MARINE ECOLOGY PROGRESS SERIES Vol. 157: 247-259, 1997 Published October 16 Mar Ecol Prog Ser

Prey switching by Acartia clausi: experimental evidence and implications of intraguild predation assessed by a model

Ingrid Gismervikll*, Tom ~ndersen~

'Section of Marine Zoology and Chemistry, University of Oslo, PO Box 1064 Blindern. N-0316 Oslo, Norway 2Section of Marine Botany, University of Oslo, PO Box 1069 Blindern, N-0316 Oslo, Norway

ABSTRACT: Sw~tching between algal (Thalasdosira we~ssflog~i)and ciliate (Strobilidium undinum) food by the marine copepod Acartia clausi was investigated In the laboratory by short incubation experiments with l4C-labeled prey. A. clausi displayed a Holling type 3 functional response (which differed significantly from a type 2 response, p c 0.05) for ciliates when there was a constant abundance of algae present, and likewise for algae when there was a constant abundance of ciliates present. The results were implemented in a mathematical model to investigate the effect of different functional responses on a simple food web comprised of nutrient, algae, ciliates and copepods. In the model, ciliates and copepods competed for resources (algae) and ciliates were also prey for copepods. This blend of predation and competition among copepods and ciliates corresponds to intraguild predation as defined by Polis & Holt (1992; Trends Ecol Evol 7:151-154). Stable solutions with all state variables present were found over a range of nutrient concentrations when the copepods displayed type 3 func- tional responses. On the contrary, when copepods displayed type 2 responses, such stable solutions were only found at very low input nutrient concentrations. Coexistence of ciliates and copepods further required that clliates had a lower threshold prey concentration for positive net growth than copepods.

KEY WORDS: Prey switching . Acartia clausi . Ciliates . Copepods . Algae . Model . Intraguild predation . Stab~lity. Coexistence

INTRODUCTION forms of the predator's diet increases disproportion- ately in comparison with the expected amount'. Cope- The feeding response displayed by a predator in pods have commonly been recognized as grazers relation to the density of prey (functional response) is which have non-increasing clearance rates with important for the stability of prey-predator relation- increasing p]-ey concentratlons (e.g.Mullin et al. 1975, ships. Holling (1959) argued that an S-shaped func- Kleppel 1993). This response, however, implies that tional response (termed type 3) has the potential to the copepods display non-variable selectivity or no regulate prey density, in contrast to rectilinear (termed selectivity at all. type 1) or curvilinear (termed type 2) functional re- While functional responses are often measured in 1 sponses. This relates to the fact that the per capita prey prey situations in the laboratory, predators in situhave mortality inflicted by a predator increases with prey a choice between several prey types. Field studies sug- concentration when the predator displays a type 3 gest that copepods are opportunistic feeders, concen- response. This response is also termed a switching trating their feeding efforts on the most abundant prey response, following the definition of Murdoch (1969): (Wilson 1973, Richman et al. 1977, Poulet 1978). This 'As a prey become relatively more abundant, switch- suggests that clearance rates should increase as the ing occurs if the relative amount which that species concentration of a specific prey item increases, and that the per capita prey mortality for the specific prey will increase when concentrations are enhanced.

0 Inter-Research 1997 Resale of full article not permitted 248 Mar Ecol Prog Ser

A type 3 response in grazers may result from differ- to the first experiment 2 Thalassiosira weissflogii cul- ent factors like reduced handling time of prey (Holling tures were inoculated, whereof one was supplied with 1959), concentration of search activity in profitable inorganic 14C (0.1 pCi ml-l). A batch of S. undinum micro habitats (Royama 1970) or optimizing of energy were fed 14C-labeled N. pyriformis daily for 6 d prior to gain (Lehman 1976). It may also be a result of switch- the second experiment. Ciliates and algae for micro- ing between 2 prey species, where the feeding effort scopical enumeration were fixed w~th2",0 and 1 % v/v on the most abundant prey is enhanced due to shifts in acid Lugol's iodine, respectively. The calanoid cope- 'search image' (Tinbergen 1960) or shifts in feeding pod Acartia clausi was collected from the Oslofjord 5 to behaviour (Marten 1973). Several copepod species 7 d before the experiments (surface temperature of have turned out to be omnivores (e.g. Stoecker & 5°C and salinity of 30%0)and acclimated to an experi- Capuzzo 1990), switching between a secondary and a mental temperature of -11°C and salinity of 25%0. The tertiary trophic position in the food web. There is evi- copepods were fed T weissflogii during acclimatiza- dence that some of these copepods have at least 2 dif- tion, and also S. undinum the day prior to the experi- ferent feeding strategies; suspension feeding on small, ment. They were kept in darkness both prior to and relatively immobile prey and ambush feeding on during the experiments. larger motile prey (Gauld 1966, Wilson 1973, Landry Approximately 20 female Acartia clausi were accli- 1981, Koehl 1984, Jonsson & Tiselius 1990, Tiselius & mated for 2 h in 500 m1 glass beakers with ciliates and Jonsson 1990, Chow-Fraser & Sprules 1992). Copepods algae added to final concentrations before an inoculum with different feeding strategies can be distinguished of labeled Thalassiosira weissflogii (Expt 1) or labeled by morphological characteristics (Schnack 1982, Strobilidium undinum (Expt 2) was added (Table 1). Landry & Fagerness 1988) as well as different swim- Only ciliate cultures where the food algae Nephro- ming patterns (Lowndes 1935, Green 1988, Tiselius & selmis pyriformis was grazed down were used. The Jonsson 1990). By filming Acartia tonsa, both Jonsson ciliates were nevertheless healthy and swimming, and & Tiselius (1990) and Kiarboe et al. (1996) have shown no change in their condition was observed during the that the time allocated to ambush and suspension feed- incubation. After 20 min the copepods were gently ing changes with the composition of food. poured onto a small sieve, quickly rinsed by filtered sea The type 3 functional response has several interest- water, and anaesthetized with MS 222. The copepods ing properties (Murdoch & Oaten 1975). Prey species were immediately counted and placed in scintillation will experience refuge or at least reduced mortality vials. Tissue solubiliser (200 1.11, SolueneTM)was added, risk below some threshold prey concentration. How- and the samples were left over night before scintillant ever, as prey concentration increases, the predator will was added. The samples were left for another night and increase its clearance rate, and may thus control prey then counted in a scintillation counter. Four different abundance. Given that the response works at realistic volumes of the labeled T. weissflogii and the labeled S. prey abundances this may hinder one prey species in undinum cultures were filtered onto GF/F filters after monopolizing resources and ensure coexistence of sev- the experiments, and the activity per cell was calcu- eral species. lated from regression curves. The regression curves for We wanted to investigate whether Acartia clausi total DPM (disintegrations per minute) as a function of (Giesbrecht), a potential switching copepod, displays volume of labeled S. undinum culture filtered were lin- type 3 functional responses for algae and/or ciliates ear. Hence, there seemed to be no losses of ciliates due when presented with a mixture of diatoms and oligo- trich ciliates. Our experimental results were incorpo- Table 1. Thalassiosira weissflogii and Strobilidium undinum. rated in a mathematical model to investigate which Concentrations in experimental beakers. Three replicates for effect copepods with different functiondl responses edch treatment have on a simplistic 4 component food web. Concentration (cells ml-')

MATERIAL AND METHODS Expt l 7 weissflogl~ 220 317 510 702 1084 1463 S. undinurn 13 13 13 13 13 13 Experiments. Two laboratory experiments were Labelled performed in December 1995. Ciliates were collected T weissflogii 123 123 123 123 123 123 from the Oslofjord (Norway) and fed a monoculture of Expt 2 Nephroselmis pyriformis (Prasinophyceae). One spe- T weissflogli 585 585 585 585 585 585 cies was isolated and identified as Strobilidium und- S. undinum 3.4 6.6 11.8 16.7 21.8 31.7 Labelled inum (Martin & Montagnes 1993) by the protargol S. undin um 1.6 1.6 1.6 1.6 1.6 1.6 staining technique (Montagnes 1993). Four days prior Gismerv~k& Andersen: Prey switching by Acartla clausl 249

to the forcing of a higher amount of water through the filter We concluded that clliates were kept on the filter, even if they did break Thus, we also sampled the cili- Ciliate egesta are assumed to be completely remin- ate and algae cultures for CN analysis on GFlF filters, eralised in the water column, while a fraction fz of and analysis was performed on a Carlo-Erba elemental copepod egesta is assumed to be lost from the water analyzer column as sedimenting fecal pellets. This is accounted Model description and incorporation of experimental for in Eq. (5) by setting the dimensionless parameter results. We defined a spatially homogeneous, chemo- fR = 1 - (1 - E,) fZ. Algal specific growth rate, P,, is stat-type model (Fig 1) comprised of 4 state variables, dependent on external nutrient concentration as inorganic nutrient (N) algae (A),ciliates (P) and cope- described by the common Monod model for nutrient pods (Z) If we, for simplicity, assume constant stoichio- limited growth, with parameters as given in Table 3: metlies In all food web components, we can express all state variables in units of the limiting nutrient (which will usually be nitrogen In a maiine system) Both copepods and ciliates feed on the same algae while The specific growth rates of the consumers, ,up and copepods are also capable of feeding on ciliates Al- pz, given by the difference between assimilation and though copepods and ciliates have different optlmal maintenance costs: prey sizes, there exists a range where the 2 groups PP = &PIP- TP (7) compete for resources (Gismervik et a1 in press) The ,P, = €212- 1, (8) mass balances coirespondlng to the flow structure Cihate ingestion rate, Ip, is assumed to follow a type 2 described In Fig 1 can be represented by the following functional response, with parameters as given in Table 3: 4 differential equations (see Tables 2 & 3 for symbol definitions and units)

d CNldt = D(CN,,, - Cl\,) - /L, CA+ RN (1) The ingestion term of copepods is more complex than that of ciliates. Preying on large moving prey (ciliates) dC,ldt = (p,- D)C, - I, C, - I,,, C, (2) and smaller non-motile prey (algae) is mutually exclu- d Cpld t = (,Up - D)Cp - Ipz C, (3) sive in the sense that the 2 prey appear to be captured by dCz/dt = (P,- D - m)Cz (4) different feeding mechanisms (see 'Introduction'). Thus copepods should feed preferentially upon the most ben- The regeneration term, R,,,, in the mass balance for eficial prey, and there is evidence that they actually do the inorganic nutrient Eq. (1) will have contributions so. According to our and others' results, some copepods from both ciliates and copepods. Since all state vari- are highly selective for ciliates (e.g.Stoecker & Sanders ables are expressed in nutrient units, the specific nutri- 1985, Stoecker & Egloff 1987, Nejstgaard et al. 1997) ent regeneration rate from a given consumer will sim- which are hypothesized to be nutritionally superior food ply be the difference between ingestion and growth: (Stoecker & Capuzzo 1990, Sanders & Wickham 1993).

Fig. 1. Flow d~agramfor a 4-compartment model : comprised of nutrient (N), algae (A),ciliates (P) and i copepods (2) Heavy llnes denote Intercompart- f mental flows (unbroken and broken lines are as- i while thin lines are boundary flows (sedlmentat~on and water exchange) 250 Mar Ecol Prog Ser 157 247-259, 1997

Table 2. Variables and process rates used in model analysis sponds to algae and ciliates being perfectly substitut- able resources in the terminology of Leon & Thumpson (1975). By the shape parameter n, the level ( Parameter Abbr. Unit 1 increasing curves of the functional response become more and External variable more convex (as illustrated in Fig. 3C for the case n = 2) C,v,," PM N Input nutrient concentration - in other words, any combination of the 2 prey spe- State variables cies will give less food intake than the same amount of Nutrient concentration Algal concentration food biomass offered as 1 species alone. It should be Ciliate concentration noticed that Eq. (10) gives a type 3 response (i.e. a Copepod concentration functional response with an inflection point) also in the Process rates single-prey situation when n > 1 - a feature that Algal growth rate ,U A d-' seems to have support in several feeding studies Ciliate growth rate !JP d- ' Copepod growth rate !Jz d- ' on Acartia spp. (see 'Discussion'). In the limit, when Ingestion rate of ciliates on algae IP d- l n + -, Eq. (10) will approach the perfect switching Ingestion rate for copepods Iz d- l functional response predicted by the theory of optimal Ingestion rate for copepods on algae d- ' foraging for a consumer with 2 mutually exclusive Ingestion rate for copepods on ciliates kz d-l Nutrient regeneration from grazers RN pMN d-l feeding modes (e.g. Stevens & Krebs 1986). The grazing losses experienced by each prey type must add up to the total ingestion rate given by Since a major purpose of this model is to investigate Eq. (10), such that if IASZ= SAIZand I,,, = SpIz, then we the effects of the copepod functional response on must have that SA + Sp = 1 (or, equivalently, Sp= 1 - SA). dynamic properties of the food web, we wanted a gen- If we choose eral representation that is able to cover the basic Holling type 2 and 3 responses as special cases. This can be easily accomplished by introducing a shape parameter n as suggested by Matsuda et al. (1986): then copepod ingestion rates of algae and ciliates can be written as

For the case of n = 1 and Cp = 0 (i.e.only algal prey present), Eq. (10) will be functionally identical to the (CP/C'P.Z)n IP,Z = I'z (13) Holling type 2 response Eq. (9). If n = 1 and both algae l + (c*/C'~,z)" + (Cp/C'p,z In and ciliates are present (i.e. CA > 0 and Cp > 0), then level curves for constant copepod ingestion rate (IZ) When n = 1 we will have that consumption is directly will be straight lines in the (CA,Cp) plane (illustrated in proportional to relative, effective food concentration Fig. 3A in the 'Results' section). This situation corre- (sensu Vanderploeg et al. 1984) - as would be ex-

Table 3. Parameter values used in model analysis - Parameter Abbr. Value Unit Source

Dilution rate D 0.01 d- ' Arbitrary Max. growth rate for algae P ,A 1 d-l Eppley et al. (1969) Monod parameter for algae c ',v 0.5 PM N Eppley et al. (1969) Max. ingestion for ciliates 1'P 2.5 d-I Scott (1985) Max. ingestion rate for copepods I'z 0.5 d' l Experiments, see Table 5, 'Results' Incipient lim. algae conc. for ciliates C'A.P 0.5 PM N Jonsson (1986).recalculated Ks Incipient lim. algae conc. for copepods c'~.~ 0.7 PM N Experiments, see Table 5 Incipient 11m. ciliate conc. for copepods c 'P. z 0.3 PMN Experiments, see Table 5 Resp~rationloss for ciliates r~ 0.5 d.' Crawford & Stoecker (19961 Respiration loss for copepods rz 0.1 d-' Kierboe et al. (1985) Mortality rate for copepods m 0.01 d-' Carlotti & Nival (1992) Assimilation efficiency for ciliates FP 0.8 Stoecker (1984) Assimilation efficiency for copepods EZ 0.8 K~orboeet a1 (1985) Faecal pellet fraction of non-assimil. food f z 0.5 Arbitrary Switching shape parameter n 1,2 Experiments, see Table 5 Gismervik & Andersen: Prey switching by Acartia clausi 25 1

pected for a consumer with invariant prey selection. Table 4. Carbon and nitrogen content of experimental organisms Level curves for constant prey ingestion rate will then be straight lines (as illustrated in Fig. 3B). If, on the 1-lg c Pg N other hand, n > 1, then Eqs. (12) & (13) imply that the relative prey consumption will be a power function of Thalassiosira weissflogii 9.0 X 10~~1.6 X IO-~ Strobilidium undin urn 1.1 X IO-~ 2.5X 10~~~ relative prey abundance: Acartia clausi" 3.6 0.7 "Values from December of the previous year (Gismervik 1997a) bCalculated from C:N ratio of 4.4 (see 'Results') such that the most abundant prey will be consumed disproportionately to its relative abundance (as illus- trated in Fig. 3D). This property conforms with the RESULTS classical definition of a switching predator, according to Oaten & Murdoch (1975). Experiments and fitted curves Least squares fitting of Eqs. (12) & (13) to experimen- tal data was performed with the non-linear fit option of Carbon and nitrogen content of the experimental or- the statistics package JMP 3.1 (SAS Institute, Inc.). ganisms are shown in Table 4. The nitrogen content of Model sirnulations and bifurcation analysis were the ciliate samples was below the detection limit, thus implemented in the numerical analysis package MAT- the nitrogen content was calculated from a C:N ratio of LAB 4.2 (The MathWorks, Inc.). Details on analytical 4.4 (Gismervik et al. in press). Ingestion and clearance and computational procedures for equilibrium points rates of Acartia clausi on diatoms and ciliates are and local stability analysis are described in Appen- shown in Fig. 2. The curves that gave the best fit to our dix 1. Full MATLAB source code for simulations and data (see below) are included. Below the incipient lim- bifurcation analysis is available from the authors upon iting prey concentration, calculated total ingestion rate request. for A. clausiwas highest when a type 2 compared to a

0 500 1000 1S00 2000 T weissjlogir abundance (cells ml-') T: ~verssflogiiabundance (cclls m]-')

600 -

400--

0 Y I 0 10 20 30 40 0 10 20 30 40 S undrrru,,~abundance (cclls rnl'l) S. undrr?u~nabundancc (cclis rnlI)

Fig. 2. Acartia clausiingestion and clearance rates on algae (A and B respectively) and ciliates (C and D respectively) as a func- tions of prey biomass. Solid lines are given by Eqs. (12) & (13), using the parameters fitted by non-linear least squares (Table 5) Mar Ecol Prog Ser 157: 247-259, 1997

Copepod functional response = Type 2 type 2 response ingest algae in a fixed proportion to Total ingestion (d.') Algal ingestion available food, a switching predator concentrates its feeding efforts on the most abundant prey (Fig. 3B, D). Using Eqs. (12) & (13) we could investigate the mag- nitude of n and decide whether our experimental re- sults corresponded to a type 2 or a type 3 functional response. Non-linear regressions were run in order to find the parameter values of Itz, and the switching shape parameter n which fitted best to our experimental data. Best fit was obtained by a sigmoid curve, with a switching shape parameter of n = 2.1 (Table 5). The corresponding curves for ingestion and 0 l 2 0 1 2 clearance rates are shown in Fig. 2. The curve fitted to Algal biornass (PMN) Algal biomass (PM N) the data deviated significantly from a Holling type 2 functional response, as the switching shape parameter response = Type was significantly different from 1 (Table 5). Total ingestion (d-l) Algal ingestion (d-l) The maximum ingestion rate of copepods, I'Z,was adjusted to a temperature of 20°C by a QIo of 2.3 4 1 (Durbin & Durbin 1992) before it was entered into the model. The ingestion curves that were entered into the 5 0.8 V) model are shown in Fig. 4A (type 2 functional re- g 0.6 0 sponse, n = 1) and Fig. 4C (type 3 functional response, a n = 2). The corresponding curves for copepod net 2 0.4 .--m growth rate as a function of algal biomass are com- G 0.2 pared to ciliate net growth in Fig. 4B, D. The threshold algal concentrations for net growth of ciliates were 0 o 1 2 lower than for copepods irrespective of the type of Algal biomass (PMN) Algal biomass (PMN) functional response displayed by the copepods. The threshold for copepods was higher when a type 3 func- Fig. 3. Acartia clausi. Contour plots of specific ingestion rate tional response was used compared to a type 2. Deduc- as function of algal and ciliate biomass for a copepod having either (A, B) a type 2 or (C, D) a type 3 functional response and tion of stationary points with both ciliates and cope- parameter values glven by Table 3. (A, C) Total ingestion rate pods together revealed that ciliates should have a given by Eq. (10); (B, D) algal ingestion rate given by Eq. (12) lower threshold algal concentration for positive net growth than copepods in order for both grazers to coexist in this simplified system (Appendix 1). Al- type 3 functional response was used (Fig. 3A, C). though such data are scarce, there is evidence that Above this concentration, copepods with switching be- some common pelagic ciliates may have lower thresh- haviour had the highest food intake (Fig. 3A, C). This olds, or at least thresholds of the same magnitude as can also be seen in Fig. 4A, C. While copepods with a copepods (Gismervik et al. in press).

Table 5. Parameter estimates and their 95% confidence limits (in parentheses) for best non-hear fit of Eqs. (12) & (13) to exper- mental data. A switching shape param.eter of n = 2 was used in the second run (non-llnear fit for 3 parameters)

Parameter Abbreviation Estimate 95% CL

Non-linear fit for 4 parameters, df = 32, RMSE = 0.0261 Maximum ingestion rate Isz(d-l) 0.23 (0.17, 0.58) Incipient limiting algae concentration c',,= (P9 c I-') 52 (25, 196) Incipient limiting ciliate concentration C'P,Z(P9 c I-') 18 (8.72) Switching shape parameter n 2.1 (1.2, 3.4) Non-linear fit for 3 parameters. df = 33, RMSE= 0.0257 Maximum ingestion rate Isz(d-l) 0.24a (0.20, 0.31) Incipient Limiting algae concentration c',,* (PS c I-') 55 (32, 81) Incipient limiting ciliate concentration C'P,~(Pg c 1-'1 19 (11, 28) aRate at experimental temperature of 11°C. For model runs it was adjusted to 0.5 d-l by a Qlo of 2.3 (see 'Results' section) Glsrnervik & Andersen: Prey switching by Acartia clausi

Copepod functional response = Type 2

- 0.5 ..A...... j ...... -I-.- B

0 0.5 1 1.5 2 2.5 Algal biomass (wM N) Algal biomass (PM N)

Fig. 4. (A, C) Copepod inges- Copepod functional response = Type 3 tion rate as function of algal biomass when ciliates are ,o,5.~ ...... absent [(A) type 2, (C) type 3 5 functional response]. (B, D) u 04. Copepod net growth rate as 2 function of algal biomass when . 0.3. ciliates are absent [(B)type 2, g (D) type 3 functional response]. Ciliate net growth as function g of algal biomass when cope- $ o-,. pods are absent is shown for cornpanson. Symbols and para- meters are defined in Tables 2 0 0.5 1 1.5 2 25 0 0.2 0.4 0.6 0.8 & 3 and in Appendix 1 Algal biomass (PM N) Algal b~omass(PM N)

Model analysis nutrient concentrations (Fig. 5). In this interval the coexistence between ciliates and copepods was also Bifurcation diagrams (Figs. 5 & 6) show the location the only locally, asymptotically stable solution. Thus and local stability of the stationary points of a dynamic ciliates will in this interval be able to successfully system as functions of an external parameter. Appen- invade a community consisting of only algae and cope- dix 1 describes the analysis of the stationary points of pods, and vice versa. Increasing nutrient input concen- the model given by Eqs. (1) to (4) when the external tration above this interval leads to the deterministic nutrient input concentration (CNin)is taken as the bi- extinction of the ciliate population, because the ciliates furcation parameter. A common feature of all bifurca- are no longer able to compensate for the increasing tion diagrams for this kind of systems is that below predation losses imposed by the copepod population. some low level of nutrient supply neither of the herbi- The coexistence between ciliates and copepods will vores is able to persist, thus resulting in a degenerate thus in this case be very fragile with respect to changes stationary state with algae as the only biotic compo- in nutrient supply. Copepods are able to maintain a nent. Above this threshold level, bifurcation diagrams stable equilibrium up to an intermediate level of nutri- for the present model have 3 major branches corre- ent supply, above which there are no locally stable sponding to stationary points with biomasses of either solutions. Local stability is lost by a Hopf bifurcation to ciliates, copepods, or both being non-zero (labeled I, 11, a limit cycle with alternations between algal blooms and 111, respectively, in Figs. 5 & 6). and periods of intense overgrazing, as in the classical Bifurcation analyses were performed on 2 different 'paradox of enrichment' described by Rosenzweig cases with identical model parameters (as given by (1971). Table 3), except for the switching shape parameter n, In contrast, when a type 3 functional response was introduced in Eq. (10), which was set either to 1 or 2 applied for copepod ingestion, stable coexistence (corresponding to copepods having either type 2 or between ciliates and copepods occurred even at high type 3 functional responses). When a type 2 functional input nutrient concentrations (Fig. 6). Stability loss response was applied for copepod ingestion, non-zero occurred instead over a limited range at low nutrient solutions for both ciliates and copepods were only pos- concentrations (0.5 to 1.5 FM). Simulations showed sible over a very narrow interval at very low input that this stability loss resulted in a limit cycle with Mar Ecol Prog Ser 157: 247-259, 1997

0 5 10 15 0 5 10 15 Input nutrient conc. (pM) Input nutrient conc. (pM) Fig. 5. Bifurcation diagram for the 4-compartment mo- del (Fig. 1) when copepods are assumed to have type 2 response (see Appendix 1 for details on the bifurcation analysis). Branch labels de- note stationary states with I: ciliates alone; 11: copepods alone; 111: both ciliates and copepods present. Dashed lines: unstable stationary 0 5 10 15 0 5 10 15 states; solid lines: stable Input nutrient conc. (pM) Input nutrient conc. (pM) stationary states much less vigorous oscillations than the previous case range, only the solution with both ciliates and cope- (oscillation period 10 to 15 d, relative amplitude ~50% pods was locally stable, meaning that neither predator of mean), and that ciliates and copepods were able to species will be able to suppress the invasion of the coexist in this limit cycle when the nutrient supply was other one. Copepod switching between ciliate and sufficient to support copepod persistence. Above this algal food kept the ciliate biomass on a constant level

0 5 10 15 0 5 10 15 Inputnutnent conc. (pM) Input nutrient conc. (pM) Fig. 6. Bifurcation diagram "D . for the 4-compartment mo- 2. m del (Fig. 1) when copepods -52 2. areresponse assumed (see to Appendixhave type 31 for details on the bifurcation 5. analysis). Branch labels de- .5 note stationary states with I: u0 Q " ciliates alone; 11: copepods Q alone; 111: both ciliates and u , , , . copepods present. Dashed 0 ...... n ,,. -...... I lines: unstable stationary 0 5 10 15 o /5 10 1s states; solid lines: stable Input nutnent conc. (pM) Inputnutnent cone. (pM) stationary states G~smerv~k& Andersen: Prey switch~ngby Acartja clausi 255

and thereby ensured a constant grazing pressure on There is scattered evidence that the large Calanus the algal community. The algal biomass at steady state spp. may switch between prey types. C. pacificus was then higher than when ciliates acted alone, but increased its clearance rates when available nauplii lower than when copepods were the only grazers carbon increased compared to algal carbon (Landry (Fig. 6B). The coexistence of ciliates and copepods was 1981). C. helgolandicus fed disproportionately on large insensitive to changes in nutrient input whenever the Ditylum cells (Richman & Rogers 1969), and recently, nutrient supply rate was sufficient to support the exis- Nejstgaard et al. (1997) found that preference for tence of a copepod population. Emiliania huxleyi by C. finmarchicus increased as the abundance of E. huxleyi increased. This could be inter- preted as switching behaviour. Some freshwater cope- DISCUSSION pods and daphnids may also display type 3 feeding responses (Chow-Fraser & Sprules 1992, Wickham Experimental results 1995). There is also evidence that some copepods display Acartia clausi increased its clearance rates with type 3 functional responses in 1 prey situations. increasing prey concentrations when there was an Acartia spp. tend to reduce their clearance rates for alternating prey present - both in Expt 1, when the algae below some threshold food abundance (Deason algal biomass was increased, and in Expt 2, when the 1980, Kerboe et al. 1985, Paffenhofer & Stearns 1988, ciliate biomass was increased. These results are con- Durbin & Durbin 1992, Kierboe et al. 1996). For dia- sistent with type 3 functional responses, and are prob- toms, the clearance rate decreases when the abun- ably due to A. clausi switching its feeding effort from dance is below -100 to 500 cells ml-l, or -4 to 40 pg algae to ciliates (from suspension feeding to ambush C 1-l. For smaller flagellates the level where clearance feeding) when ciliate biomass is increased, and vice rate starts to decrease seems to be higher (Kerboe et versa [see Fig. 4 in Marten (1973) for an informative al. 1985: maximum clearance rate at 150 1-19 C 1-' presentation of the feeding upon 2 alternate prey]. for Rhodomonas baltica; Gismervik unpubl. results: There are but a few experiments which reveal a type 3 maximum clearance rate at about 5000 cells ml-' for response in copepods. Jonsson & Tiselius (1990) found Tetraselmis damir). According to Paffenhofer & enhanced clearance rates for A. tonsa at increasing Stearns (1988) the decrease in clearance rate in A. concentrations of the ciliate Strombidium reticulatum. tonsa at low algal food concentrations may be due to When both ciliates and the algae Cryptomonas baltica few chemoreceptors, and an inability to change the were present, the copepods doubled their time de- flow field in order to re-route cells that are displaced voted to suspension feeding behaviour when the bio- towards the copepod. Thus, this initial increase in mass of the algae was increased from 330 to 1100 pg clearance rates for algae found by some investigators C 1-l. A similar result was obtained by fierboe et al. is generated by switching from non-feeding to feed- (1996). They also showed that the clearance rate ing activities (Marten 1973). A feeding threshold may of algae Thalassiosira rveissflogii by the copepod A. also exist for other copepods, like Calanus spp. (Par- tonsa depended on the availability of ciliate prey. sons et al. 1969, Frost 1975). However, in their study, the clearance rate of ciliates Copepod feeding threshold for ciliates is consider- was almost constant independent of the biomass of ably lower than for algae (0.3 pg C I-' for Acartia algae. The authors argued that this discrepancy could tonsa feeding on Strornbidium reticulatum), and max- be related to the fact that this particular ciliate species imum clearance rate is achieved at low concentrations was not able to escape the feeding current of A. tonsa. (Jonsson & Tiselius 1990). A tonsa's ability to feed at Thus the copepod might have been able to clear cili- high rates at such low concentrations of ciliates may ate~also in the suspension feeding mode (Kierboe et be due to specialization for carnivory by numerous al. 1996). mechanoreceptors on the first antenna (Paffenhofer & The clearance rates obtained for Acartia clausi in Stearns 1988, Jonsson & Tiselius 1990). Prey moving our study were low. This may partly be due to low in the vicinity of the copepod will be detected and activity of the animals during winter (they were col- captured. lected at a surface temperature of 5"C, and acclimated to an experimental temperature of 11°C). It can also indicate that our estimates of radioactivity per food Model results: predictions cell (ciliates and algae) were overestimated. A biased estimate of the activity per food item should, however, The experiments were designed to reveal functional not interfere with the type of functional response I-esponses in copepods when presented with 2 prey obtained. types, 1 motile and 1 non-motile. We chose Thalas- 256 Mar Ecol Prog Ser 157: 247-259, 1997

siosira weissflogii as this alga has been frequently tional responses in the copepod Acartia clausi, we used in feeding experiments with copepods, because have no data to verify the outcome of the model. Cope- it is too large to be eaten by the ciliate, and to reduce pods in our model feed on both ciliates and algae, the effects of size between the motile and non-mot~le while they simultaneously compete with ciliates for prey. However, in the model we wanted to explore food (intraguild predation). Thus, net phytoplankton interactions between copepods and ciliates feeding on have been excluded from our model. In natural sys- the same prey, while at the same time they were tems, increased nutrient input will generally lead to an engaged in a predator-prey relationship. This blend of increase in the biomass of net phytoplankton, while competition and predation has been termed intraguild less changes appear in the microbial food web (Nielsen predation by Polis & Holt (1992). Intraguild predation & Kiarboe 1991, Kiarboe 1993. Riegman et al. 1993). theory (see Polis & Holt 1992) has predicted several Hence, increased nutrient input will tend to increase effects of predation in a 4 component system as the importance of the traditional food web, or overflow shown in Fig. 1: coexistence between the predators system, compared to the maintenance system, in the (copepods) and the consumers (ciliates) is promoted, terminology of Riegman et al. (1993). Our model re- given that the consumers (ciliates) are the superior sults partly explain how the microbial food web may competitors for the shared prey (algae), and predation stay basically the same through density dependent can indirectly increase prey abundance (algae bio- selectivity of ciliates by copepods, which allow the mass) by depressing the most efficient consumers remaining ciliate community to keep up a constant (ciliates). Both effects are demonstrated by our model, grazing pressure on the nanoplankton. Our model however, a switching response of the predator is an does not include the heterotrophic part of the microbial additional prerequisite for coexistence over an ex- community (DOM-bacteria-heterotrophic flagellates), tended range of nutrient supply (Figs. 5 & 6). When thus caution should be exercised when extrapolating copepods are present, ciliate biomasses are main- our results to natural systems. In contrast to the smaller tained at low stable levels, and consequently the algal plankton (both auto- and heterotrophic), net plankton biomasses are higher and more stable than when cili- escapes grazing from most protozoans, and the grazing ates act alone (Fig. 6B, C). Several field studies sup- pressure from copepods is generally too low to control port the theory of predation control of ciliates, except net phytoplankton blooms (Kiarboe 1993). As our during periods of low copepod abundance, e.g. spring model does not include net phytoplankton, we may not (Smetacek 1981, Kivi et al. 1993, Nielsen & Ki~rboe predict the outcome of increased nutrient input to a 1994). natural system. Nevertheless, due to the high selectiv- A strength of our experiments and model is that ity of ciliates over algae, we would expect switching they encompass realistic abundances. The experi- copepods to control ciliate abundances even when net ments indicate that copepods switch towards ciliates phytoplankton is included. Thus, the increased nutri- when biomasses pass 14 pg C I-' (-13 small ciliates ent concentrations should be channeled to net phyto- ml-l), and the model simulations indicate stable cili- plankton, and not smaller phytoplankton which is con- ate biomasses between 1 and 10 ml-'. This is in trolled by grazers, as predicted by e.g. Riegman et al. accordance with observed abundances/biomasses of (1993). ciliates in coastal waters, although maximum values In this study we have empirically investigated the may be higher (Stoecker et al. 1989, Leakey et al. functional response of a switching copepod, and by 1993, Nielsen & Kierrboe 1994, Edwards & Burkill applying a simple model we have strengthened our 1995, Gismervik 1997b). Furthermore, it may partly notion of copepods as important structuring compo- explain the high diversity of ciliate species with nents of the pelagic food web. Switching copepods are seemingly identical trophic position In plankton potential key species that should be given special samples, as one species will be hindered in monopo- attention in future studies. lizing resources. When discussing the paradox of the plankton, Hutchinson (1961) suggested that predation could promote coexistence of prey species. Indeed, Acknowledgements. We thank Wenche Eikrem and Peter the recognition of the omnivore habits of marine crus- Tiselius for providing the algae (Nephroselmis pyriformis and Thalasnosira weissflogi~ respect~vely),and David J. S. Mon- taceans as well as discovery of numerous mixo- and tagnes for help with the protargol stain of clliates. We are also heterotrophic protists further suggests that predation grateful to Per J. Feeravig for analysing the C and N content of may be important in structuring plankton comrnuni- the ciliates and algae, and to Stein Kaartvedt for critically ties. reading the manuscript. This study forms a part of the research programme on marine pollution (PMF] financed by Our model is a simplified system, investigating the the Norwegian Research Council (NFR), and was also sup- impact of different functional response of copepods on ported by the programme on manne resources and environ- food web stability. While we have investigated func- ments (MAREMI). Gismervik & Andersen: Prey 5

Appendix 1. Stationary points and local stabil~tyanalysis

Stationary polnts are deflned by the non-l~nearequation vectors (C;,,, CiZ, 0. CmZ)as functions of CN,,, which can system resulting from setting the time derivatives (l e left be used to construct the branches of the bifurcatlon diagram hand sides) of Eqs (l) to (4)equal to zero In general, such a that correspond to copepods alone system of non-linear equat~onsmay have zero, 1, or many Stationary points with ciliates and copepods together. solutions which may be locally stable or not depending on For the persistent system (I e tor which all state va~lables the signs of eigenvalues of the Jacobian matrlx corlespond- are positive), Eq (4) must still be valid, as In the previous lng to a given stationary polnt A blfurcation analysls deals case but now subject to the condition that copepod con- with how the locatlon and local stability of statlonaly polnts sumptlon of algae and cillates together 1s sufficient to bal- change in response to an external parameter For the system ance losses In other words, steady state concentratlons of Eqs (l) to (4) the most obvious choice of bifurcation pala- algae and ciliates (C', and Cmp)must be such that the com- meter IS the ~nputnutrient concentration C, ,,,, which deter- bined consumption rate of both prey is the same as on algae mines the nutnent supply rate to the system Inspection of alone at concentration C the system Eqs (1) to (4) reveals that lt must have seve~al non-pers~stentsolut~ons colresponding to 1 or more of the btate variables belng zero Both the trivial solutions when both predators are extinct (C', = CmZ= 0 where C'denotes This requllement impl~esthat all stationary polnts w~th the value of state variable Cat a stat~onarypoint), and solu- ciliates ~lndcopepods present must be such that C; < C', tlons where only 1 of the consumers (elther c~liatesor cope- Furthermore, we must also requlre that the algal concentra- pods) has non-zero abundance are important blanches of t~onis higher than what is necessary for ciliates to exlst in the bifurcatlon dlagram Some of these non-perslstent solu- the absence of copepods if they are going to persist when tlons will be investigated first suffering addlt~onallosses due to copepod predation that is, Stationary points with ciliates alone. When copepods ale C', > C', p The constraint C', 1, < C', < C ', ; means that absent (Cmz- 0) Eq (3)implies that cillate growth rate must coexistence of ciliates and copepods In the system descnbed be exactly balanced by dilution/advection losses (D),~vhlch here is only possible when the communit) composition rule again means that algal biomass must be equal to the thresh- C', P < C; is satisfied, i e when cil~atesare superior to old concentration for positive net growth for cillates copepods in pure competition for algal prey Computation of stationary points can be performed as above, but thls t~mestartlng out from an arbitrary algal bio- mass C ; ,,< C ; < C 1 With C 1 glven, we can solve Eq (A5) From here, one mlght go on and solve the steady state for CeP We can also compute p 'P by substltutlng C; Into Eqs (1) & (2) for CmpandC', as funct~onsof C,,,, by brute Eqs (7) & (g),and Im,,,by substltutmg C', and CaplntoEq (13) force Th~sapproach leads to a quadratic equation where -which agaln means that we can solve Eq (3) for Cml only 1 solution is feaslble in the sense that it corresponds to an lnorganlc nutrlent concentration C'&) > 0 Alternatively one mlght flnd the feasible solutlon directly by taking an approach employed by Andersen (1997),where one chooses Having solved for C',, Cnp,and Cmz,we can solve Eq (2) for an arbitrary algal growth rate p',, such that D 5 p', < p', p', ]after computing I., and I;,, from Eqs (7)& (12)]. Given ,U ;, C; p, and l', [from substituting C;, Into Eq (g)], l.,, C> +I.,,,, C; we can solve Eq (2) for the ciliate biomass p: = D+ (~71 C.., From p >we can solve the Monod equation (Eq. 6) for C'!,,, and, finally, solve Eq. (l) for C&,,,,as in Eq. (A3).Repeating this Given ,U', we can also solve the Monod equation (Eq G) procedure for different values of C; ylelds stationary points foi the correspond~nginorganic nutr~entconcentration C',, (C:,., C;,,z, Cmp,Cmz) as functions of C,\, ,,, whlch can be used Flnally, from substltutlng lep,pmp, and C'pinto Eq (5)to yleld to construct the branches of the blfurcdtion diagram that the regeneration rate R',, we can solve Eq (1)for our bifur- col-respond to coexistence between clliates and copepods. cation parametei, the input nutnent concentration C, ,,, Local stability analysis. Analytical local stability analysis of the 4-dimensional system studied here could In princ~ple C,\,,, = D-'(,LI;C; - R',) + Cm,\. (A31 have been performed by extracting the Routh-Hurwitz Repeating this procedure for different values of p:, yields determinants of the 4th order chai-acter~sticpolynomial of stational-y state vectors (C;, C;., Pp, 0) as function of the Jacobian, as described in standard textbooks on C,,,,, which can be used to construct the branches of the dynamic systems (e.g Luenberger 1979) As this procedure blfurcation diagram that correspond to ciliates alone yields some very unwieldy and rather uninformative Stationary points with copepods alone. In analogy with expressions for the present system, we have instead chosen the previous case, the steady state mass balance Eq. (4) a numerical approach Analytical expressions for the ele- impl~esthat copepod growth rate must be exactly balanced ments of the Jacobian matrix of the system are found by by dilution/advect~on (D) and mortahty (m,) losses, which taking partial derivatives of the nght hand sldes of Eqs. (1) again means that algal blomass must be equal to the thresh- to (4) with respect to the state variables For a given set of old concentration for positive net growth for copepods: parameters and a glven stationary state vector (C'\, C;,z, C',, CmZ),the eigenvalues of the Jacobian matnx can thus be found numer~callyby standard methods (such as the eig function in MATLAB). This means that all stationary If we proceed as In the previous case, we can for a given points can be classlf~edas stable or not depending on the D 5 p h < ,U',, solve the Monod equation (Eq G)for C', Fur- eigenvalues of the Jacobian matrix at that part~cularpoint thermole, we can in analogy with Eqs (A2) & (A3) solve (all elgenvalues having negative real parts Implies local sta- Eqs (2) & (1) for Cazand C, ,,, respectivel) Repeating this bility), such that the bifurcation d~agramcan be divided into procedure for different values of p; yields stationary state stable and unstable branches as shown in F~gs.5 & G 258 Mar Ecol Prog Ser 157: 247-259, 1997

LITERATURE CITED and grazing control of the Baltic plankton community dur- ing annual succession. Limnol Oceanogr 38:893-905 Andersen T (1997) Pelagic nutrient cycles. Herbivores as Kleppel GS (1993) On the diets of calanoid copepods, Mar sources and sinks. Ecological studies 129. Springer Ver- Ecol Prog Ser 99:183-195 lag, Berlin Koehl MAR (1984) Mechanisms of particle capture by cope- Carlotti F, Nival S (1992) Moulting and mortality rates of pods at low Reynolds numbers: possible modes of selective copepods related to age within stage. experimental results. feeding. In: Meyers DG, Strickler JR (eds) Trophic interac- Mar Ecol Prog Ser 84:235-243 tions within aquatic ecosystems. AAAS Selected Sympo- Chow-Fraser P, Sprules WG (1992) Type-3 functional re- sium. Westview Press Inc, Boulder, p 7-27 sponse in limnetic suspension-feeders, as demonstrated Landry MR (1981) Switching between herbivory and car- by in situ grazing rates. Hydrobiologia 232:175-191 nivory by the planktonic marine copepod Calanus pacifi- Crawford DW, Stoecker DK (1996) Carbon content, dark res- cus. Mar Biol 65.77-82 piration and mortality of the mixotrophic planktonic ciliate Landry MR, Fagerness VL (1988) Behavioural and morpho- Strombidium capitatum. Mar Biol 126:415-422 logical influences on predatory interactions among marine Deason EE (1980) Grazing of Acartia hudsonica (A. clausi) on copepods. Bull Mar Sci 43~509-529 Skeletonerna costatum in Narragansett Bay (USA): influ- Leakey RJG, Burkill pH,Sleigh MA (1993) Planktonic ciliates ence of food concentration and temperature. Mar Biol 60: in Southampton Water: quantitative taxonomic studies. 101-113 J Mar Biol Assoc UK 73579-594 Durbm EG, Durbin AG (1992) Effects of temperature and food Lehman JT (1976) The filter feeder as an optimal forager, and abundance on grazing and short-term weight change in the predicted shapes of feeding curves. Limnol Oceanogr the marine copepod Acartia hudsonica. Limnol Oceanogr 21501-516 37:361-378 Leon JA, Thumpson DB (1975) Competition between two Edwards ES, Burkill PH (1995) Abundance, biomass and dis- species for two complementary or substitutable resources. tribution of microzooplankton in the Irish Sea. J Plankton J Theor Biol 50:185-201 Res 17:771-782 Lowndes AG (1935) The swimming and feeding of certain Eppley RW, Rogers JN, McCarthy JJ (1969) Half-saturation calanoid copepods. Proc Zoo1 SOCLond 22:687-715 constants for uptake of nitrate and ammonium by marine Luenberger DG (1979) Introduction to dynamic systems. The- phytoplankton. Limnol Oceanogr 14-912-920 ory, models & applications. J Wiley & Sons, New York Frost BW (1975) A threshold feeding behaviour in Calanus Marten GG (1973) An optimization equation for predation. pacificus. Limnol Oceanogr 20:263-266 Ecology 54:92-101 Gauld DT (1966) The swimming and feeding of planktonic Martin AJ, Montagnes DJS (1993) Winter ciliates in a British copepods. In: Barnes H (ed) Some contemporary studies Columbian fjord: six new species and an analysis of ciliate in marine science. George Allen & Unwin, London, putative prey. J Eukaryot Microbial 40~535-549 p 313-334 Matsuda H, Kawasaki K, Shigesada N, Teramoto E, Riccardi LM Gismewik I (1997a) Stoichiometry of some marine planktonic (1986)Switching effect on stability of the prey-predator sys- crustaceans. J Plankton Res 19:279-285 tem with three trophic levels. J Theor Biol122:251-262 Gismervik I (1997b) Implications of zooplankton stoichiome- Montagnes DJS (1993) A quantitative protargol stain (QPS) try on distribution of N and P among planktonic size frac- for ciliates and other protists. In: Kemp PF, Sherr BF, Sherr tions. J Plankton Res 19~343-356 EB, Cole JJ (eds) Handbook of methods in aquatic micro- Gismervik I, Andersen T, Vadstein 0 (1997) Pelagic food bial ecology. Lewis Publishers, Boca Raton, p 229-240 webs and eutrophication of coastal waters: impact of graz- Mullin MM, Stewart EF, Fuglister FJ (1975) Ingestion by ers on algal communities. Mar Pollut Bull 33:22-35 planktonic grazers as a function of concentration of food. Green CH (1988) Foraging tactics and prey selection patterns Limnol Oceanogr 20:259-262 of omnivorous and carnivorous calanoid copepods. Hydro- Murdoch WW (1969) Switching in general predators: experi- biologia 167/168:295-302 ments on predator specificity and stability of prey popula- Holling CS (1959) The components of predation as revealed tions. Ecol Monogr 39:335-354 by a study of small mammal predation of European pine Murdoch WW, Oaten A (1975) Predation and population sta- sawfly. Can Ent 91:293-332 bility. Adv Ecol Res 9:l-131 Hutchmson GE (1961) The paradox of the plankton. Am Nat Nejstqaard JC, Gismervik I, Solberg P (1997) Feeding and 15:137-145 reproduction by Calanus finmarchicus, and microzoo- Jonsson PR (1986) Particle size selection, feeding rates and plankton grazing during blooms of diatoms and the coc- growth dynamics of marine planktonic oligotrichous ciliates colithophore Emiliania huxleyi. Mar Ecol Prog Ser 147: (Cihophora: Oligotrichina). Mar Ecol Prog Ser 33~265-277 197-217 Jonsson PR, Tiselius P (1990) Feeding behaviour, prey detec- Nielsen TG, Kierboe T (1991) Effects of a storm event on the tion and capture efficiency of the copepod Acartia tonsa structure of the pelagic food web with special emphasis on feeding on planktonic ciliates. Mar Ecol Prog Ser 60:35-44 planktonic ciliates. J Plankton Res 13.35-51 Kierboe T (1993) Turbulence, phytoplankton cell size, and the Nielsen TG, Kierboe T (1994) Regulation of zooplankton bio- structure of the pelagic food webs. Adv Mar Biol 29:l-72 mass and production in a temperate, coastal ecosystem. 2. ffierboe T, Mehlenberg F, Hamburger K (1985) Bioenergetics Ciliates. Limnol Oceanogr 39:508-519 of the planktonic copepod Acartia tonsa: relation between Oaten A, Murdoch WW (1975) Switching, functional response, feeding, egg production and respiration, and composition and stability in predator-prey systems. Am Nat 109: of specific dynamic action. Mar Ecol Prog Ser 2645-97 299-318 Kierboe T, Saiz E, Viitasalo M (1996) Prey switching behav- Paffenhofer GA, Stearns DE (1988) Why is Acartia tonsa iour in the planktonic copepod Acartia tonsa. Mar Ecol (Copepoda: Calanoida) restricted to nearshore environ- Prog Ser 143:65-75 ments? Mar Ecol Prog Ser 42:33-38 Kivi K, Kaitala S, Kuosa H, Kuparinen J, Leskinen E, Lignell Parsons TR, LeBrasseur RJ, Fulton JD, Kennedy OD (1969) R, Marcussen B, Tamminen T (1993) Nutrient Limitation Production studies in the Strait of Georgia. Part 11. Sec- Gismervik & Andersen: Prey swltch~ngby Acartia clausi 259

ondary production under the Fraser river plume. February Steele JH (1974)The structure of marine ecosystems. Harvard to May 1967. J Exp Mar Biol Ecol3:39-50 University Press, Cambridge, MA Polis GA, Holt RD (1992) lntraguild predation. the dynamics of Stevens DW, Krebs JR (1986) Foraging theory. Monographs in complex trophic interactions. Trends Ecol Evol 7.151-154 behaviour and ecology. Pnnceton University Press, Prince- Poulet SA (1978) Companson between five coexisting species ton, NJ of marine copepods feeding on naturally occurring partic- Stoecker DK (1984) Particle production by planktonic clliates ulate matter. Limnol Oceanogr 23:1126-1143 Limnol Oceanogr 29:930-940 Richman S, Heinle DR. Huff R (1977) Grazing by adult estuar- Stoecker DK, Capuzzo JM (1990) Predation on Protozoa: its ine calanoid copepods of the Cheasapeake Bay. Mar Biol importance to zooplankton. J Plankton Res 12:891-908 42:69-84 Stoecker DK. Egloff DA (1987) Predation by Acartia tonsa on Richman S, Rogers JN (1969) Feeding of Calanus helgolandi- planktonic ciliates and rotifers. J Exp Mar Biol Ecol 110: cus on synchronously growing populations of the marine 53-68 diatom Ditylum bnghtwellii. Limnol Oceanogr 14:701-709 Stoecker DK, Sanders NK (1985) Differential grazing by Acar- Riegman R, Kuipers B, Noordeloos AAM, Witte HJ (1993) tla tonsa on a dinoflagellate and a tlntinnid. J Plankton Res Size-different~alcontrol of phytoplankton and the struc- 7:85-100 ture of plankton communities. Neth J Sea Res 31.255-265 Stoecker DK, Taniguchi A, M~chaelsAE (1989) Abundance of Rosenzweig ML (1971) Paradox of enrichment: destabillzation autotrophic, mixotrophic and heterotrophic planktonic cih- of exploitation ecosystems in ecological time. Science 171: ates in shelf and slope waters. Mar Ecol Prog Ser 50:241254 385-387 Tinbergen L (1960) The dynamics of insect and bird predation Royama T (1970) Factors governing the hunting behaviour in pine woods. Arch Neerl Zoo1 13:259-472 and selection of food by the great tit. J Anim Ecol 39: Tiselius P, Jonsson PR (1990) Foraging behaviour of six cala- 619-668 noid copepods: observations and hydrodynamic analysis. Sanders RW, Wickham SA (1993) Planktonic protozoa and Mar Ecol Prog Ser 66:23-33 metazoa: predat~on,food quality and population control. Vanderploeg HA, Scavia D, Lieb~gJR (1984) Feeding rate of Mar Microb Food Webs 7:197-223 Diaptomus sicilis and its relation to selectivity and effec- Schnack S (1982) The structure of the mouth parts of cope- tive food concentration in algal mixtures and in Lake pods in Kiel Bay. Meeresforsch 29:89-101 Michigan. J Plankton Res 6.919-941 Scott JM (1985)The feeding rates and efficiencies of a marine Wickham SA (1995) Cyclops predat~onon ciliates: species ciliate Strombidium sp., grown under chemostat steady specific differences and functional responses. J Plankton state conditions. J Exp Mar Biol Ecol 90:81-95 Res 17:1633-1646 Smetacek V (1981) The annual cycle of protozooplankton in Wilson DS (1973) Food size selection among copepods. Ecol- the Kiel Bight. Mar Biol 63:l-11 ogy 54:909-914

Editorial responsibility, Otto Kinne, Submitted: January 27, 1997; Accepfed: July 14, 1997 Oldendorf/Luhe, Germany Proofs received from a uthor(s): September 19, 1997