climates (Bertness & Ewanchuk 2002). These climate-driven variations in interactions may translate into significant changes in population dynamics and community structure, making it more difficult to predict species and community responses to climate warming.

Here we report an experimental study demonstrating that temperature-dependent interactions may explain unexpected community responses to environmental warming. We show that temperature-dependent variation in interaction strength made it extremely difficult to predict responses to warming, even in a simple community where only two consumers engaged in consumptive resource competition. We conducted the experiments in aquatic microcosms containing assemblages of bacterivorous ciliates and their bacterial resources. We focused our attention on the responses of the two ciliates to warming. However, responses of bacterial resources were also explored to provide a mechanistic explanation for temperature-dependent interaction strength. The short generation times of the experimental ciliates (∼ 4–12 h) allowed us to collect long-term data (over 80 generations for the ciliate with the longest generation time) within the experimental period (44 days). Such long-term data are essential for evaluating competitive outcomes, as competitive exclusion may take many generations to occur.

Materials and methods

microscom setup and sampling

Our experimental organisms were two ciliate species: Colpidium striatum Stein (hereafter Colpidium) and Paramecium tetraurelia Sonneborn (hereafter Paramecium). Colpidium was obtained from Carolina Biological Supply Company (CBS) (Burlington, NC, USA), and Paramecium from the American Type Culture Collection (Rockville, MD, USA). Both are free-swimming bacterivores under the conditions used in this study. Before the experiments both species were maintained separately in stock cultures at 22 °C, where they fed on three bacterial species: Serratia marcescens Bizio, Bacillus cereus Frankland & Frankland and Bacillus subtilis[Ehrenberg] Cohn, all obtained from CBS.

Our general experimental protocol closely followed Lawler & Morin (1993). The experimental microcosms were 240 mL screw-capped glass bottles, each containing 100 mL of nutrient medium, plus two wheat seeds, which slowly released nutrients into the medium. We sterilized bottles and wheat seeds before setting up the experiments. We made up the medium by adding 0·55 g of CBS protozoan pellets in 1 L of well water in 2-L Erlenmeyer flasks. The medium was sterilized and then inoculated with three bacterial species: S. marcescens, B. cereus and B. subtilis. We distributed medium into experimental bottles 24 h after bacterial inoculation, at which time we introduced small numbers of the two ciliates into their designated microcosms. All microcosms were placed in incubators with lights turned off to prevent algal contamination.

To simulate environmental warming, we established a temperature gradient consisting of five constant temperatures (22, 24, 26, 28, and 30 °C), controlled by five independent incubators. Because the experimental organisms were maintained at the baseline temperature 22 °C before the experiment, temperatures higher than 22 °C represented different scenarios of environmental warming. This relatively large temperature gradient was chosen because species’ responses to warming are more likely to be detected with larger temperature change and because pilot experiments revealed that both Colpidium and Paramecium remained active even at the high end of this gradient.

Microcosms were sampled at designated intervals to estimate ciliate population densities. During sampling, a small sample of known volume (∼ 0·30 mL) was withdrawn from each microcosm, and the number of individuals of each ciliate was counted under a dissecting microscope. Dilution of samples was often necessary for counting Colpidium, which frequently became too dense to count in undiluted samples.

intrinsic growth rates experiment

We conducted a short-term experiment to measure intrinsic growth rates of Colpidium and Paramecium. For each ciliate species, we set up three replicates at each of the five temperatures, totaling 30 microcosms. We inoculated 100 individuals of each species into their designated microcosms and sampled densities every 12 h until populations no longer grew exponentially. The intrinsic growth rate of each species was calculated as the slope of the linear regression of natural logarithm of population densities vs. time.

main experiment

The main experiment explored how environmental warming affects the interaction between Colpidium and Paramecium. We set up three species treatments: two controls without interspecific competition (Colpidium alone and Paramecium alone), and a competition treatment in which Colpidium and Paramecium competed for bacteria (Colpidium–Paramecium together). All three species combinations were crossed with the five temperature treatments. We set up three and four replicates for the control and competition treatments, respectively, totalling 50 microcosms. The experiment ran for 44 days.

We sampled each microcosm every 2 days to gauge ciliate population dynamics. We replaced 10% of the medium in each microcosm with fresh sterile medium every week, to replenish nutrients and prevent metabolic waste build-up.

bacteria experiment

We also investigated the impacts of ciliate consumption, temperature, and their interaction on bacterial abundance and composition, providing insights into possible mechanisms responsible for different competitive outcomes between Colpidium and Paramecium along the temperature gradient. The experiment used three consumer treatments: bacteria alone, bacteria with Colpidium and bacteria with Paramecium crossed with five temperature treatments described above. We used two replicates for the bacteria alone treatment and three replicates for Colpidium and Paramecium treatments. Lower replication was used in the bacteria alone treatment because trial experiments showed that bacteria alone treatment exhibited smaller variation among replicates than the other two treatments.

We estimated bacterial abundances using plate counts of serially diluted sampled bacteria 1 week after microcosms were established, after ciliate densities stabilized. Plate counts allowed us to distinguish among different colony types. We classified bacteria into three colony types: red (probably Serratia), large white and small white colonies. We did not attempt to relate these colony types specifically to the three bacterial species inoculated at the start of the experiment, because their identities were not absolutely essential for the purposes of this study and because each colony type may also represent contaminants entering microcosms during the experiment.

estimation of interaction strength

Using the data from the main experiment we estimated per capita interaction strengths (competition coefficients) of Colpidium on Paramecium at each of the five temperatures. We did not estimate interaction strengths of Paramecium on Colpidium because Colpidium appeared to be unaffected by the presence of Paramecium (see Results). We first fitted continuous-time logistic models to Paramecium population dynamics in the controls without Colpidium to estimate carrying capacities at different temperatures, with measured intrinsic growth rates. We then fitted continuous-time Lotka–Volterra competition models to Paramecium population dynamics in the competition treatments to estimate competition coefficients for the per capita effects of Colpidium on Paramecium, using measured intrinsic growth rates and estimated carrying capacities. Linear least square fitting was used in both steps.


intrinsic growth rates

Temperature strongly influenced intrinsic growth rates of the two ciliates, but displayed opposite effects on Colpidium and Paramecium. Increasing temperature had a significant negative effect on the intrinsic growth rate of Colpidium (Fig. 1a; one-way anova, temperature: F4,10 = 30·65, P < 0·0001) and a significant positive effect on Paramecium growth rates (Fig. 1a; one-way anova, temperature: F4,10 = 6·93, P = 0·0081).

Figure 1.

The effect of temperature on the (a) intrinsic growth rate and (b) geometric mean density of Colpidium and Paramecium in the control (no interspecific competition between Colpidium and Paramecium). Error bars represent + SE.

main experiment

Colpidium and Paramecium persisted for the entire experimental period (44 days) in all microcosms without interspecific competitors, regardless of temperature. Increasing temperature significantly reduced the geometric mean of Colpidium density (Fig. 1b; one-way anova, temperature: F4,10 = 117·01, P < 0·0001), consistent with its negative effect on the intrinsic growth rate. In contrast, increasing temperature did not affect the geometric mean of Paramecium density (Fig. 1b; one-way anova, temperature: F4,10 = 1·37, P = 0·3106), despite its positive effect on Paramecium intrinsic growth rates.

Species’ responses in the competition treatment could not be readily predicted from single-species responses to warming (Fig. 2). At 22 °C Paramecium coexisted with Colpidium as the numerically subdominant species. At 24 °C Paramecium abundance declined steadily after initial increase, and was approaching extinction near the end of the experiment. At 26 and 28 °C Paramecium went extinct well before the experiment ended, with the extinction occurring more rapidly at 28 °C than at 26 °C. At 30 °C Paramecium again coexisted with Colpidium, but was numerically dominant in two replicates and codominant with Colpidium in the other two replicates toward the end of the experiment. Colpidium abundance declined with increasing temperature, but was not affected adversely by Paramecium.

Figure 2.

Population dynamics of Colpidium and Paramecium in the competition treatment (interspecific competition present between Colpidium and Paramecium) at 22, 24, 26, 28 and 30 °C. Each panel represents one of four possible replicates at each temperature.

Estimates of per capita competitive effects of Colpidium on Paramecium increased significantly with temperature, being more than five times larger at 30 °C than at 22 °C (Fig. 3a; one-way anova, temperature: F4,15 = 21·18, P < 0·0001). Population interaction strengths (collective population-level effect, see Navarrete & Menge 1996) of Colpidium on Paramecium, however, exhibited a humped pattern in relation to temperature (Fig. 3b; one-way anova, temperature: F4,15 = 8·16, P < 0·0011), due to the decline of Colpidium abundance with warming. The low population interaction strengths at 22 and 30 °C correspond to the coexistence of Colpidium and Paramecium observed at these two temperatures, and the large population interaction strengths at 26 and 28 °C correspond to the competitive exclusion of Paramecium observed at these two temperatures (Fig. 2).

Figure 3.

The effect of temperature on the (a) per capita interaction strength and (b) population interaction strength of Colpidium on Paramecium. Error bars represent + SE. Different letters indicate significant differences at P = 0·05 level in a Tukey HSD test.

bacteria experiment

Patterns in bacterial abundances differed substantially in the bacteria alone, Colpidium, and Paramecium treatments. In the bacteria alone treatment, red colonies dominated regardless of ambient temperature, and were typically more than one order of magnitude more abundant than large and small white colonies (Fig. 4). Red colony abundance declined slightly as temperature increased (Fig. 4a; one-way anova, temperature: F4,5 = 4·31, P = 0·0703), whereas large and small white colonies both significantly increased with temperature (Fig. 4b, one-way anova, temperature: F4,5 = 7·36, P = 0·0252 for large white colony; Fig. 4c, one-way anova, temperature: F4,5 = 69·49, P = 0·0001 for small white colony). Colpidium predation reduced red colony abundances significantly and increased both large white and small white colony abundances (Fig. 4a,b,c), presumably because Colpidium consumed primarily red colonies, benefiting large and small white colonies indirectly. The effect of Colpidium on red colonies appeared to be independent of temperature (two-way anova, predation × temperature: F4,15 = 0·27, P = 0·8944). In contrast, Paramecium reduced abundances of all three colony types and in fact eliminated small white colonies from all microcosms (Fig. 4a–c). In addition, the negative effect of Paramecium on red colonies appeared to be more pronounced at high temperatures (Fig. 4a; two-way anova, predation × temperature: F4,15 = 3·63, P = 0·0314).

Figure 4.

Responses of (a) red colonies, (b) large white colonies, (c) small white colonies and (d) all colonies in the bacteria alone, Colpidium and Paramecium treatments along the temperature gradient. Note that bacterial densities were recorded on a logarithm scale. Error bars represent + SE.


Our experimental system is far simpler than most natural communities, with only two consumers competing for three bacterial species. The results from our experiments nevertheless demonstrate the difficulty of predicting species’ responses to climate warming in situations affected by strong interspecific interactions. In the absence of interspecific competition, Paramecium's abundance was not affected by temperature and Colpidium's abundance was affected adversely by increasing temperature (Fig. 1). Based on this information, a climate-mapping approach might predict that Paramecium would increase its abundance relative to Colpidium as temperature increased. Instead, Paramecium coexisted with Colpidium at both ends of the temperature gradient (22 and 30 °C), and was competitively excluded by Colpidium at intermediate temperatures (26 and 28 °C).

Resource competition models predict that species with the minimal resource requirement (i.e. species that reduces resource to the lowest level) will exclude all other species (R* rule, Tilman 1982). Applying the R* rule to this study suggests that Paramecium would exclude Colpidium from 24 to 30 °C, because Paramecium reduced total bacterial abundance to a lower level than Colpidium at these temperatures (Fig. 4d). However, we observed that Colpidium and Paramecium coexisted at some temperatures and that Colpidium excluded Paramecium at other temperatures. We suspect that the failure of the R* rule to explain our results is due primarily to the differential use of heterogeneous bacterial resources by Colpidium and Paramecium. Colpidium can be considered as a specialist consumer and Paramecium a generalist consumer, because Colpidium only negatively affected the abundance of the red colony bacteria (which were probably Serratia marcescens), while Paramecium was able to reduce the abundances of all the three colony types (Fig. 4). Thus, resource differentiation appeared to be responsible for the possible coexistence between Colpidium and Paramecium. Substantial resource overlap, however, still existed between Colpidium and Paramecium because the red colony bacteria, the only resource clearly consumed by Colpidium, was also the primary food resource for Paramecium at all temperatures (Fig. 4a,b,c, note the log scale). Consequently, differential resource use by Colpidium and Paramecium was not always sufficient to maintain coexistence (Fig. 2).

The unexpected responses of Paramecium to warming in the presence of Colpidium could reflect temperature-dependent variation in the strength of interactions between Colpidium and Paramecium. Our estimates of per capita competitive effects increased continuously from 22 °C to 30 °C, whereas population-level interaction strength increased from 22 to 28 °C and declined from 28 °C to 30 °C (Fig. 3). A simple explanation for the temperature-dependent interactions is that as temperature increased, individuals of Colpidium must consume more resources to meet their elevated metabolic requirements, translating into larger per capita effects of Colpidium on Paramecium at higher temperatures. This explanation is supported, to some extent, by the data on bacteria. Despite the decline in Colpidium abundance with increasing temperature (Fig. 1), the abundance of its bacterial resource (i.e. red colony bacteria) was reduced to approximately the same level across the temperature gradient (Fig. 4a), suggesting that each individual of Colpidium consumed more bacteria at higher temperatures.

The observation that Colpidium reduced the abundance of red bacteria to a constant level, independent of temperature, does not directly explain why Paramecium coexisted with Colpidium at 22 and 30 °C, and was driven to extinction at 26 and 28 °C. Further inspection of the data on bacteria suggests that unexpected Paramecium extinction patterns are probably the result of two processes with opposing effects on species coexistence. On one hand, warming facilitated the coexistence of Paramecium with Colpidium, by increasing the abundance of alternate bacterial resources (i.e. bacteria with large white and small white colonies) for Paramecium (Fig. 4b,c). On the other hand, warming may have intensified competition between Colpidium and Paramecium and made coexistence more unlikely, because the red bacteria were more heavily exploited by Paramecium at higher temperature (Fig. 4a). The interaction between these two opposing forces may be behind Paramecium extinction patterns.

Our results reinforce the findings of some recent studies that support an important role of interspecific interactions in species and community responses to global warming. For example, Davis et al. (1998a,b) found that both interspecific competition and predation altered the responses of several Drosophila species to warming. Fox & Morin (2001) reported that intraspecific density dependence could buffer species against warming. Recent ecological time-series studies also suggest that observed dynamics of natural animal populations are often the consequence of both density dependence (both intra- and inter-) and climate variation (e.g. Sætre, Post & Kral 1999; Lima, Stenseth & Jaksic 2002; reviewed in Stenseth et al. 2002). This present study builds on the results of those previous studies and suggests that climate-dependent variation in interaction strength could also play important roles in species and community responses to climate warming.

The suggestion that interaction strength varies importantly with climate is emerging from studies of a variety of systems. For example, Sanford (1999) reported that changes in ocean temperature affected the strength of interactions between a keystone predator, the sea star Pisaster ochraceus, and its mussel prey Mytilus californianus, but he did not examine its community-wide consequences. Coulson et al. (2001) found that climate variation altered the strength of intraspecific density dependence in the Soay sheep (Ovis aries) population on the St Kilda archipelago, UK. Climate variation was reported to affect wolf-hunting behaviour and consequently wolf predation rate on herbivores in Isle Royale, USA (Post et al. 1999). Our study is unique in that we were able to quantify temperature-dependent interactions and to observe effects on responses to environmental warming. So far, most studies of climate-dependent interactions have focused on interactions within species (Coulson et al. 2001) or within simple communities (Post et al. 1999; Sanford 1999; this study). Climate-dependent interactions may have even stronger effects on more complex communities, especially if more complex systems are inherently less stable than simple communities (May 1973).

One important remaining caveat is that we estimated interaction strengths by fitting linear Lotka–Volterra models to our data. Some ecologists argue that species interactions are frequently non-linear functions of population densities and that consequently single measures of interaction strength may be of limited predictive value (e.g. Abrams 2001). We assumed linear species interactions in our experiments, both because we did not have the large amount of data required to fit nonlinear models, and because the population dynamics of competing protists can be approximated well by Lotka–Volterra models (Gause 1934; Vandermeer 1969).

Our results suggest that without knowledge of how interactions change with climate, it may be extremely difficult to predict how species respond to climate change. We encourage additional tests on the generality of our findings in various ecological settings. We expect climate-dependent interactions to be common, because interaction strengths depend on physiology and behaviour which are in turn influenced by climate. Accurately predicting how climate warming will affect species and communities will remain a daunting task, until the mechanistic basis for climate-dependent interactions is fully understood.


We thank Peter Abrams, Jennifer Adams, Brian Allan, Tim Casey, Christina Kaunzinger, Zac Long, Timon McPhearson, Peter Smouse, Chris Steiner and two anonymous referees for their comments on previous drafts of the manuscript. The project was funded by US NSF grant DEB-9806427 to Peter Morin and Tim Casey, and a Bevier dissertation fellowship awarded to Lin Jiang from Rutgers University.