Environmental Stress, Facilitation, Competition, and Coexistence

Ecology, 94(12), 2013, pp. 2719–2731 Ó 2013 by the Ecological Society of America

Environmental stress, facilitation, competition, and coexistence

1,3 1,2 SIMON P. HART AND DUSTIN J. MARSHALL 1School of Biological Sciences, University of Queensland, Queensland 4072 Australia 2School of Biological Sciences, Monash University, Clayton, Victoria 3800 Australia

Abstract. The major theories regarding the combined influence of the environment and species interactions on population and community dynamics appear to conflict. Stress/ disturbance gradient models of community organization, such as the stress gradient hypothesis, emphasize a diminished role for competition in harsh environments whereas modern coexistence theory does not. Confusion about the role of species interactions in harsh environments is perpetuated by a disconnect between population dynamics theory and data. We linked theory and data using response surface experiments done in the field to parameterize mathematical, population-dynamic competition models. We replicated our experiment across two environments that spanned a common and important environmental stress gradient for determining community structure in benthic marine systems. We generated quantitative estimates of the effects of environmental stress on population growth rates and the direction and strength of intra- and interspecific interactions within each environment. Our approach directly addressed a perpetual blind spot in this field by showing how the effects of competition can be intensified in stressful environments even though the apparent strength of competition remains unchanged. Furthermore, we showed how simultaneous, reciprocal competitive and facilitative effects can stabilize population dynamics in multispecies communities in stressful environments. Key words: Bugula neritina; competition; competition coefficients; environmental stress; facilitation; filter-feeding bryozoans; marine invertebrates; multispecies communities; Ricker model; stress gradient hypothesis; Watersipora subtorquata.

INTRODUCTION es with increasing environmental stress (Bruno et al. Because environmental variation is ubiquitous, un- 2003, Callaway 2007, Maestre et al. 2009). Empirical derstanding how changes in the environment influence support has been generated through estimates of the species interactions is crucial for predicting population strength of interspecific interactions between individuals and community dynamics. Many classic (e.g., Connell or surveys of community-wide patterns in the frequency 1978, Grime 1979, Huston 1979) and modern (e.g., of positive and negative interactions along environmen- derivatives of the Menge and Sutherland [1987] stress tal gradients (Callaway 2007, Maestre et al. 2009). This gradient hypothesis; Callaway 2007) theories of com- work has been useful for identifying common patterns in munity dynamics continue to emphasize a diminished the strength and direction (positive vs. negative) of role for competition in harsh environments, but modern species interactions among disparate communities, and species coexistence theory does not (Chesson and Huntly for highlighting the role of facilitation in community 1997, Chase et al. 2002, Violle et al. 2010, Fox 2012). dynamics (Stachowicz 2001, Bruno et al. 2003). How- Understanding how these theories can mutually inform ever, these conceptual theories and the empirical each other is therefore an important goal. methods used to support them are rarely suited to Stress gradient models (commonly called ‘‘stress understanding the consequences of changes in the gradient hypotheses,’’ SGH) of community organization strength and direction of species interactions for the have been important for simplifying the enormous regulation of populations, and therefore for species context dependency of species interactions across coexistence (Freckleton et al. 2009, Siepielski and environmental gradients (Menge and Sutherland 1987, McPeek 2010). For example, stress gradient hypotheses Bertness and Callaway 1994, Angelini et al. 2011). These are often framed in terms of the commonness of theories specifically predict that the frequency and/or particular types of interactions (Maestre et al. 2009), strength of competition declines and facilitation increas- but the commonness of an interaction type across an entire community provides little information on the Manuscript received 17 May 2012; revised 30 April 2013; density-dependent regulation of populations within accepted 11 June 2013. Corresponding Editor: L. Stone. those communities. Furthermore, intraspecific compet- 3 Present address: Institute for Integrative Biology, ETH itive effects are rarely estimated in these studies even Zu¨rich (Swiss Federal Institute of Technology), Universita¨t- strasse 16, 8092 Zu¨rich, Switzerland. though it is the ratio of the strength of population-level E-mail: [email protected] intra- to interspecific effects that determines the outcome 2719 2720 SIMON P. HART AND DUSTIN J. MARSHALL Ecology, Vol. 94, No. 12 of species interactions (Macarthur and Levins 1967, species interactions is a challenge that is rarely met Chesson 2000). empirically (Chesson and Huntly 1997, Freckleton et al. Modern coexistence theory, in contrast, tends to 2009, Siepielski and McPeek 2010). explain community structure as a consequence of the Approaches that allow formal links between mathe- influence of the environment on interspecific interac- matical population demographic theory and data are tions, but also on density-independent processes and ideally suited for quantifying the strength and direction intraspecific interactions (Chesson 2000, Snyder 2008, of species interactions and for determining how these Levine and HilleRisLambers 2009). These mathematical changes affect population dynamics (Inouye 2001, theories demonstrate that environmental gradients, Levine and HilleRisLambers 2009). However, such including stress and disturbance gradients, can promote approaches are rarely replicated across stress gradients. coexistence (Caceres 1997, Bolker and Pacala 1999, Using a classic model system for studying competition— Chesson 2003, Adler et al. 2006, Snyder 2008). benthic marine invertebrates—we replicate a powerful Furthermore, modern coexistence theory does not but rarely used experimental design in the field in two preclude changes in the strength, direction, or frequency strongly contrasting environments that span an impor- of particular types of interspecific interactions in tant natural stress gradient in marine systems. Using our different environments such as those suggested by stress experimental field data, we parameterize a mathematical gradient hypotheses (e.g., Sears and Chesson 2007). model of competitive population dynamics to provide However, coexistence theory provides no a priori quantitative estimates of the direct effects of environ- predictions about the outcome of those interactions mental conditions on density-independent rates of (Chase et al. 2002). In particular, the strength of an increase, and the indirect effects of the environment on interaction cannot be used to determine the importance population dynamics via changes in the direction and of an interaction for population regulation (Freckleton strength of both intra- and interspecific interaction et al. 2009). This is because the direct density- coefficients. We test the predictions of stress gradient independent effects of a harsh environment can make models by evaluating how the strength and direction of a population less tolerant of competition, which means species interactions vary depending on environmental that competitive exclusion can occur more readily in harshness, and we show that the consequences of these harsh environments even though the intensity of changes for competitive population dynamics are not competition may be weaker (Chesson and Huntly straightforward. 1997). Ultimately, coexistence depends on how the MATERIALS AND METHODS environment affects the relative strengths of both intra- and interspecific interactions, and the direct Assemblage, study species, and the environment density-independent effects of the environment on a We studied population dynamics in diverse, subtidal, species’ tolerance to competition (Chesson 2000). sessile marine invertebrate assemblages in which strong Importantly, independent changes in each of these interactions, both competitive and facilitative, are a variables depend strongly on species-specific responses distinctive feature (Buss 1990, Stachowicz and Byrnes to the environment (Angert et al. 2009), an area recently 2006). We were explicitly interested in interactions identified as being crucial for improving stress gradient among early-successional invasive species before they models of community organization (Maestre et al. were largely outcompeted by later-successional domi- 2009). nants (typically after four to six weeks). Therefore, we The foci and underlying motivations behind stress studied competition between two filter-feeding bryozo- gradient hypotheses (e.g., Callaway 2007) and modern, ans: an encrusting species, Watersipora subtorquata, and population dynamic coexistence theories (e.g., Chesson an arborescent species, Bugula neritina. Both species are 2000) are therefore different, but their predictions are invasive and widespread globally and often dominate not necessarily mutually exclusive. Where possible, the recruitment of recently exposed substrata at our field conclusions of these theories might mutually inform site, such that there is strong potential for interactions each other. One step toward this goal is to determine the between these species to be important determinants of consequences of individual-level changes in the direc- their success in early-successional assemblages. We tion, frequency, and strength of interspecific species quantified competitive population dynamics between a interactions (the domain of stress gradient theories of single cohort of genetic individuals and, consistent with community organization) for population dynamics (the the mathematical theory that we apply, our response domain of modern, population-dynamic, coexistence variable was the production of a new generation of theories). This requires quantification of the direct genetic individuals. Our approach was deliberately effects of the environment on density-independent phenomenological; our results include the effects of all processes, as well as the indirect effects of the mechanisms of competition/facilitation and all density- environment on populations as mediated through independent influences on our study species. changes in both intra- and interspecific interactions. We studied interactions in two habitats that encom- However, estimating both direct and indirect effects of pass one of the most important environmental gradients the environment on the population-level outcome of for determining community structure in marine systems: December 2013 STRESS, FACILITATION, AND COMPETITION 2721 vertical and horizontal (facedown) habitats (Miller and recruits commonly observed in the field (typically 0.25– Ron 2008). These habitats vary in several environmental 0.75 individuals/cm2; Appendices A and B). variables, including rates of sediment and UV exposure, In the field we attached and randomly arranged two which are two major sources of stress for many species full sets of the 16 density combinations (i.e., 2 3 16 of filter-feeding sessile invertebrates (Irving and Connell settlement plates) on each of four 800 3 800 mm, 6 mm 2002). Definitions of environmental stress are often thick, PVC backing panels. Backing panels were problematic because they rarely take into account suspended from floating pontoons at a depth of 1 m in species-specific responses to the environment (Maestre either a facedown horizontal (benign) or vertical et al. 2009). Therefore we defined stress in terms of the (stressful) orientation, with two backing panels in each direct response of our study species to the environmental orientation. In total, there were 2144 recruits divided by conditions (sensu Sears and Chesson 2007). Based on two species distributed across 128 settlement plates our observations of performance of Bugula and Water- across four backing panels. This arrangement meant sipora, we specified the vertical habitat as more stressful that there were two backing panels statistically nested a priori, which is also consistent with reduced develop- within each of the two environments such that there ment and lower productivity of assemblages on vertical were two replicates of each of our environment habitats at our field site. Our estimates of density- treatments. Within each replicate of the environment, independent performance (k0) were used to confirm that there were two replicates of each of the 16 density the vertical habitat was intrinsically stressful for both combinations (Appendix C). species. Additional information on the life histories of Competition between our focal species occurred in the our study species and the natural history and location of field where other biotic (e.g., settlement and growth of our study system is provided in Appendix A. other species, predation) and abiotic factors (e.g., temperature, disturbance, water flow) were allowed to Experimental methods vary naturally. We maintained densities of the focal We studied competitive population dynamics using a species by removing new recruits each week. After one month in the field, we recorded mortality and measured response surface experimental design in which we colony size and fecundity of all surviving individuals. To manipulated densities of our study species across a estimate size and fecundity of Watersipora, we took a range of density combinations such that each species high-resolution digital photograph of each colony and occurred at both different densities and relative abun- used image processing software (ImageJ: Rasband 1997– dances (Appendix B; for detailed descriptions of 2008) to estimate colony area and to count embryos response surface designs, see Inouye [2001], Damgaard (embryos in Watersipora are visible as spherical pink [2008]). Our design closely replicates the experience of structures behind zooid frontal walls; Hart and Keough species in real communities where densities and relative 2009). To estimate size in Bugula, we counted the abundances can take a wide range of values in space and number of bifurcations on each colony’s longest branch time. Most importantly, data generated using this and converted this metric into zooid number using approach can be fit statistically to theoretical models standard methods (Keough and Chernoff 1987). To of competition (Law and Watkinson 1987, Inouye 1999, estimate fecundity in Bugula, we used a dissecting 2001). We extended the use of the response surface microscope to count all embryos, which are individually approach by replicating the experiment in the field brooded in specialized, highly visible zooids called across multiple environments. Our extension enabled us ovicells. to estimate and directly compare population growth rates and interaction coefficients for both species in both Analytical methods: competition model analyses benign and stressful environments in a way that directly Our main goal was to determine how the environment links empirical field data with mathematical population affects density-independent population growth and the dynamic theory. strength and direction of intra- and interspecific We used standard methods to collect individuals of interaction coefficients, and to predict how these Watersipora and Bugula that were less than five days old changes were likely to influence the outcome of the and of similar size, and to manipulate recruit densities species’ interactions. To do this we first fitted our data to (Hart and Marshall 2009, Hart et al. 2012; see Appendix seven different competition models and used a model A). We attached individual recruits of each species selection procedure to choose a single model that best haphazardly within a 50 3 50 mm square in the center of described dynamics in our system. Candidate models 110 3 110 mm settlement plates according to 16 and the details of model selection are provided in combinations of conspecific and heterospecific densities Appendices A and D. Second, using only the best model, (Appendices B and C). To generate accurate parameter we used a model simplification procedure to determine estimates with an efficient experimental design we which model parameters were required to describe the followed the recommendations of Inouye (2001) to dynamics in our system. include a wide range of density combinations, including, The primary response variable for our competition as recommended, densities above and below densities of model analyses was per capita embryo production 2722 SIMON P. HART AND DUSTIN J. MARSHALL Ecology, Vol. 94, No. 12

(fecundity), which we used to quantify the per capita sizes at time t and t þ 1. To accommodate this feature, contribution to population size at time t þ 1. Consistent we studied competition among a single cohort of new with the original formulation of the population dynamic recruits in early-successional assemblages before they models that we apply, this single response variable were largely excluded by later-successional dominants. includes any effects of the environment or species Ideally, our response variable would have been the interactions on mortality and growth as well as any number of recruits of our study species to the same other demographic rate that affects fecundity. habitats at time t þ 1. However, the minute and Of the seven candidate competition models, dynamics dispersive larval phase in our study species’ life histories were best described by a Ricker model (Ricker 1954): makes this difficult. Therefore, we used our integrative measure of per capita reproduction at four weeks as the Ni;tþ1 ¼ k½env:expðaii½env:Ni;t aij½env:Nj;tÞ response variable in our competition model analyses. Ni;t Estimates of interaction coefficients (aii and aij) are not where Ni,t and Nj,t are the initial densities of recruits of affected by this response variable under two assump- species i and j, respectively, Ni,tþ1 is the total reproduc- tions, the first of which is that reproduction at week four tive output of species i at the end of the experiment, k is is linearly related to the number of recruits at time t þ 1. the density-independent growth rate, and aii and aij are Data from our research group suggest that fecundity at the absolute intra- and interspecific competition coeffi- week four is linearly related to lifetime fecundity in our cients, respectively. We fit the experimental data to the 2 experimental species (For Watersipora, R ¼ 0.30, F1, 161 Ricker model and estimated parameter values using 2 ¼ 69.9, P , 0.001; for Bugula, R ¼ 0.73, F1,22 ¼ 63.11, P nonlinear least-squares estimation (see Appendix A for , 0.001; Marshall et al. 2003, Marshall and Monro details). Importantly, we included environment as a two- 2013), but the relationship between fecundity and level categorical grouping factor in our model fits such recruitment is unknown because this requires an that the subscript [env.] indicates that parameter estimate of planktonic mortality. However, a linear estimates are allowed to vary between benign and relationship should again be a good approximation stressful environments. Including environment as a when density-independent processes determine larval factor allowed us to simultaneously fit all of our survival in the plankton, which is a reasonable experimental data for each species using a single model, assumption for species such as ours that are unlikely which resulted in separate parameter estimates and to compete in the plankton (as we will describe) and do therefore separate competition surfaces estimated for not mass spawn (Hughes et al. 2002). The second each environment. The full model has six parameter assumption is that we captured the major period of estimates for each focal species (i.e., three demographic competition, which should also be reasonable, given that parameters [k, a , and a ] in each of two environments ii ij competition is unlikely among the planktonic, non- [benign, stressful]). feeding larvae of these species. We did not quantify the To determine the most parsimonious form of the effects of interactions between benthic adults and Ricker model, we used likelihood ratio tests (LRT) to settling larvae, assuming that recruitment at time t 1 compare versions of the Ricker model with and without þ is to newly disturbed habitat and thus generations are each individual parameter such that we removed all terms that did not improve model fit at a significance nonoverlapping. We address this caveat in the Discus- level of 0.10. These tests included comparisons of models sion. with separate estimates of a population parameter in Although estimates of interaction coefficients should each environment to a simpler (reduced) model with be unaffected by our approach, we did not estimate true only a single estimate of the same parameter across both rates of density-independent increase, k (Inouye 1999). environments. When these tests were not significant, we To precisely predict population dynamics, our estimates 0 compared models with and without the population of k, which we call k , must be adjusted by an unknown parameter of interest to determine if each individual linear term h(1 m), where h is a constant relating parameter should be included to describe dynamics. We fecundity at week four (our response variable) to lifetime fecundity, and m quantifies the proportional (value repeated these tests separately for each parameter (k, aii, and aij). For each parameter in the final, most between 0 and 1) mortality of larvae in the plankton, parsimonious model, we constructed likelihood profiles, and thus 1 m gives the proportion of all embryos which we then used to calculate 95% confidence intervals counted that survive to settlement at time equals t þ 1. around each estimate using standard methods (for a True values of lambda are then equal to h(1 m) k0.In detailed explanation of this procedure, see Venables and the absence of estimates for h and m, our parameter (k0) Ripley 2002, Bolker 2008, Ritz and Streibig 2008). provided a relative estimate of the direct effects of the Linking mathematical competition theory with em- environment on density-independent population growth pirical field data is difficult and we relied on some rates. Our conclusions will hold when the observed reasonable assumptions to justify our approach. The differences in fecundity are maintained across the life competition model is a discrete-time population model cycle, and when planktonic mortality of the similar-sized that typically relies on comparisons between population larvae of these species is random with respect to species December 2013 STRESS, FACILITATION, AND COMPETITION 2723 identity. We examine the sensitivity of our results to described using a Ricker competition model. The these assumptions in Appendix E. dynamics of both species were dominated by the direct We explored some of the consequences of our effects of the environment on population growth rates competition model results for population dynamics by (Table 1, Fig. 1; z intercept at Bugula ¼ 0and assessing the relative capacity for population growth Watersipora ¼ 0). There was a large, negative effect of between environments as a function of competitor the vertical environmental conditions on k0 that was density, density-independent growth rates, and compe- proportionally similar in magnitude for both species. tition coefficients. In Appendix E, we expand this For Watersipora, k0(benign) ¼ 24.7 vs. k0(harsh) ¼ 4.3 assessment by constructing phase planes with vector (likelihood ratio test (LRT) comparing a model with a fields to show the trajectories (direction and relative single estimate of k0 to a model with a separate estimate magnitude) of population change across a range of of k0 for each environment: v2 ¼ 81.098, df ¼ 1, P , densities of both species. 0.0001). For Bugula, k0(benign) ¼ 59.7 vs. k0(harsh) ¼ 15.7 (LRT: v2 ¼ 7.839, df ¼ 1, P ¼ 0.0051). This result Analytical methods: survival and size in response confirmed our a priori expectation that the vertical to environment and competition habitat was intrinsically more stressful than the hori- To understand the extent to which differences in zontal habitat for both species. survival and/or individual size contributed to the In Watersipora, intraspecific competition aww (sub- observed outcomes of the competition model analyses, scripts on parameters show ‘‘w’’ for Watersipora and ‘‘b’’ we assessed the influence of the environment and for Bugula) was an order of magnitude stronger in competition on survival and colony size. It should be benign environments (for aww, benign ¼ 0.025, harsh ¼ noted that the effects of the environment and competi- 0.004; LRT comparing single estimate vs. separate tion on the response variable in our competition model estimates for each environment, v2 ¼ 4.082, df ¼ 1, P ¼ analyses implicitly include the effects of changes in 0.0433; Table 1, Fig. 1). Indeed, although there were survival and colony size in so far as these two negative effects of intraspecific competition in benign demographic rates ultimately influence per capita environments, intraspecific competition in harsh envi- fecundity. Therefore, our univariate analyses of survival ronments did not differ significantly from zero (t95 ¼ and size complement, but should not be considered 0.46, P ¼ 0.645). In contrast, intraspecific competition in independently of, our competition model analyses. Bugula was strong and did not differ between environ- Survival of Watersipora was high (.95%)inall ments (abb across both environments ¼ 0.032; LRT environment–density combinations, so these data were comparing single vs. separate, environment-dependent not analyzed. We assessed survival of Bugula using a estimates, v2 ¼ 0.005, df ¼ 1, P ¼ 0.94; LRT comparing generalized linear mixed model (GLMM) with binomial 2 presence vs. absence of abb in the model, v ¼ 4.763, df ¼ errors and a logit link function. Environment was a fixed 1, P ¼ 0.0291). These tests indicated a need to retain abb effect and backing panel nested within environment was in the model (Table 1). a random effect. Models were simplified using likelihood Interspecific competition by Bugula on Watersipora ratio tests (LRT) to remove nonsignificant terms at did not differ between environments (LRT, v2 ¼ 2.211, alpha 0.30 (random terms were left in the model). The df ¼ 1, P ¼ 0.137; Table 1) and was relatively weak (awb influence of competition and environment on the mean ¼ 0.012; LRT, v2 ¼ 3.518, df ¼ 1, P ¼ 0.0607; Fig. 1, size of individuals of both species was assessed using Table 1). In contrast, the effects of Watersipora on ANCOVA, where Bugula density and Watersipora Bugula differed dramatically between environments density were continuous independent variables, environ- (LRT comparing single vs. separate, environment- ment was a fixed categorical independent variable, and 2 dependent estimates of abw, v ¼ 4.922, df ¼ 1, P ¼ backing panel nested within environment was a random 0.0265), indicating that separate estimates are necessary term. Where necessary, we transformed data to meet the (Table 1). There was a large negative competitive effect assumptions of the analytical methods. We sequentially of Watersipora on Bugula in benign environments (abw ¼ removed higher-level nonsignificant terms at alpha 0.034). However, in harsh environments, Watersipora 0.30. Analyses were done using R v. 2.13.0 (R strongly facilitated Bugula population growth (abw ¼ Development Core Team 2011). More details of our 0.043; note that a negative ‘‘competition’’ coefficient analytical methods and results are provided in Appendix indicates facilitation). A (supplementary materials and methods) and Appen- dix E (phase plane construction and analysis). Implications for population dynamics

RESULTS The parameterized model can be used to understand some important features of competitive population Parameter estimates from the Ricker competition model dynamics between nonoverlapping generations of and inﬂuence of environment Watersipora and Bugula settling on bare substrata. The Competitive population dynamics between nonover- term eaij Nj in the Ricker model gives the per capita lapping generations of Watersipora and Bugula settling growth rate of species i in the presence of competition on bare (e.g., recently disturbed) substrata can be from N individuals of species j, relative to the density- 2724 SIMON P. HART AND DUSTIN J. MARSHALL Ecology, Vol. 94, No. 12

TABLE 1. Estimates and 95% profile-likelihood confidence intervals for parameters describing competition between two invasive marine bryozoans, Watersipora subtorquata and Bugula neritina, from the best-fit, most parsimonious Ricker competition model.

95% CI Model Parameter Estimate Lower Upper RSE/df

Watersipora subtorquata, F ¼ keawwNw awb Nb k0 benign 24.679 20.237 29.966 0.140/91 k0 harsh 4.331 3.340 5.577 aww benign 0.025 0.012 0.038 aww harsh 0.004 0.013 0.021 awb 0.012 0.0008 0.024 Bugula neritina, F ¼ keabbNbawbNw k0 benign 59.717 33.092 107.705 1.196/91 k0 harsh 15.718 8.546 28.565 abb 0.032 0.002 0.062 abw benign 0.034 0.015 0.085 abw harsh 0.043 0.092 0.008 Notes: Here, F is per capita fecundity, which is an integrative metric that includes all density-independent and density-dependent effects on growth, survival, reproduction, and any other demographic rates that affect the contribution of an individual to population size at time t þ 1. The density-independent rate of increase is k, aii and aij are the intra- and interspecific competition coefficients, respectively, and Ni is the initial density of species i. Subscripts on parameters represent our study species, with ‘‘w’’ for Watersipora and ‘‘b’’ for Bugula. Separate parameter estimates are provided for different levels of environmental stress (benign/ harsh) when log-likelihood ratio tests indicated that separate estimates significantly improved model fit. RSE stands for residual standard error, df is model degrees of freedom, and CI stands for confidence interval. It should be noted that, by their very nature, competition coefficients naturally take values much less than 1.

independent growth rate (ki; Fig. 2). The value of this competitors (or vice versa). So while the intensity of term across a range of competitor densities shows that intraspecific competition in Watersipora (aww) declined low densities of either Bugula or Watersipora can have in the harsh environment (Table 1), the large decrease in large proportional effects on Bugula growth rates in the density-independent growth rate of Watersipora (kw) both environments. Approximately 20 individuals (den- more than compensated, such that population growth is sities that are regularly observed in the field) of either expected to be higher in the harsh environment. In Bugula (both environments) or Watersipora (benign contrast, the decrease in kw makes Watersipora more environments) can halve the growth rates of Bugula, vulnerable to interspecific competition in the harsh whereas even fewer individuals (;16) of Watersipora environment relative to the benign environment, even can double growth rates in the harsh environment though awb is unchanged (Table 1). Similarly, intraspe- through facilitation (Fig. 2c, d). Similarly, low densities cific density dependence will be stronger for Bugula in of conspecifics have large effects on growth rates in the harsh environment because Bugula maintains strong Watersipora in benign environments but not harsh intraspecific competitive effects (abb)intheharsh environments (Fig. 2b). environment, but the population growth rate (kb)is A similar proportional change in population growth reduced. rate caused by species interactions (Fig. 2) will have a Reciprocal interactions between species determine the larger consequence for persistence when density-inde- outcome of species interactions. Facilitation by Water- pendent growth rates are already low as a consequence sipora will benefit Bugula in the harsh environment. of living in a harsh environment, which was the case for However, as the population size of Bugula increases, the both species (Table 1). Taking the natural log of both absolute strength of the competitive effect of Bugula on Watersipora will increase (increasing value of eawbNb ; sides of the Ricker equation and considering the effects Fig. 2). Therefore, the reciprocal competitive effect of of intra- and interspecific competition separately gives Bugula on Watersipora will ultimately limit facilitation the following: of Bugula by limiting Watersipora population size. For a Ni;tþ1 more detailed description of the population-level impli- ln ¼ lnðkÞaiiNi Ni;t cations of our results, refer to Appendix E. for population growth in the presence of conspecifics, Survival and size in response to environment and and competition There were contrasting effects on survival between the Ni;tþ1 ln ¼ lnðkÞaijNj species. Survival of Watersipora was high in both Ni;t environments and across all density combinations for population growth in the presence of heterospecifics. (.95% survival). However, the environment and com- These formulations show that a low value of k or a petitors strongly affected survival in Bugula. Surprising- high value of aii or aij results in less capacity for ly, survival of Bugula was actually higher in harsh population growth in the presence of a given number of environments in the absence of competition (Fig. 3); December 2013 STRESS, FACILITATION, AND COMPETITION 2725

FIG. 1. Population-level effects of intra- and interspecific competition on two invasive marine bryozoans, (a, b) Watersipora subtorquata and (c, d) Bugula neritina in benign (blue) and harsh (orange) environments. Points are observed values, and surfaces are the predicted values from the fitted model. Black and red symbols (dots and lines) indicate residuals above and below predicted surfaces, respectively. Note: to best illustrate the relationship between independent variables and the response, the axis rotation varies between (a/b) and (c/d). Also note the log response variable in panels (c) and (d). predicted proportion surviving was 0.54 6 0.008 in the harsh environment, although this effect was (estimate 6 SE) in the benign environment and 0.69 6 marginally nonsignificant for Watersipora (P ¼ 0.087). 0.007 in the harsh environment. However, both intra- This most likely reflects a lack of power as a result of specific and interspecific densities reduced survival of low replication at the appropriate scale (backing panel Bugula and there was a stronger negative effect of [environment], n ¼ 2; Appendix F). The mean size of intraspecific (Bugula) densities on survival in harsh individuals of Watersipora decreased with increasing environments than in benign environments (LRT for densities of both Watersipora and Bugula, and this effect significance of Bugula density 3 environment interac- was consistent in both environments (Appendix F; Fig. tion, v2 ¼ 9.87, df ¼ 1, P ¼ 0.002; Fig. 3). There were consistent negative effects of environmen- 4). In contrast, there were no indications of negative tal harshness on the mean size of individuals of both effects of competitor density on size in Bugula in either species, but the effects of competitors varied between environment; indeed, the mean size of Bugula tended to species (Appendix F; Fig. 4). In both Watersipora and increase with increasing densities of Watersipora in Bugula, the mean size of individuals tended to be lower harsh, but not benign, environments although this effect 2726 SIMON P. HART AND DUSTIN J. MARSHALL Ecology, Vol. 94, No. 12

FIG. 2. Effects of species interactions in harsh and benign environments across a range of competitor densities on population growth rates, relative to density-independent growth rates. Values on the y-axes are the factor by which density-independent growth is multiplied as a consequence of an interaction. The horizontal dotted line indicates no effect of an interaction on population growth rates.

FIG. 3. Effects of environmental stress and (a) Bugula and (b) Watersipora density on Bugula survival. Symbols and lines show observed and predicted values, respectively. Jitter has been added to observed values to distinguish overlapping data. Lines are predicted values from a GLMM with binomial errors, including environment, competitor densities, and all their interactions as independent variables. Interactions between Bugula and Watersipora density were not statistically signiﬁcant, so the effect of each competitor on survival is plotted separately. December 2013 STRESS, FACILITATION, AND COMPETITION 2727

FIG. 4. Effects of environmental stress and competitor density on the mean size of individuals of (a, b) Watersipora and (c, d) Bugula. Symbols show observed values, and lines show predicted values from ANCOVA, including environment, competitor densities, and all their interactions as independent variables. Interactions between Bugula and Watersipora density were not statistically signiﬁcant (see Appendix F); therefore the effect of each competitor on size is plotted separately for both species. Data were log-transformed for analysis. A small amount of jitter has been added to observed values to distinguish overlapping data points.

was marginally nonsignificant (environment 3 Water- cific competition in Watersipora (aww)declinedinthe sipora interaction, P ¼ 0.070; Appendix F; Fig. 4). harsh environment, as would generally be predicted by stress gradient models, the larger proportional decline DISCUSSION in Watersipora’s density-independent growth rate (kw) Environmental stress can have predictable effects on compensated, such that intraspecific density depen- the intensity of species interactions, but the conse- dence will be lower in stressful environments, effective- quences of these changes for population dynamics are ly increasing carrying capacity. Further, while unlikely to be straightforward. Consistent with stress interspecific interactions strongly modify single-species gradient models, the intensity of competition tended to dynamics, it is the reciprocal nature of these interac- decline and facilitation emerged in the harsh environ- tions that is of overriding importance. For example, ment, and both types of interactions are expected to facilitation by Watersipora will benefit Bugula in the have strong effects on dynamics even at low densities harsh environment, but the reciprocal competitive (Table 1, Fig. 2). However, it is the interaction between effect of Bugula on Watersipora will ultimately limit each of the species-specific, density-independent (i.e., this facilitation by limiting Watersipora population direct effects on ki ) and density-dependent (i.e., growth. Our empirical approach demonstrates the indirect effects via aii and aij) responses to the importance of placing changes in the strength and environment that determines the population-level direction of interactions among environments in their response. For example, while the intensity of intraspe- population-dynamic context (Freckleton et al. 2009). 2728 SIMON P. HART AND DUSTIN J. MARSHALL Ecology, Vol. 94, No. 12

The idea that the role of competition is diminished in prey. Theory predicts that subtle interactions between harsh environments is deeply embedded in conceptual facilitation and competition can fundamentally change theory (Menge and Sutherland 1987, Bruno et al. 2003, our expectations of community dynamics (Gross 2008). Grime 2007) and has been used to interpret a range of Our results are distinct from Gross’s conclusions in that ecological patterns, including species range limits it is reciprocal competitive–facilitative interactions that (Normand et al. 2009) and patterns of diversity (Huston can stabilize dynamics, rather than simultaneous com- 1979). Our results highlight not only that the strength of petitive and facilitative effects of a benefactor species on competition need not always decline, but also, more beneficiaries. However, both studies emphasize the importantly, that the consequences of competition for importance of simultaneous positive and negative population dynamics can actually increase in harsh interactions at the population level; further work in this environments. For example, the per capita strength of area is likely to be fruitful (Holland and DeAngelis interspecific competition by Bugula on Watersipora 2010). (awb) did not differ between environments (Table 1), but the population-level consequences for Watersipora Life history variables and mechanisms differ dramatically. The large negative effect of the harsh Harsh environments and competition reduced pop- environment on Watersipora’s density-independent ulation growth of both species (Fig. 1), but the growth rate (kw; Table 1) means that Watersipora has mechanisms behind these effects varied between spe- less ability to maintain positive population growth in the cies. In Watersipora, reductions in colony size in harsh environment and thus is less tolerant of interspe- response to both the harsh environment and compet- cific competition even though the strength of competi- itor density appeared to cause some of the reductions in tion (awb) is unchanged. Such a result is consistent with per capita reproduction (Fig. 4). This result is theoretical work that challenges the logic of traditional consistent with the idea that demographic rates in conceptual models for the effect of stress on competi- modular organisms are strongly influenced by the size tion, on the basis that these models rarely incorporate of individuals (Hughes 1984, Harper et al. 1986). the direct negative effect of harsh environments on However, in Bugula the relationships between the species’ abilities to tolerate competition (Chesson and population-level response and the survival and growth Huntly 1997; see also Violle et al. 2010, Fox 2012). Our of individuals within those populations was less clear. results do not question the veracity of conceptual stress Despite strong intra- (both environments) and inter- gradient models such as the SGH (some of our data specific (benign environments) competition, increasing support the SGH), but instead demonstrate the impor- competitor densities had no negative effect on colony tance of placing predicted changes in the strength or size in Bugula (Fig. 4). Although this suggests that size frequency of interactions in a population-dynamic does not strongly determine fecundity in benign context (Chesson and Huntly 1997, Freckleton et al. environments, small increases in colony size in harsh 2009). environments (Fig. 4b) may partially explain the facilitative effect of Watersipora on Bugula. Similarly, Role of competition and facilitation in interspecific effects on survival of Bugula were also complex and population regulation cannot be used to predict the population-level out- Positive interactions are an important component of comes of the interactions, as described by our ecological theory (Stachowicz 2001, Bruno et al. 2003), competition model analysis (compare Fig. 1c, d with and are perhaps the most prominent and influential Fig. 3). For example, although the environment had feature of stress gradient models (Bertness and Callaway direct negative effects on per capita reproduction 1994, Maestre et al. 2009). However, an important gap (k(harsh) , k(benign), remembering that k integrates in this field is the need to explain how populations are across all density-independent effects on survival, regulated when positive interactions can promote growth, and reproduction), Bugula survival was often unbounded population growth (Holland et al. 2002). higher in harsh environments (Fig. 3; also see Burgess One of our most striking results suggests that the et al. 2012). Furthermore, the strength of intraspecific emergence of facilitation in stressful environments can competition on Bugula survival was stronger in harsh be stabilized by reciprocal competitive interactions in a environments (compare slopes in Fig. 3a), even though way that may regulate populations in multispecies the overall per capita strength of intraspecific compe- communities. This is because Bugula will limit the tition (abb) was unchanged (Fig. 1, Table 1). These strength of facilitation by Watersipora by limiting results suggest that the nature and context of an Watersipora population size through its own interspe- interaction can strongly determine which demographic cific competitive effects (see Appendix E for details). variables are affected, and also how these variables Such a reciprocal interaction between facilitation and translate to population-level effects. Choosing an competition is analogous to consumer–resource interac- appropriate metric is a critical component of studies tions (Holland and DeAngelis 2010), where the prey of species interactions (Maestre et al. 2005, Snyder contributes to population growth of the predator while 2008, Freckleton et al. 2009), and our results suggest the predator simultaneously reduces the density of its that this is even more important when comparing the December 2013 STRESS, FACILITATION, AND COMPETITION 2729 nature of interactions among environments. Integra- and different classes of stressors are expected to have tive, population-level metrics are best, but where this is different effects on density-independent rates of in- not possible, environment-specific ground-truthing of crease and competition (Chesson and Huntly 1997, the relationship between the measured variable and the Kawai and Tokeshi 2007). By manipulating panel population-level response is strongly recommended orientation we certainly modified the exposure of our (e.g., Hart and Keough 2009). study organisms to sedimentation and UV stress, but Although our approach was deliberately phenomeno- we also probably modified other variables such as food logical, we speculate on some of the mechanisms behind availability and predation. Indeed, multiple covarying our results because there is increasing evidence that biotic and abiotic factors are a feature of many Watersipora is an important foundation species for the important environmental gradients (Maestre et al. development of subtidal assemblages. Stachowicz and 2009). We were interested in the influence of the Byrnes (2006) demonstrated that Watersipora creates environment on the outcome of competition so we habitat that facilitates colonization of other species. deliberately used a phenomenological approach. Stud- Increased colonization success cannot be the mechanism ies that focus on the mechanisms underpinning the responsible for increased performance of Bugula in our population-level effects would be complementary. study. We speculate that Watersipora may mediate the Similarly, many of the insights of stress gradient negative effects of sedimentation on Bugula in harsh theories of community organization and modern, environments, perhaps by changing patterns of water population-dynamic theories of coexistence are com- flow via generation of feeding currents (Okamura 1985, plementary. Our approach highlights that a more Vogel 1994). That Watersipora can facilitate other comprehensive understanding of community dynamics species by at least two mechanisms is consistent with emerges when the approaches and findings of these extensions of stress gradient models that highlight the related theories mutually inform each other. role of foundation species for providing the physical ACKNOWLEDGMENTS context for community development, such that commu- We thank Malcolm Keag, Simone Higgie, Jacquei Burgin, nities themselves are hierarchically organized (Bruno and David Aguirre for help in the laboratory and field, and and Bertness 2001, Altieri et al. 2010, Angelini et al. Scott Burgess, Ed Spitznagel, and Victor Wickerhauser for 2011). statistical advice. We thank Peter Adler, Jeremy Fox, Jon Chase, Peter Chesson, Jay Stachowicz, Brian Inouye, and three Limitations and future studies anonymous reviewers for comments and suggestions that substantially improved the manuscript. Any errors remain our Our approach assesses dynamics in a single cohort, own. This work was funded by an Australian Research Council but takes no account of competition between adults grant to D. J. Marshall. S. P. Hart was supported by a and recruits from the plankton. Watersipora,for University of Queensland Research Scholarship. example, may affect recruitment by preempting space LITERATURE CITED (a negative competitive effect), or by providing Adler, P. B., J. HilleRisLambers, P. C. Kyriakidis, Q. F. Guan, secondary space for recruitment (a positive effect; and J. M. Levine. 2006. Climate variability has a stabilizing Stachowicz and Byrnes 2006). Size of Watersipora, effect on the coexistence of prairie grasses. Proceedings of the and therefore space occupation, tended to be lower in National Academy of Sciences USA 103:12793–12798. harsh environments, suggesting that size-dependent Altieri, A. H., B. K. van Wesenbeeck, M. D. Bertness, and B. R. Silliman. 2010. Facilitation cascade drives positive positive or negative effects of Watersipora may be relationship between native biodiversity and invasion success. reduced in harsh relative to benign environments. Ecology 91:1269–1275. Unfortunately, the net effect of these influences on Angelini, C., A. H. Altieri, B. R. Silliman, and M. D. Bertness. population dynamics is unknown. Consequently, our 2011. Interactions among foundation species and their consequences for community organization, biodiversity, and results apply most strongly to nonoverlapping gener- conservation. BioScience 61:782–789. ations establishing on bare substrata, dynamics that are Angert, A. L., T. E. Huxman, P. Chesson, and D. L. Venable. consistent with the dominance of these species in early- 2009. Functional tradeoffs determine species coexistence via successional assemblages at our field site. Furthermore, the storage effect. Proceedings of the National Academy of our results apply when the scale of environmental Sciences USA 106:11641–11645. Bertness, M. D., and R. Callaway. 1994. Positive interactions in variation exceeds the scale of dispersal, such that communities. Trends in Ecology and Evolution 9:191–193. offspring recruit into an environment that is most Bolker, B. M. 2008. Ecological models and data in R. Princeton similar to that of their parents. This is a simplifying University Press, Princeton, New Jersey, USA. assumption so that we could model the effects of Bolker, B. M., and S. W. Pacala. 1999. Spatial moment equations for plant competition: Understanding spatial environmental conditions on competitive population strategies and the advantages of short dispersal. American dynamics in separate environments. The alternative is Naturalist 153:575–602. certainly possible, but would require a much more Bruno, J. F., and M. D. Bertness. 2001. Habitat modification complex model with experiments that include pheno- and facilitation in benthic marine communities. Marine Community Ecology 201–218. typic links among life history stages. Bruno, J. F., J. J. Stachowicz, and M. D. Bertness. 2003. Environmental stress can take many forms, including Inclusion of facilitation into ecological theory. Trends in resource stress, physiological stress, and disturbance, Ecology and Evolution 18:119–125. 2730 SIMON P. HART AND DUSTIN J. MARSHALL Ecology, Vol. 94, No. 12

Burgess, S. C., E. A. Treml, and D. J. Marshall. 2012. How do Inouye, B. D. 1999. Estimating competition coefficients: strong dispersal costs and habitat selection influence realized competition among three species of frugivorous flies. population connectivity? Ecology 93:1378–1387. Oecologia 120:588–594. Buss, L. W. 1990. Competition within and between encrusting Inouye, B. D. 2001. Response surface experimental designs for clonal invertebrates. Trends in Ecology and Evolution 5:352– investigating interspecific competition. Ecology 82:2696– 356. 2706. Caceres, C. E. 1997. Temporal variation, dormancy, and Irving, A. D., and S. D. Connell. 2002. Sedimentation and light coexistence: A field test of the storage effect. Proceedings of penetration interact to maintain heterogeneity of subtidal the National Academy of Sciences USA 94:9171–9175. habitats: algal versus invertebrate dominated assemblages. Callaway, R. M. 2007. Positive interactions and interdepen- Marine Ecology Progress Series 245:83–91. dence in plant communities. Springer, Dordrecht, The Kawai, T., and M. Tokeshi. 2007. Testing the facilitation– Netherlands. competition paradigm under the stress-gradient hypothesis: Chase, J. M., P. A. Abrams, J. P. Grover, S. Diehl, P. Chesson, decoupling multiple stress factors. Proceedings of the Royal R. D. Holt, S. A. Richards, R. M. Nisbet, and T. J. Case. Society B 274:2503–2508. 2002. The interaction between predation and competition: a Keough, M. J., and H. Chernoff. 1987. Dispersal and review and synthesis. Ecology Letters 5:302–315. population variation in the bryozoan Bugula neritina. Chesson, P. 2000. Mechanisms of maintenance of species Ecology 68:199–210. diversity. Annual Review of Ecology and Systematics 31: Law, R., and A. R. Watkinson. 1987. Response-surface analysis 343–366. of two-species competition: an experiment on Phleum Chesson, P. 2003. Understanding the role of environmental arenarium and Vulpia fasciculata. Journal of Ecology 75: variation in population and community dynamics. Theoret- 871–886. ical Population Biology 64:253–254. Levine, J. M., and J. HilleRisLambers. 2009. The importance of Chesson, P., and N. Huntly. 1997. The roles of harsh and niches for the maintenance of species diversity. Nature 461: fluctuating conditions in the dynamics of ecological commu- 254–257. nities. American Naturalist 150:519–553. MacArthur, R. H., and R. Levins. 1967. The limiting similarity, Connell, J. H. 1978. Diversity in tropical rainforests and coral convergence, and divergence of coexisting species. American reefs—high diversity of trees and corals is maintained only in Naturalist 101:377–385. a non-equilibrium state. Science 199:1302–1310. Maestre, F. T., R. M. Callaway, F. Valladares, and C. J. Lortie. Damgaard, C. 2008. On the need for manipulating density in 2009. Refining the stress-gradient hypothesis for competition competition experiments. Functional Ecology 22:931–933. and facilitation in plant communities. Journal of Ecology 97: Fox, J. W. 2012. The intermediate disturbance hypothesis 199–205. should be abandoned. Trends in Ecology and Evolution 28: Maestre, F. T., F. Valladares, and J. F. Reynolds. 2005. Is the 86–92. change of plant–plant interactions with abiotic stress Freckleton, R. P., A. R. Watkinson, and M. Rees. 2009. predictable? A meta-analysis of field results in arid environ- Measuring the importance of competition in plant commu- ments. Journal of Ecology 93:748–757. nities. Journal of Ecology 97:379–384. Marshall, D. J., T. F. Bolton, and M. J. Keough. 2003. Grime, J. P. 1979. Plant strategies and vegetation processes. Offspring size affects the post-metamorphic performance of a John Wiley, Chichester, UK. colonial marine invertebrate. Ecology 84:3131–3137. Grime, J. P. 2007. Plant strategy theories: a comment on Craine Marshall, D. J., and K. Monro. 2013. Interspecific competition (2005). Journal of Ecology 95:227–230. Gross, K. 2008. Positive interactions among competitors can alters nonlinear selection on offspring size in the field. produce species-rich communities. Ecology Letters 11:929–936. Evolution 67:328–337. Harper, J. L., B. R. Rosen, and J. White, editors. 1986. The Menge, B. A., and J. P. Sutherland. 1987. Community growth and form of modular organisms. The Royal Society, regulation: variation in disturbance, competition, and preda- London, UK. tion in relation to environmental stress and recruitment. Hart, S. P., J. R. Burgin, and D. J. Marshall. 2012. Revisiting American Naturalist 130:730–757. competition in a classic model system using formal links Miller, R. J., and J. E. Ron. 2008. Shading facilitates sessile between theory and data. Ecology 93:2015–2022. invertebrate dominance in the rocky subtidal Gulf of Maine. Hart, S. P., and M. J. Keough. 2009. Does size predict Ecology 89:452–462. demographic fate? Modular demography and constraints on Normand, S., U. A. Treier, C. Randin, P. Vittoz, A. Guisan, growth determine response to decreases in size. Ecology 90: and J.-C. Svenning. 2009. Importance of abiotic stress as a 1670–1678. range-limit determinant for European plants: insights from Hart, S. P., and D. J. Marshall. 2009. Spatial arrangement species responses to climatic gradients. Global Ecology and affects population dynamics and competition independent of Biogeography 18:437–449. community composition. Ecology 90:1485–1491. Okamura, B. 1985. The effects of ambient flow velocity, colony Holland, J. N., and D. L. DeAngelis. 2010. A consumer– size, and upstream colonies on the feeding success of resource approach to the density-dependent population Bryozoa. II. Conopeum reticulum (Linnaeus), an encrusting dynamics of mutualism. Ecology 91:1286–1295. species. Journal of Experimental Marine Biology and Holland, J. N., D. L. DeAngelis, and J. L. Bronstein. 2002. Ecology 89:69–80. Population dynamics and mutualism: Functional responses R Development Core Team. 2011. R: A language and of benefits and costs. American Naturalist 159:231–244. environment for statistical computing. R Foundation for Hughes, T. P. 1984. Population dynamics based on individual Statistical Computing, Vienna, Austria. size rather than age: a general model with a reef coral Rasband, W. S. 1997–2008. ImageJ. National Institutes of example. American Naturalist 123:778–795. Health, Bethesda, Maryland, USA. Hughes, T. P., A. H. Baird, E. A. Dinsdale, V. J. Harriott, Ricker, W. E. 1954. Stock and recruitment. Journal of the N. A. Moltschaniwskyj, M. S. Pratchett, J. E. Tanner, and Fisheries Research Board, Canada 11:559–623. B. L. Willis. 2002. Detecting regional variation using meta- Ritz, C., and J. C. Streibig. 2008. Nonlinear Regression with R. analysis and large-scale sampling: Latitudinal patterns in Springer, New York, New York, USA. recruitment. Ecology 83:436–451. Sears, A. L. W., and P. Chesson. 2007. New methods for Huston, M. 1979. A general hypothesis of species diversity. quantifying the spatial storage effect: an illustration with American Naturalist 113:81–101. desert annuals. Ecology 88:2240–2247. December 2013 STRESS, FACILITATION, AND COMPETITION 2731

Siepielski, A. M., and M. A. McPeek. 2010. On the evidence for extrinsic factors. Marine Ecology Progress Series 311:251– species coexistence: a critique of the coexistence program. 262. Ecology 91:3153–3164. Venables, W. N., and V. D. Ripley. 2002. Modern applied Snyder, R. E. 2008. When does environmental variation most statistics with S. Springer, New York, New York, USA. influence species coexistence? Theoretical Ecology 1:129–139. Violle, C., Z. Pu, and L. Jiang. 2010. Experimental demonstra- Stachowicz, J. J. 2001. Mutualism, facilitation, and the tion of the importance of competition under disturbance. structure of ecological communities. BioScience 51:235–246. Proceedings of the National Academy of Sciences USA 107: Stachowicz, J. J., and J. E. Byrnes. 2006. Species diversity, 12925–12929. invasion success, and ecosystem functioning: disentangling Vogel, S. 1994. Life in moving fluids. Second edition. Princeton the influence of resource competition, facilitation, and University Press, Princeton, New Jersey, USA.

SUPPLEMENTAL MATERIAL

Appendix A Supplementary materials and methods, including expanded descriptions of the study system and species’ life histories, experimental and analytical methods, and results of model selection (Ecological Archives E094-252-A1).

Appendix B Density combinations of Watersipora subtorquata and Bugula neritina used in the response surface experiment (Ecological Archives E094-252-A2).

Appendix C Experimental setup showing arrangement of settlement plates on backing panels in benign and harsh environments (Ecological Archives E094-252-A3).

Appendix D Candidate competition models (Ecological Archives E094-252-A4).

Appendix E Construction and analysis of phase planes, including zero net growth isoclines and vector ﬁelds, to describe competitive population dynamics of Watersipora and Bugula (Ecological Archives E094-252-A5).

Appendix F Results of ANCOVA testing for the inﬂuence of environmental stress, Watersipora density, and Bugula density on mean individual size (Ecological Archives E094-252-A6).