ecosphere, 4(8): 100 nilsson, k., persson, l. (2013...

http://www.diva-portal.org

This is the published version of a paper published in Ecosphere.

Citation for the original published paper (version of record):

Nilsson, K., Persson, L. (2013)

Refuge availability and within-species differences in cannibalism determine population variability

and dynamics.

Ecosphere, 4(8): 100

http://dx.doi.org/10.1890/ES13-00105.1

Access to the published version may require subscription.

N.B. When citing this work, cite the original published paper.

Permanent link to this version:http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-84804

Refuge availability and within-species differencesin cannibalism determine population variability and dynamics

KARIN A. NILSSON� AND LENNART PERSSON

Department of Ecology and Environmental Science, Umea University, S-901 87 Umea, Sweden

Citation: Nilsson, K. A., and L. Persson. 2013. Refuge availability and within-species differences in cannibalism

determine population variability and dynamics. Ecosphere 4(8):100. http://dx.doi.org/10.1890/ES13-00105.1

Abstract. Theoretical studies show that both cannibalism and intraspecific resource competition can

have major effects on population dynamics. Cannibalistic intensity, offspring size, harvesting and refuge

availability are important factors affecting the interplay between cannibalism and competition. We studied

two populations of the common guppy (Poecilia reticulata) that differed in their cannibalistic voracity as

well as offspring size. We manipulated the availability of refuges for juveniles and harvesting intensity of

large adults to investigate how these factors influenced the dynamics of the two populations.

Overall population dynamics was mainly affected by the origin of the founder populations and the

presence of refuges. The population with a higher cannibalistic propensity and smaller offspring exhibited

higher population variability, and the presence of refuges reduced cannibalism and stabilised the dynamics

in both populations. Harvest of large cannibalistic females destabilised the dynamics and caused

extinctions of several populations without refuges. Both populations displayed cannibal-driven cycles with

repression of recruitment when no refuges were present. Cycle periods were shorter with refuges present

and the dynamics were more cohort like with synchronised peaks in density of vulnerable juveniles and

cannibals. We suggest that increased number of refuging juveniles led to intensified resource competition

in the population. The harvest yield was low in the refuge treatments as few females grew large due to

resource competition, leading to a small impact of harvesting in these treatments.

Key words: cannibalism; coefficient of variation; competition; extinction; guppy; harvest; Poecilia reticulata; population

dynamics; refuges; structural complexity.

Received 25 March 2013; revised 9 May 2013; accepted 21 May 2013; final version received 15 July 2013; published 20

August 2013. Corresponding Editor: E. Garcıa-Berthou.

Copyright: � 2013 Nilsson and Persson. This is an open-access article distributed under the terms of the Creative

Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided

the original author and source are credited. http://creativecommons.org/licenses/by/3.0/

� E-mail: [email protected]

INTRODUCTION

Cannibalism is a common phenomenon innature and may have strong impacts on popula-tion dynamics (Ricker 1954, Fox 1975a, Polis1981, Hastings and Costantino 1991, Cushing1992, Claessen and de Roos 2003, Claessen et al.2004, Persson et al. 2004). Theoretical studieshave shown that density dependent cannibalismand a small time-delay is enough to generatepopulation cycles and that intense cannibalismmay lead to large amplitude cycles and even

extinctions (Hastings and Costantino 1991, Claes-sen and de Roos 2003). In spite of this, there arefew clear empirical examples of cannibal drivencycles (but see Dennis et al. 1997, Elliott 2004,Wissinger et al. 2010). Still, recruitment regula-tion by cannibalism, which is the key elementbehind cannibal-driven cycles has been recordedin several species including fish (Polis 1981,Andersson et al. 2007), the flower beetle Tribolium(Hastings and Costantino 1991, Dennis et al.1997, Benoit et al. 1998), salamanders (Wissingeret al. 2010), leeches (Elliott 2004), isopods

v www.esajournals.org 1 August 2013 v Volume 4(8) v Article 100

(Leonardsson 1991) and backswimmers (Fox1975b, Orr et al. 1990).

Cannibalistic interactions are typically size-dependent and in particular the smallest sizedvictim that a cannibal can possibly capture hasbeen shown to be crucial for dynamics (Claessenet al. 2002). For example, the lower size limit tocannibalism differs between species and has beenshown to determine the population dynamics ofdifferent fish species (Persson et al. 2004).Furthermore, the size of the offspring is also ofimportance for the dynamics. With large andintermediate hatchling sizes, cannibalism can beregulating and harvesting adults will in suchpopulations have a destabilising effect (vanKooten et al. 2007, 2010). Correspondingly,experimental studies in Tribolium systems haveshown that harvesting of adults can result in bothstabilisation as well as destabilisation dependingon the intensity of the harvesting (Dennis et al.1995, 1997).

The presence of refuges is generally consideredto stabilise population dynamics. In predator-prey systems refuge availability for prey resultsin decreased predator efficiency, which dampensor even extinguishes predator-prey cycles (Rose-nzweig and MacArthur 1963, Stenseth 1980,Krivan 1998). Refuge availability will also reducecannibalism and in the case where cannibalisminduces population cycles in the absence ofrefuges, addition of refuges can be hypothesisedto stabilise dynamics. However, the situation ismore complex since prey and predator belong tothe same population and, may also compete forthe same resources. Increasing competition forresources within a population may lead to cohortcycles (de Roos et al. 2002). Thus refugeavailability may have a destabilising effect whendecreased cannibalism results in higher popula-tion densities and subsequent intensified compe-tition (Cushing 1991, van den Bosch and Gabriel1997). Experimental studies on how cannibalisticpopulation dynamics are affected by refuges arehowever few (but see Benoit et al. 1998, 2000),and none have examined the population dynam-ical effects of increased competition among andwithin cohorts.

Here we used two strains of the commonguppy (Poecilia reticulata), that differ in bothoverall cannibalistic voracity and offspring size,to study the effects of cannibalism and resource

competition on population dynamics. We ma-nipulated the availability of structural refugesand harvesting intensity on large cannibals inlong term population experiments and made thefollowing predictions: (1) The scope for canni-balism to affect recruitment is large in bothpopulations due to the large offspring size andthe low per capita fecundity. In the absence ofrefuges we therefore expected the populations todisplay cannibal-driven cycles. We further ex-pected that the population with higher cannibal-istic propensity and smaller offspring shouldshow more fluctuating dynamics. (2) We expect-ed harvest of large females to have a destabilisingeffect as it could allow pulses of recruits to breakthrough the cannibal regulation. (3) We expectedthe presence of refuges to reduce cannibalismand to potentially have a stabilising effect onpopulation dynamics. Still, because refuge avail-ability may lead to intensified resource competi-tion as a result of increased survival of juveniles,fluctuations influenced by cohort competitioncould counteract the stabilising effect of reducedcannibalism. To study this we carried out crosscorrelation analyses of the timing of peaks indensities of different stages, as the timing of thesediffers between the fluctuations driven by canni-balism and those driven by cohort competition.

MATERIALS AND METHODS

Study organismOur study organism, the common guppy

Poecilia reticulata, is a small poecilid fish thatlives in freshwater ponds and streams. Littersarrive at approximately 25 days interval underlaboratory conditions (Reznick and Yang 1993)and the generation time is 10 weeks or longer inlaboratory depending on food availability (Re-znick and Bryga 1987; see also the Appendix).Many studies on life- history characteristics havebeen performed with guppies in Trinidad wherethey coexist with different predators and haveevolved different life-history characteristics inresponse to the predation environment (Magur-ran 1998). Guppies from high predation environ-ments have a smaller size and lower age atmaturity, a larger reproductive allocation andsmaller sized offspring (Magurran 1998, Reznicket al. 2001). One of the populations in this studyoriginated from a high predation site in the


NILSSON AND PERSSON

Turure River in the Oropuche drainage situatedon the southern slopes of the Northern Range inTrinidad. The other population originated from alow predation site in the Quare River also on thesouthern slopes of the Northern Range. It isknown that the per capita fecundity of Turureguppies from high predation environments ishigher than that of Quare guppies from lowpredation environments (S. Auer, unpublisheddata) and that the propensity for cannibalismand cannibalistic voracity is higher for the Tururehigh predation population in laboratory condi-tions when no refuges are present (Nilsson et al.2011). It is also known that cannibal efficiency isaffected by the presence of refuges and thatjuveniles increase refuge use in response tocannibal presence (Nilsson et al. 2011).

Experimental setup, treatments and samplingThe experiment was performed at the Depart-

ment of Ecology and Environmental Science,Umea University, Sweden, from May 2007 toAugust 2009. Forty aquaria were used in theexperiment, each with a volume of 132 Lreceiving 20 L of water per hour from acirculation system. Feeding took place 8 timesper day when two portions of fish food (23 9 mgsera micrograin) were dropped on the watersurface in each aquarium. The treatments wereset up as a full factorial design with refuges andharvesting for each population (Turure andQuare). There were 5 replicates of each treatmentincluding controls, harvesting treatments, refugetreatments and refuge and harvesting treatments.In the refuge treatments, 4 refuges made of greenplastic filter material were placed in eachaquarium (weight 55 grams each) spread over avolume of 5 L (see the Appendix for a photo-graph). Two of the four refuges were positionedclose to the bottom with the help of a small pieceof metal. The harvesting targeted females largerthan 25 mm in standard length, a size at whichguppies have become efficient cannibals (Nilssonet al. 2011). The harvesting rate was 0.02 per dayand the total number of individuals to beremoved was calculated after each samplingand subsequent removal of females took placeonce a week. Each aquarium was stocked withpopulations with identical size-structure, consist-ing of 82 individuals ranging between 6.5 and 28mm in size (standard length). When populations

went extinct, aquaria were restocked with thesame size structure.

Sampling took place every fifth week and theexperiment lasted for 116 weeks resulting in 22sampling occasions in total. During sampling, allindividuals in each aquaria were removed withthe help of a large net and sorted into males,juveniles and females (males were identified byinspection of the gonopodium, juveniles andfemales were separated in a casual manner basedon size). Each group was photographed andafterwards returned to the aquaria. The photo-graphs were analysed using an image analysingprogram where all individual were counted andmeasured.

The populations were divided into 6 classes:vulnerable juveniles (6 , L � 10 mm), non-vulnerable juveniles (10 , L � 12 mm), interme-diate sized individuals (12 , L � 17 mm), if notclassified as males) small females (17 , L � 20mm), cannibals (L . 20 mm) and males. Guppyjuveniles are most susceptible to cannibalismwhen they are newborn, and as they grow in sizetheir risk of being cannibalised decreases. Theupper size limit for the vulnerable size class (10mm) was determined by the size threshold wherejuveniles were no longer susceptible to cannibal-ism, obtained from previous experiments esti-mating attack rates with both Turure and Quarepopulations (Nilsson et al. 2011).

StatisticsAnalyses of variances were performed for

overall treatment effects on the coefficient ofvariation (CV ¼ standard deviation/mean, logtransformed before the analysis) for total popu-lation density. Population that went extinct andrestocked populations were included in thisanalysis. The significance level was generallyset to P , 0.05 although the major part of themain analyses resulted in significant levels thatwere substantially lower. We therefore did notcarry out any formal corrections for multipletesting (Abdi 2010) but rather discuss cases withlimited significance when relevant. To detectcycles and estimate periodicity, autocorrelationswere estimated for total population density,density of vulnerable juveniles and cannibaldensity. Cross-correlations were estimated forcannibal vs. vulnerable juvenile density andvulnerable juvenile density vs. the density of


NILSSON AND PERSSON

intermediate sized individuals as well as vulner-able juvenile density vs. non-vulnerable juveniledensity (the two latter presented in the Appen-dix). We used cross-correlations to assess thetiming of peaks in densities of different stagessince this is a good indicator of what kind ofcycles are present.

RESULTS

Coefficient of variationThe Turure populations showed a higher

variation (coefficient of variation, CV) in totalpopulation density compared to the Quarepopulations (Fig. 1, Table 1). The presence ofrefuges decreased population variability, where-as harvest increased population variability inboth populations (Fig. 1, Table 1, Appendix).Population and refuges had very strong effectson variation in density whereas the effect ofharvest was weaker and would turn non signif-icant if any correction for multiple tests weremade.

Harvesting yieldHarvesting had a dramatic effect in the Turure

populations without refuges and all replicates ofthe Turure harvest treatment went extinct. Threeof these populations were restocked of which onewent extinct again. One replicate of the Turure

control treatment also went extinct. The totalnumber of harvested individuals was lower inthe refuge treatments (F1,16, p ¼ 0.011). Onaverage, 13 individuals in the Turure harvesttreatment, 7.8 individuals in the Turure harvestand refuge treatment, 10.8 individuals in theQuare harvest treatment and 6 individuals in theQuare harvest and refuge treatment were har-vested during one year of the experiment. Tworeplicates from the Turure harvest treatmentwithout refuges present were excluded from theanalysis as they went extinct at an early stageand hence had a very low yield.

AutocorrelationsThe effects of population origin and refuges on

Fig. 1. Average values of coefficient of variation for total population density for (A) the Quare control treatment

(QC, in white), the Quare harvest treatment (QH, white striped), the Quare refuge treatment (QR, grey), the

Quare harvest and refuge treatment (QHR, grey striped) and (B) the Turure control treatment (TC, in white), the

Turure harvest treatment (TH, white striped), the Turure refuge treatment (TR, grey) and the Turure harvest and

refuge treatment (THR, grey striped). Error bars are þ1 SD.

Table 1. Analysis of variances for treatment differences

in the coefficient of variation in total population

density. The analysis was performed after log

transformation of data.

Source df F p

Population 1 16.7 0.0002Refuge 1 21.8 ,0.0001Harvest 1 4.26 0.046Population 3 Refuge 1 0.99 0.33Population 3 Harvest 1 0.016 0.9Refuge 3 Harvest 1 1.28 0.26Population 3 Refuge 3 Harvest 1 0 1Residuals 35


NILSSON AND PERSSON

population variability were much stronger thanthe effect of harvesting. This result was partly aresult of the extinctions in several of theharvested populations and partly a result of thatthe harvesting yield was low when refuges werepresent. We will therefore focus the moredetailed analyses of population variability usingautocorrelation and cross-correlation analyses onthe effects of population origin and refuges.

The total density autocorrelations for theQuare control treatment showed a clear negativecorrelation around lag 7 (significant at the p ¼0.05 level in 4 out of 5 replicates) and a positivecorrelation around lag 15 (significant in all 5replicates) corresponding to a cycle period of 75weeks (Fig. 2A). The total density autocorrela-tions for the Turure control treatment showed anegative correlation around lag 10 (significant in4 out of 4 replicates, one replicate went extinctbefore the 10th sampling) and a tendency for apositive correlation around lag 18 (present in 4replicates, significant in one replicate). Theseresults suggest that the periodicity of the Tururepopulations (90 weeks) was longer than that ofthe Quare populations (Fig. 2A). In contrast tocontrol populations, no clear patterns were seenin the total density autocorrelations for the Quarerefuge treatment or the Turure refuge treatment(Fig. 2B). More detailed analyses of total densityautocorrelations that support the results present-ed above are given in the Appendix.

The Quare population showed substantial

peaks in juvenile densities in all treatments,although the relative magnitudes of peaks werehigher in the control treatments due to overalllower densities (Fig. 3). The peaks in vulnerablejuveniles in the Turure populations were gener-ally higher in the refuge treatment, but, therelative magnitude of these peaks was higher inthe control treatment when taking the lowdensities in controls into account. Vulnerablejuvenile density autocorrelations did not showany strong patterns for the Quare controltreatment, which was also the case for the Tururecontrol treatment (Fig. 4A). In contrast, vulner-able juvenile density autocorrelations for theQuare refuge treatment showed negative corre-lations at lag 5 and also around lag 15 (significantin 3 out of 5 replicates) and positive correlationsaround lag 10 (present in 4 replicates but notsignificant) corresponding to a cycle period of 50weeks (Fig. 4B). The vulnerable juvenile densityautocorrelations for the Turure refuge treatmentshowed similar patterns to the Quare refugetreatment described above but were less clear(Fig. 4B).

In summary, cycles with a periodicity ofapproximately one and a half years (75 weeks)were present in total density for the Quarecontrol treatment, while the periodicity in Tururecontrol was somewhat longer (90 weeks). Cycleswith the periodicity of one year (50 weeks)appeared in the density of vulnerable juveniledensity for the Quare refuge treatment and a

Fig. 2. Total density auto-correlations and lags at maximum and minimum correlations for (A) the Quare

control treatment (light grey circles) and the Turure control treatment (dark grey circles and black for the extinct

replicate), (B) the Quare refuge treatment (light grey circles) and the Turure refuge treatment (dark grey circles).


NILSSON AND PERSSON

tendency for a similar pattern was also present inthe Turure refuge treatment.

Cross-correlations between cannibaland vulnerable juvenile density

Large females are both efficient cannibals andhave a high reproductive output. We performedcross-correlation analyses with the densities ofvulnerable juveniles and cannibals to evaluate ifhigh juvenile densities co-occurred with highcannibal density. A lag of þ1 in the cannibal-vulnerable cross-correlation means that a peak incannibal density is followed by a peak injuveniles on the following sampling occasion.This is a pattern expected for dynamics driven by

cohort competition but not for cannibal-drivencycles (Claessen et al. 2000).

The Quare control treatment showed consid-erable variation between replicates in the canni-bal-vulnerable density cross-correlation. Thesignificant correlations were in general negativeand occurred at various lags (�1,�2,þ3,þ7; Fig.5A). The Turure control treatment showedpositive correlations around lag �5, althoughthe correlations would span over several lags andwere not significant. Negative correlations wereobserved at lag þ4 and þ5 (also spanning overseveral lags, significant in 2 replicates; Fig. 5A).Importantly, there were no significant positivecorrelations at lag 0 or þ1 for any of the control

Fig. 3. Densities of the vulnerable juvenile size class for individual replicates of (A) the Quare control treatment,

(B) the Turure control treatment, (C) the Quare refuge treatment, (D) the Turure refuge treatment.


NILSSON AND PERSSON

populations, which suggest that a high cannibaldensity did not result in high recruitment,furthermore suggesting the lack of fluctuationsdriven by cohort competition.

In contrast to control populations, cross-corre-lations between cannibal and vulnerable juveniledensities for the Quare refuge treatment showedpositive correlations around lag 0 (significant in 4out of 5 replicates). Negative correlations werenot as distinct, but occurred at a lag around �3(significant in 2 out of 5 replicates; Fig. 5B). TheTurure refuge treatment showed the same pat-tern as the Quare refuge treatment with positivecorrelations at a lag around þ1 (significant in 3out of 5 replicates) and negative correlations at alag around�3 (significant in 3 out of 5 replicates;Fig. 5B). More detailed analysis of the cannibal-vulnerable juvenile density cross-correlations can

be found in the Appendix.

In summary, the vulnerable juvenile densitiesand the cannibal densities were overall synchro-nised in the refuge treatments, but not in controltreatments.

DISCUSSION

Dynamics in the control treatments:cannibal-driven cycles

Recruitment regulation has been shown inseveral fish populations and may be caused byboth intra- and interspecific predation (Polis1981, Andersson et al. 2007). Cannibalism is aclassic example of density-dependence in de-layed feedback cycles already described byRicker (1954). Typically, the cycle length of thisdynamics has the periodicity of 2–4 generations.

Fig. 4. Vulnerable density auto-correlations and lags at maximum and minimum correlations for (A) the Quare

control treatment (light grey circles), (B) the Turure control treatment (dark grey circles), (C) the Quare refuge

treatment (light grey circles) and (D) the Turure refuge treatment (dark grey circles).


NILSSON AND PERSSON

Empirical examples of such cannibalistic cycleswith a periodicity substantially exceeding thegeneration time includes the salamander popu-lations studied by Wissinger et al. (2010) and theleech populations studied by Elliott (2004).Resource competition between cannibals andvictims was virtually absent, at least in thesalamander study by Wissinger et al. (2010).The dynamics may become more complex whencannibals and victims share resources, as was thecase in our guppy populations. The most studiedexample in this context is the Eurasian perch thatmay display a mixture of cannibal cycles andcohort competition (Persson et al. 2003). In otherstudied species, cannibalism is the main regulat-ing process despite that cannibals and victimsshare resources (Persson et al. 2006).

The results from our guppy populationssuggest that the dynamics in the control treat-ment represents cannibal driven cycles. Thisinterpretation can first be related to the relativelyfew and large offspring in guppies, which meansthat they are cannibalised directly from birth and

that there is a high potential for cannibals tocontrol victims. Second, the fact that highjuvenile densities did not co-occur with highcannibal densities suggests that cannibalism hada strong negative effect on recruitment in controltreatments. Third, the clear periodicity in totaldensity in both populations suggests that cycleswere present in controls. Fourth, the cycle periodin control treatments was 75–90 weeks which isin the order of 2 times the generation time(estimated from growth experiments, see theAppendix for details).

The dynamics in the control treatments in thetwo populations were overall similar althoughthe cycle period was somewhat shorter in theQuare control than in the Turure control popu-lations. Furthermore, peaks in vulnerable juve-nile density in the Quare control were muchhigher and were followed by higher and clearerpeaks in the density of non-vulnerable juvenileindividuals. We suggest that this difference incycle period between the two populations isrelated to the overall lower cannibalism in the

Fig. 5. Average values of cross correlations between cannibal density and vulnerable juvenile density for (A)

the Quare control treatment (light grey line) and the Turure control treatment (dark grey line), (B) the Quare

refuge treatment (light grey line) and the Turure refuge treatment (dark grey line).


NILSSON AND PERSSON

Quare control populations. This conclusion issupported by individual level experiments thatshowed that Quare females had lower cannibal-istic voracity than Turure females when norefuges were present (Nilsson et al. 2011). Inaddition, Quare juveniles are larger at birth andwill hence spend a shorter time in the predationwindow, which may contribute to a lower levelof cannibalism.

The effect of harvesting:extinctions and irregular dynamics

The effects of harvesting were overall substan-tially weaker than that of population origin andrefuges. Still, harvesting tended to induce ahigher variability in total population densitywhen no refuges were present. A destabilisingas well as a stabilising effect of harvesting adultshas been shown in both models and experimentsfor the cannibalistic Tribolium system (Dennis etal. 1995, 1997). Models that include both canni-balism and resource competition have shownthat harvest of large individuals in cannibalisticpopulations may lead to a destabilisation ofpopulation dynamics (van Kooten et al. 2007,2010). In stable cannibalistic populations, har-vesting may result in more cohort-driven dy-namics due to a decrease in cannibalistic pressurefrom harvesting adults. In populations withsmall offspring (and thereby a high number ofoffspring), harvesting may decrease the degree ofcannibalism to such an extent that small juvenilessuccessfully recruit into the population and areable to depress the shared resource before theybecome large enough to be efficiently cannibal-ised whereby cohort cycles may arise (vanKooten et al. 2010). Empirical support for thatharvest of adults may lead to increased recruit-ment in cannibalistic systems has been found inpopulations of several fish species such as pikeand perch (Alm 1951, Sharma and Borgstrøm2008). In cannibalistic populations with largeoffspring size (and hence few offspring) such asguppy populations, this scenario is, however, lesslikely.

In our experiment harvesting generally led toincreased population variability. In the Tururepopulations without refuges present, harvestingalso resulted in extinctions in several popula-tions. This can simply be explained by the lowpopulation densities due to cannibalism in

combination with the stochastic componentinduced by harvesting. Another aspect is thatharvesting likely decreased the reproductivepotential in the populations, since large females(which were subjected to harvest) have a highreproductive output.

The presence of cannibals that were largeenough to be harvested in the treatments withoutrefuges was likely due to the relatively lowresource competition and possibly some energet-ic gain from cannibalism. In contrast, few largecannibals were harvested in the refuge andharvest treatments. This result can be related tothat increased resource competition due to highertotal densities made it difficult for cannibals togrow as large as the harvesting limit. As aconsequence, harvest in the refuge treatmentshad hardly any effects.

The stabilising effect of refugesRefuge availability is in contemporary ecolog-

ical theory generally considered to decreasepopulation variability in predator-prey systems(Stenseth 1980, Krivan 1998). This was also thecase in our study, although the mechanisms aredifferent from predator-prey systems in thatcannibals and victims are coupled over the life-cycle. It is easy to envision that refuge additionleading to a decrease in cannibalism maystabilize cannibal-driven cycles. However, thesituation is more complex as cannibals andvictims also compete for a shared resource. Inour experiments, refuge availability clearly had astabilising effect on total population variability,but there were also indications of the presence ofpopulation fluctuations with a shorter periodlength (50 weeks) in juvenile and Quare cannibaldensities when refuges were present. The densityof female adults was also somewhat pulsed,reflected in variability in adult densities as highas or higher than in control treatments. Further-more, densities of vulnerable juveniles andcannibals in the refuge treatment were highlysynchronised. The co-occurrence of this synchro-nized juvenile-cannibal periodicity and of overallhigh densities of in particular intermediate sizedindividuals in refuge treatments points towards acyclicity induced by higher resource competitionand lower per capita food availability (observethat the inflow of food was constant; see theAppendix for details).


NILSSON AND PERSSON

In the absence of cannibalism, size-structuredconsumer-resource systems have generally beenshown to exhibit high amplitude generationcycles driven by cohort competition (Persson etal. 1998, de Roos and Persson 2003). Althoughcannibalism and refuge use have not beentheoretically analysed in combination, our resultssuggest that cannibal thinning of recruits incombination with the relatively low per capitafecundity of guppies may prevent juvenilecohorts from depressing the shared resourcelevel to such an extent that proper (highamplitude) generation cycles emerge. This resultis agreement with theoretical studies that showthat cannibalism may be stabilising in systemsthat incorporate other density dependent pro-cesses like competition (Cushing 1991).

In conclusion, both cannibalism and cohortcompetition may generate cycles in size-struc-tured populations (Claessen and de Roos 2003).Our results provide clear evidence for cannibal-istic cycles in both guppy populations in controlswhere the fluctuations were higher in the morecannibalistic population (Turure). The presenceof refuges increased population densities and ledto population fluctuations with a shorter periodand lower amplitude than the cannibalistic cyclesand these fluctuations could be related to cohortcompetition. The cyclicity of these fluctuationswas also more clear in the less cannibalisticpopulation (Quare) having a higher total density.We argue that the circumstance that cannibalismwas also present in refuge treatments preventedthe populations from exhibiting high amplitudegeneration cycles. Finally, we suggest that thesmaller offspring and the higher cannibalisticvoracity in the Turure than in the Quarepopulation were the main factors behind thepopulation differences in dynamics. The signifi-cance of this difference is reinforced by that thedifference in population variability (CV) betweenthe guppy populations was higher than thatpreviously observed between different species offish (van Kooten et al. 2010)

ACKNOWLEDGMENTS

We thank anonymous reviewers for their helpfulcomments. We are grateful to David Reznick forsending us guppies and for providing good advice.We also thank all the people who helped with thesampling and the setup of the experimental system,

especially Marten Soderquist and William Larsson.The study was financially supported by the SwedishResearch Council. The Knut and Alice WallenbergFoundation provided funding for the image analysingsystem.

LITERATURE CITED

Abdi, H. 2010. Holm’s sequential Bonferroni proce-dure. In N. Salkind, editor. Encyclopaedia ofresearch design. Sage, Thousand Oaks, California,USA.

Alm, G. 1951. Year class fluctuations and span of life ofperch. Report 33:17–38. Institute of FreshwaterResearch, Drottningholm, Sweden.

Andersson, J., P. Bystrom, D. Claessen, L. Persson, andA. M. De Roos. 2007. Stabilization of populationfluctuations due to cannibalism promotes resourcepolymorphism in fish. American Naturalist169:820–829.

Barlow, J. 1992. Nonlinear and logistic growth inexperimental populations of guppies. Ecology73:941–950.

Benoit, H. P., E. McCauley, and J. R. Post. 1998. Testingthe demographic consequences of cannibalism inTribolium confusum. Ecology 79:2839–2851.

Benoit, H. P., et al. 2000. Recruitment dynamics andsize structure in experimental populations of themosquitofish, Gambusia affinis. Copeia 2000:216–221.

Claessen, D., and A. M. de Roos. 2003. Bistability in asize-structured population model of cannibalisticfish: a continuation study. Theoretical PopulationBiology 64:49–65.

Claessen, D., A. M. de Roos, and L. Persson. 2000.Dwarfs and giants: Cannibalism and competitionin size-structured populations. American Natural-ist 155:219–237.

Claessen, D., A. M. de Roos, and L. Persson. 2004.Population dynamic theory of size-dependentcannibalism. Proceedings of the Royal Society B271:333–340.

Claessen, D., C. Van Oss, A. M. de Roos, and L.Persson. 2002. The impact of size-dependentpredation on population dynamics and individuallife history. Ecology 83:1660–1675.

Cushing, J. M. 1991. A simple model of cannibalism.Mathematical Biosciences 107:47–71.

Cushing, J. M. 1992. A size-structured model forcannibalism. Theoretical Population Biology42:347–361.

de Roos, A. M., K. Leonardsson, L. Persson, and G. G.Mittelbach. 2002. Ontogenetic niche shifts andflexible behavior in size-structured populations.Ecological Monographs 72:271–292.

de Roos, A. M., and L. Persson. 2003. Competition insize-structured populations: mechanisms inducing


NILSSON AND PERSSON

cohort formation and population cycles. Theoreti-cal Population Biology 63:1–16.

Dennis, B., R. A. Desharnais, J. M. Cushing, and R. F.Costantino. 1995. Nonlinear demographic dynam-ics: mathematical-models, statistical-methods, andbiological experiments. Ecological Monographs65:261–281.

Dennis, B., R. A. Desharnais, J. M. Cushing, and R. F.Costantino. 1997. Transitions in population dynam-ics: Equilibria to periodic cycles to aperiodic cycles.Journal of Animal Ecology 66:704–729.

Elliott, J. M. 2004. Contrasting dynamics in twosubpopulations of a leech metapopulation over 25year-classes in a small stream. Journal of AnimalEcology 73:272–282.

Fox, L. R. 1975a. Cannibalism in natural populations.Annual Review of Ecology and Systematics 6:87–106.

Fox, L. R. 1975b. Factors influencing cannibalism, amechanism of population limitation in predatorNotonecta Hoffmanni. Ecology 56:933–941.

Hastings, A., and R. F. Costantino. 1991. Oscillations inpopulation numbers: age-dependent cannibalism.Journal of Animal Ecology 60:471–482.

Krivan, V. 1998. Effects of optimal antipredatorbehavior of prey on predator-prey dynamics: Therole of refuges. Theoretical Population Biology53:131–142.

Leonardsson, K. 1991. Effects of cannibalism andalternative prey on population-dynamics of Saduriaentomon (Isopoda). Ecology 72:1273–1285.

Magurran, A. E. 1998. Population differentiationwithout speciation. Philosophical Transactions ofthe Royal Society B 353:275–286.

Nilsson, K., S. Lundback, A. Postavnicheva-Harri, andL. Persson. 2011. Guppy populations differ incannibalistic degree and adaptation to structuralenvironments. Oecologia 167:391–400.

Orr, B. K., W. W. Murdoch, and J. R. Bence. 1990.Population regulation, convergence, and cannibal-ism in Notonecta (Hemiptera). Ecology 71:68–82.

Persson, L., A. Bertolo, and A. M. De Roos. 2006.Temporal stability in size distributions and growthrates of three Esox lucius L. populations: A result ofcannibalism? Journal of Fish Biology 69:461–472.

Persson, L., A. M. de Roos, and A. Bertolo. 2004.Predicting shifts in dynamics of cannibalistic fieldpopulations using individual-based models. Pro-ceedings of the Royal Society B 271:2489–2493.

Persson, L., A. M. De Roos, D. Claessen, P. Bystrom, J.Lovgren, S. Sjogren, R. Svanback, E. Wahlstrom,and E. Westman. 2003. Gigantic cannibals driving awhole-lake trophic cascade. Proceedings of the

National Academy of Sciences USA 100:4035–4039.Persson, L., K. Leonardsson, A. M. de Roos, M.

Gyllenberg, and B. Christensen. 1998. Ontogeneticscaling of foraging rates and the dynamics of a size-structured consumer-resource model. TheoreticalPopulation Biology 54:270–293.

Polis, G. A. 1981. The evolution and dynamics ofintraspecific predation. Annual Review of Ecologyand Systematics 12:225–251.

Reznick, D., M. J. Butler, and H. Rodd. 2001. Life-history evolution in guppies. VII. The comparativeecology of high- and low-predation environments.American Naturalist 157:126–140.

Reznick, D. N., and H. Bryga. 1987. Life-historyevolution in guppies (Poecilia reticulata). 1. Pheno-typic and genetic changes in an introductionexperiment. Evolution 41:1370–1385.

Reznick, D., and A. P. Yang. 1993. The influence offluctuating resources on life-history: patterns ofallocation and plasticity in female guppies. Ecology74:2011–2019.

Ricker, W. E. 1954. Stock and recruitment. Journal ofthe Fisheries Research Board of Canada 11:559–623.

Rosenzweig, M. L., and R. H. MacArthur. 1963.Graphical representation and stability conditionsof predator-prey interactions. American Naturalist97:209–223.

Sharma, C. M., and R. Borgstrøm. 2008. Increasedpopulation density of pike Esox lucius: a result ofselective harvest of large individuals. Ecology ofFreshwater Fish 17:590–596.

Stenseth, N. C. 1980. Spatial heterogeneity andpopulation stability: some evolutionary conse-quences. Oikos 35:165–184.

van den Bosch, F., and W. Gabriel. 1997. Cannibalismin an age-structured predator-prey system. Bulletinof Mathematical Biology 59:551–567.

van Kooten, T., J. Andersson, P. Bystrom, L. Persson,and A. M. de Roos. 2010. Size at hatchingdetermines population dynamics and response toharvesting in cannibalistic fish. Canadian Journalof Fisheries and Aquatic Sciences 67:401–416.

van Kooten, T., L. Persson, and A. M. de Roos. 2007.Size-dependent mortality induces life-historychanges mediated through population dynamicalfeedbacks. American Naturalist 170:258–270.

Wissinger, S. A., H. H. Whiteman, M. Denoel, M. L.Mumford, and C. B. Aubee. 2010. Consumptiveand nonconsumptive effects of cannibalism influctuating age-structured populations. Ecology91:549–555.


NILSSON AND PERSSON

SUPPLEMENTAL MATERIAL

APPENDIX

Detailed analysis of total density autocorrelationsTo compare the lags in total density autocor-

relations, and hence the periodicity in differenttreatments, the highest (positive) correlationvalue and the lowest (negative) correlation valuewas selected for each replicate and included in anAnalysis of Variances. The first positive correla-tions at very short lags were ignored. Severalreplicates lacked positive correlations. For exam-ple, only one out of five replicates in the Tururerefuge treatment had any positive correlationexcept for the initial ones (this can be seen by thelack of data points in Fig. 4). We selected thestrongest correlation. There was no weightingdepending on the actual value of the correlation.

Positive correlationsThe Turure population had longer lag at the

highest positive autocorrelation than did theQuare population (Table A1). This pattern waslargely driven by the control treatments. Noeffects of refuges were found (Table A1), this islikely partly due to that the refuge treatmentshad few positive correlations. These findingssupport the statements in main text about cyclelength in the different populations.

Negative correlationsThere were no significant treatment effects on

the differences in lags at the most pronouncednegative correlation for total population density(Table A2). However, when comparing theperiodicity of the Turure control and the Quarecontrol treatment, excluding the extinct replicate(one Turure with negative correlation at short lagin this case) resulted in significantly longer lagfor the Turure control ( p¼ 0.0084). This supportsthat Quare control has a shorter cycle length thanTurure control, which was found when lookingat the positive correlations.

Detailed analysis of cannibal-vulnerablejuvenile cross-correlations

To compare the correlations at zero lag for thecannibal and vulnerable juvenile densities weselected the correlation value at lag 0 andperformed an analysis of variance (Table A3).

There was a strong correlation between cannibals

and vulnerable densities when refuges were

present, whereas when no refuges were present

this was not the case (Table A3).

Post hoc tests for the ANOVAon total population variability

Post hoc tests for the analysis of variances on

treatment effects on coefficient of variation for

total population density (corresponding to Table

1) can be found in Table A4.

Growth rate estimates

To check if our division into size-classes was

accurate and to obtain an estimate of the growth

rate from the vulnerable to the invulnerable stage

Table A1. Analyses of variance for treatment differ-

ences in lags at the most pronounced positive

correlation in the autocorrelations for total popula-

tion density.

Source df F p

Population 1 6.62 0.033Refuge 1 2.37 0.16Population 3 Refuge 1 0.55 0.48Residuals 8

Table A2. Analyses of variance for treatment differ-

ences in lags at the most pronounced negative

correlation in the autocorrelations for total popula-

tion density.

Source df F p


Table A3. Analysis of variance for treatment differenc-

es in the correlations at zero lag for the cannibal and

vulnerable juvenile densities.

Source df F p



NILSSON AND PERSSON

we used cross-correlation analyses between the

densities of vulnerable and non-vulnerable juve-

niles. In general the cross-correlations showed

significant positive correlations around lag þ 1

which means that a peak in vulnerable juveniles

was followed by a peak in non-vulnerable

juveniles after 5 weeks. This fits well with the

assumption that vulnerable juveniles grow into

the non-vulnerable stage in 5 weeks (one

sampling interval, results not shown).

To obtain an estimate of the growth of

vulnerable juveniles we performed cross-correla-

tions with the density of intermediate sized

individuals and vulnerable juveniles. The Quare

control treatment showed clear positive correla-

tions at lags �1 and �2 (significant in 3 out of 5

replicates), which means that a peak in vulner-

able juvenile density was followed by a peak in

intermediate sized individuals in the subsequent

sampling or after two or three sampling occa-

Table A4. Tukey’s post hoc tests for the analysis of variances on treatment effects on coefficient of variation for

total population density. Bold lettering indicates significant p-values (,0.05) and italic indicates p-values

,0.10. The figures in the Combination column represent the following: 1, Quare; 2, Turure; 0, No refuges; 3,

Refuges present; 4, No harvest; 5, Harvest.

Combination p Treatment I Treatment II

2:0–1:0 0.18 Turure without refuges Quare without refuges1:3–1:0 0.0021 Quare with refuges Quare without refuges2:3–1:0 0.93 Turure with refuges Quare without refuges1:3–2:0 ,0.0001 Quare with refuges Turure without refuges2:3–2:0 0.048 Turure with refuges Turure without refuges2:3–1:3 0.011 Turure with refuges Quare with refuges2:4–1:4 0.06 Turure without harvest Quare without harvest1:5–1:4 0.55 Quare with harvest Quare without harvest2:5–1:4 0.0006 Turure with harvest Quare without harvest1:5–2:4 0.57 Quare with harvest Turure without harvest2:5–2:4 0.4 Turure with harvest Turure without harvest2:5–1:5 0.027 Turure with harvest Quare with harvest3:4–0:4 0.12 Refuge without harvest No refuge no harvest0:5–0:4 0.13 No refuge with harvest No refuge no harvest3:5–0:4 0.34 Refuge with harvest No refuge no harvest0:5–3:4 0.0002 No refuge with harvest Refuge no harvest3:5–3:4 0.93 Refuge with harvest Refuge no harvest3:5–0:5 0.0015 Refuge with harvest No refuge with harvest2:0:4–1:0:4 0.92 Turure Control Quare Control1:3:4–1:0:4 0.4 Quare Refuge Quare Control2:3:4–1:0:4 1 Turure Refuge Quare Control1:0:5–1:0:4 0.79 Quare Harvest Quare Control2:0:5–1:0:4 0.065 Turure Harvest Quare Control1:3:5–1:0:4 0.66 Quare Harvest and Refuge Quare Control2:3:5–1:0:4 0.99 Turure Harvest and Refuge Quare Control1:3:4–2:0:4 0.035 Quare Refuge Turure Control2:3:4–2:0:4 0.95 Turure Refuge Turure Control1:0:5–2:0:4 1 Quare Harvest Turure Control2:0:5–2:0:4 0.65 Turure Harvest Turure Control1:3:5–2:0:4 0.09 Quare Harvest and Refuge Turure Control2:3:5–2:0:4 1 Turure Harvest and Refuge Turure Control2:3:4–1:3:4 0.35 Turure Refuge Quare Refuge1:0:5–1:3:4 0.016 Quare Harvest Quare Refuge2:0:5–1:3:4 0.0001 Turure Harvest Quare Refuge1:3:5–1:3:4 1 Quare Harvest and Refuge Quare Refuge2:3:5–1:3:4 0.15 Turure Harvest and Refuge Quare Refuge1:0:5–2:3:4 0.84 Quare Harvest Turure Refuge2:0:5–2:3:4 0.082 Turure Harvest Turure Refuge1:3:5–2:3:4 0.6 Quare Harvest and Refuge Turure Refuge2:3:5–2:3:4 1 Turure Harvest and Refuge Turure Refuge2:0:5–1:0:5 0.84 Turure Harvest Quare Harvest1:3:5–1:0:5 0.045 Quare Harvest and Refuge Quare Harvest2:3:5–1:0:5 0.98 Turure Harvest and Refuge Quare Harvest1:3:5–2:0:5 0.0003 Quare Harvest and Refuge Turure Harvest2:3:5–2:0:5 0.24 Turure Harvest and Refuge Turure Harvest2:3:5–1:3:5 0.31 Turure Harvest and Refuge Quare Harvest and Refuge


NILSSON AND PERSSON

sions (5–15 weeks; Fig. A1A). The Turure controltreatment also showed positive correlations atlags�1 and�2 (significant in 3 out of 5 replicates;Fig. A1B). Both harvest treatments were similarto the controls.

In the Quare refuge treatment 4 replicatesshowed significant positive correlations at lagþ 2or þ 4 (Fig. A1). This result suggests that a peak

in intermediate sized individuals is followed by apeak in juveniles after 2–4 sampling occasions(10–20 weeks). To summarise, in the controltreatments a peak in vulnerable juvenile densitywas followed by a peak in the density ofintermediate sized individuals after approxi-mately 2 sampling occasions (10 weeks) for bothpopulations. In the refuge treatments, a peak in

Fig. A2. Female growth at different food levels. The thin black line represents growth at ad libitum food levels

(from S. Auer), the fat black line represent growth at intermediate food levels, the dark grey line represents

growth at low food levels and the light grey line represents growth at very low food levels (modified from

Barlow 1992). The grey dotted lines represent the average size at first reproduction at ad libitum food levels (27

mm) and size at maturation obtained from literature (17 mm). The large black crosses mark the transition from

the vulnerable and the invulnerable juvenile stage in our experiment, verified by cross-correlations. The small

grey cross represents the intermediate size-class, based on the relationship between vulnerable juveniles and

intermediate sized individuals in cross-correlations (when no refuges were present).

Fig. A1. Average values of cross correlations between density of intermediate sized individuals (excluding

males) and vulnerable juvenile density for (A) populations without refuges, Quare in white and Turure in grey,

(B) without refuges, Quare in white and Turure in grey.


NILSSON AND PERSSON

intermediate sized individuals seemed to befollowed by a peak in vulnerable juvenile densityafter 3 sampling occasions (15 weeks).

These results suggest that juveniles grew intothe intermediate stage in 10 weeks in controltreatments. However, they were still likely to beaffected by resource competition as they did notgrow as fast as under ad libitum conditions (Fig.A2). With a size at first reproduction of 27 mm,this results in a generation time of approximately40 weeks (or lower) as the size at reproductioncan be expected to be lower when not feeding adlibitum (Fig. A2). The observed cycle period inthe control treatment (75–90 weeks) is thus 2times the generation times or more.

In the refuge treatments the periodicity was 50weeks and hence only slightly longer than thegeneration time which is expected for cohortcycles. Moreover, the high population density inrefuges treatments is expected to have led to alower per capita food availability in refuge

treatments compared to controls, hence theguppies are expected to have grown slower inthe refuge treatments making the time to reachmaturity to be longer than 40 weeks. Additionalsupport for a cohort cycle in the refuge treat-ments comes from the fact that it was difficult todetect a clear transition from vulnerable tointermediate sizes in the these treatments, whichis in line with theoretical predations on formationof a stunted cohort. Instead the transition fromintermediate sized individuals to large adults,that can reproduce and give rise to a new cohortof juveniles, was more evident.

Photograph of an aquariumWe have provided a photograph of an aquar-

ium containing refuges (Fig. A3). The feeder canbe seen to the left above the water surface. Theinlet of water is the hose in the middle and theoutlet is close to the surface at the right. Theaquarium is covered with black plastic on 3 sides.

Fig. A3. An experimental aquarium containing refuges.


NILSSON AND PERSSON

ecosphere, 4(8): 100 nilsson, k., persson, l. (2013...

Documents