On the different nature of top-down and bottom-up effects in pelagic food webs

Transcription

1 Freshwater Biology (22) 47, On the different nature of top-down and bottom-up effects in pelagic food webs Z. MACIEJ GLIWICZ Department of Hydrobiology, University of Warsaw, Warsaw, Poland SUMMARY 1. Each individual planktonic plant or animal is exposed to the hazards of starvation and risk of predation, and each planktonic population is under the control of resource limitation from the bottom up (growth and reproduction) and by predation from the top down (mortality). While the bottom-up and top-down impacts are traditionally conceived as compatible with each other, field population-density data on two coexisting Daphnia species suggest that the nature of the two impacts is different. Rates of change, such as the rate of individual body growth, rate of reproduction, and each species population growth rate, are controlled from the bottom up. State variables, such as biomass, individual body size and population density, are controlled from the top down and are fixed at a specific level regardless of the rate at which they are produced. 2. According to the theory of functional responses, carnivorous and herbivorous predators react to prey density rather than to the rate at which prey are produced or reproduced. The predator s feeding rate (and thus the magnitude of its effect on prey density) should hence be regarded as a functional response to increasing resource concentration. 3. The disparity between the bottom-up and top-down effects is also apparent in individual decision making, where a choice must be made between accepting the hazards of hunger and the risks of predation (lost calories versus loss of life). 4. As long as top-down forces are effective, the disparity with bottom-up effects seems evident. In the absence of predation, however, all efforts of an individual become subordinate to the competition for resources. Biomass becomes limited from the bottom up as soon as the density of a superior competitor has increased to the carrying capacity of a given habitat. Such a shift in the importance of bottom-up control can be seen in zooplankton in habitats from which fish have been excluded. Keywords: biomanipulation, bottom-up, Daphnia, fish feeding, food web Introduction One of the most fundamental questions in the early days of zooplankton studies, centred on the relative importance of competition and predation. The two factors used to be looked on as mutually exclusive, so the question was often asked in a conclusive way as to whether zooplankton abundance would be controlled by the limitation of growth and reproduction as a Correspondence: Z. Maciej Gliwicz, Department of Hydrobiology, University of Warsaw, Warsaw, 2-97 Warsaw, Poland. gliwicz@hydro.biol.uw.edu.pl result of short food supply or by enhanced mortality through predation. These contrasting views were most apparent between those plankton ecologists involved in the International Biological Program focussing on productivity, and those taking a more evolutionary approach, mostly Hutchinson s students who had been inspired by Ivlev s (1955, 1961) book on the Experimental ecology of the feeding of fishes and Hrbáček s (1962) paper on zooplankton in relation to the fish stock. However, the opposing views soon started to be reconciled. An important impetus came from Hrbáček s (1962) fishpond observations, which were 2296 Ó 22 Blackwell Science Ltd

2 Top-down versus bottom-up effects 2297 expanded to lake zooplankton by Brooks & Dodson (1965) and formalised as the size-efficiency hypothesis. Although the spirit of the confrontation was still much alive at the Dartmouth College workshop on the Evolution and ecology of zooplankton communities in 1979 (Kerfoot, 198), the gap between the food and predation explanations was being closed. Eventually, the two approaches were combined successfully, as reflected by publications such as the Effects of food availability and fish predation (Vanni, 1987) and the Relative importance of food limitation and predation (Lampert, 1988). Two decades after the pioneering papers by Hrbáček (1962) and Brooks & Dodson (1965), the importance of both food limitation and predation had been widely accepted by zooplankton ecologists working at both the community (Fig. 1a) and population level (Fig. 1b). Thus, the abundance and specific structure of zooplankton communities has been perceived as being controlled from both the top down and the bottom up: by predation, because of species-specific vulnerability to predators such as planktivorous fish, and by food levels, because of species-specific efficiency in food utilisation (Fig. 1a). The density and age structure of the population would be controlled by predation because of different age-specific mortality, and by food levels because of different body-size-dependent abilities to utilise food (Fig. 1b). The notion of combined predation and food limitation effects had implications for the way community and population structure would be viewed. According to McQueen et al. (1989) and their bottom-up: top-down theory, the trophic level biomass control is determined by the combined impacts of predation and energy availability. According to Lampert (1988), the population density of Daphnia species would be determined by the combined impacts of food limitation and predation. This view has been imprinted in our minds, and similar reasoning has been commonplace in recent publications on zooplankton communities and populations (e.g. Sommer, 1989; Lampert & Sommer, 1997). The same reasoning was introduced into the study of individual life histories and behaviour, and could often be found in depictions of life in pelagial zones as life between the never-ending hazards of starvation and risks to predation (e.g. Gliwicz, 21). The parity of top-down and bottom-up impacts on behavioural and life-history traits, especially body size at first reproduction, has become a key assumption in studies on the costs of antipredator defences in zooplankton and fish (Fig. 1c): the life history and behaviour of an individual is assumed to be controlled by predation, because of body-size-specific vulnerability to sizeselective predators such as planktivorous fish, and by food levels, because of body-size dependent abilities to compete for food (food-threshold concentration). Besides being the two most evident factors of natural selection, the top-down and the bottom-up impacts also seem equally important for the selection of an appropriate phenotype among a range of phenotypes available within a plastic genotype. (a) (b) (c) PREDATION PREDATION PREDATION Body-size dependent vulnerability Mortality Body-size dependent predation risk ZOOPLANKTON ABUNDANCE AND COMMUNITY STRUCTURE POPULATION DENSITY INDIVIDUAL LIFE HISTORY AND BEHAVIUOR Energy-transfer efficiency and body-size-related superiorityin resource competition FOOD LIMITATION Reproduction FOOD LIMITATION Body-size dependent food-threshold concentration FOOD LIMITATION Fig. 1 Diagrammatic representation of the parity of bottom-up (food limitation) and top-down impacts (predation) on zooplankton abundance and community structure (a), population density and age structure (b), and individual behaviour and life histories (c). Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

3 2298 Z. Maciej Gliwicz The concept depicted in Fig. 1 now appears to be widely accepted, whether at the community, population or individual level. Inherent in it is the tacit assumption that the nature of the two impacts is the same with only the direction, down or up, differing (see Reynolds, 1994; Drenner & Hambright, 1999; Carpenter et al. 21; Benndorf, 22; McQueen et al. 21). For two reasons, however, this assumption is incorrect. The first reason is that top-down and bottom-up forces affect differently the behaviour of an individual animal (e.g. a Daphnia or a fish), which trade off increased safety against decreased feeding rates. The second reason is less apparent and has been largely overlooked in the top-down versus bottom-up debate. It relates to the fact that rates (e.g. growth rates) are controlled from the bottom-up whereas state variables (e.g. density) at both the population and community level are controlled from the top-down. The importance of this fundamental difference has emerged from recent field studies on fish behaviour in an experimental biomanipulated lake, and on the role of prey abundance in prey selectivity in planktivorous fish. It is further supported by comparisons of zooplankton communities in the presence and absence of fish. These points are discussed in detail in the following sections. An individual s and a population s perspective We know intuitively that risking life is different from risking hunger. The risk of becoming subject to predation may become lethal within seconds, while the risk of starvation may persist for days or weeks with future compensation always being possible. Compensation might be readily achieved as soon as food levels have increased, as an effect of animals refraining from intense feeding in food-proficient but predation-risky areas. The possibility of compensation might account for the difficulty in devising a common currency for life-history or behavioural decisions when considering both the risk of predation and the hazards of starvation. The major difference between the two is in the likelihood of a mistake becoming fatal. It is high in the case of predation, but many mistakes in regard to food limitation could be allowed within an individual s life span (McNamara & Houston, 1986; Lima, 1998). This difference may be the reason why behavioural responses to increased predation risk tend to be stronger than those to decreased food levels. Following the pioneering work by Werner & Gilliam (1984) on size-structured fish populations, the two disparate quantities have often been compared successfully when they were converted to the common currency of fitness. Dynamic modelling of state-dependent decision-making under the risk of predation has been successfully developed to tackle the problem of the relative importance of top-down and bottom-up impacts for animal behaviour, especially in fish and zooplankton (Mangel & Clark, 1988). However, the common currency of fitness has not solved the problem that the nature of top-down and bottom-up effects is different. While feeding rate, individual growth rate and reproductive potential may all be assessed in energy units, predation risk can only be asserted as a probability in regard to the sole undivided life of an individual that can either be alive or dead. For an individual, satiation can take any value between and 1%. Long periods with an empty gut (zero satiation) can be compensated for in future when food levels increase again. However, at the individual s level, survival can never be lower than 1%. Therefore, the hazards of starvation and the risk of predation cannot be compared with a common currency, for instance as a per cent increase in feeding rate and risk of predation. This obvious incompatibility might be the reason why the disparity has never been ignored at the level of the individual. The situation is different at the population and community level, because mortality and reproduction readily combine with each other. The common currency is the individual. The effect of food limitation is reflected in the birth rate, b, and the effect of predation in the death rate, d, the two merging into the intrinsic rate of population-density increase (r ¼ b ) d). For this reason, the sandwich or squeeze-in idea of full symmetry between topdown and bottom-up impacts (Fig. 1) has been approved so readily for the population and community level. However, the illusion of full symmetry at the population level is revealed as soon as it is recognised that the state variables are controlled from the top down while the rates of change are controlled from the bottom up. This fundamental difference has accidentally emerged from our recent data collected within an unsuccessful biomanipulation project (Gliwicz et al., Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

4 Top-down versus bottom-up effects ; Gliwicz, Rutkowska & Wojciechowska, 2; Gliwicz & Dawidowicz, 21). The argument leading to this conclusion is developed more fully in the following sections. Modifying the feeding behaviour of fish with an alarm substance The principal objective of the project was to determine whether manipulation of fish abundance (reducing the density of planktivorous fish) might be substituted by a manipulation of fish behaviour (reducing feeding rates of planktivorous fish), by frightening fish with an alarm substance (skin preparation after Von Frish, 1941). This idea originated from observations that the fear of predation can lead to reduced feeding rates because planktivorous fish hide in the littoral zone or aggregate and remain in deepwater refuges (for review see Lima, 1998), where zooplankton is scarce and difficult to detect (Gliwicz & Jachner, 1992). We predicted that the impact of frightened fish on zooplankton would be reduced significantly, thus allowing for an increase in zooplankton density and mean body size. We expected that following application of the alarm substance smelt (Osmerus eperlanus L.) and roach (Rutilus rutilus L.) would tend to remain aggregated in their daytime refuges of the hypolimnion (smelt) or among the littoral vegetation (roach) during dusk, when both species normally feed most intensely in our lakes (Gliwicz & Jachner, 1992, 1993). Roach was found to be a more convenient subject for in situ manipulations than smelt for three reasons. First, roach had displayed very regular diel habitat shifts, especially in lakes free of smelt and other pelagic fish. They spent the daytime among the littoral vegetation in large aggregations and disintegrated in the evening as individual fish surged offshore to feed on Daphnia (Fig. 2), causing Daphnia abundance to increase with increasing distance from shoreline (Szynkarczyk, 2). Secondly, the required large quantities of alarm substance were more easily obtained from roach. Skin preparation were needed for treating a lake area of 8 2 ha. This corresponds to up to 1 kg of live fish from commercial catches for a single treatment (Gliwicz et al., 1998). Thirdly, roach performed better in captivity, thus allowing for many successful laboratory experiments before work in the lake commenced. Laboratory tests on roach and bleak (Alburnus alburnus L.) brought clear evidence that fish responded to the predator odour and alarm substance by aggregating, hiding in vegetation, and reducing feeding rate (Fig. 3; see also Hölker et al. 22). The field experiment was run in three interconnected lakes in north-eastern Poland. The lakes were very similar to each other (area 8 87 ha, maximum depth m, Secchi disc transparency ). One was used as the experimental (treatment) lake and the Fig. 2 Example of an evening change in near-shore roach distribution in an experimental lake treated with alarm substance (3 August 1997) starting with typical daytime distribution at 19 : 26 h, and ending an hour after sunset, at 21 : 3 h, when all daytime fish aggregations disintegrated and dispersed, and the majority of individual roach moved into the pelagial zone to forage offshore closer to the lake surface. Some roach escaped the echosounder when staying in the upper 2 m. Each HADAS-generated echogram of 5 pings covers 2 m (3 min; after Gliwicz & Dawidowicz, 21). Depth (m) Distance (m) 3 2 Sunset 21 3 Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

5 23 Z. Maciej Gliwicz Alarm substance added mm mm 6 8 mm Fig. 3 Example of fright response to alarm substance addition in roach of three size categories shown as reduction in feeding rate in per cent of initial food intake before alarm substance addition (arrow) to treatment aquaria (filled circles) compared with reference aquaria (empty circles) (mean from five replicate experiments with different roach individuals; details in Jachner & Janecki, 1999). Different response times of roach of different body size to overcome fear and maximise feeding again after exposure to an alarm substance is another example of sizestructured interactions in fish (see Persson et al. 1991; 1996). two remaining ones as reference lakes. In the summers of 1996 and 1997, one of the two ends of the experimental lake received alarm substance (roach skin preparation concentrate) to a final concentration equal to that used in earlier laboratory experiments such as those shown in Fig. 3 (6 1 )5 cm 2 of roach skin area per 1 litre, or 2 1 )7 mg hypoxanthine-3(n)- oxide per 1 litre). The alarm substance was mixed throughout the epilimnion, down to 4 m depth, by pumping it into the wake of the propeller of a cruising boat on a high-speed slalom. Hydroacoustic surveys following the treatments showed the hypothesised response on many occasions. Although the daytime aggregations were breaking apart in the evening when most fish moved into the pelagial zone (Fig. 2), the overall fish density in the evening was significantly lower offshore (Fig. 4) and the mean depth of roach rushing offshore was greater in the treatment than in the reference area (Fig. 5). Having succeeded in frightening roach and manipulating fish distribution in the lake, we also expected to see the effects of the weakened impact of fish predation on zooplankton and water transparency in the experimental lake. In particular, we anticipated: 1 a mass exodus of fish from the experimental to the adjacent reference lake (to check this, all fish moving out of and into the experimental lake were counted in the connecting stream several days before and after the treatment during both day and night); 2 roach intestines to be significantly less filled with zooplankton in the treatment than in the reference area (roach were trawl-sampled several days before and after the treatment, intestines immediately dissected, fixed and later analysed); 3 a higher density of the most vulnerable (i.e. larger) cladoceran species between the treatment and reference areas (plankton was sampled at eight stations along the experimental lake s long axis, identified and sized); 4 a higher zooplankton abundance in the experimental lake than in the two reference lakes (weekly triplicate plankton samples were taken from each lake s centre throughout the seasons). None of these four predicted responses were observed and the hypothesised enhancement of water transparency by modifying fish behaviour had eventually to be discarded, as had been foreseen by fellow disbelievers from fish-ecology circles. First, no increase was noted in the number of fish leaving the experimental lake, nor was a decrease in fish entering the lake from the reference lakes observed. Instead, the opposite response was observed following application of the alarm substance. The likely reason for this response is a change in fish-depth distribution, i.e. the frightened fish were Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

6 Top-down versus bottom-up effects 231 Depth (m) NW Reference 3 Jul 31 Jul 1 Aug 2 Aug 3 Aug Treatment SE Distance (m) Fig. 4 Example of fright response of wild roach in an experimental lake (right panels), observed at dusk (2 : 3 h) as a decreased fish density following addition of alarm substance in the south-eastern end of the lake on 1 August Densities in the north-western end of the lake, which was used as a reference, are shown on the left (details in Gliwicz & Dawidowicz, 21). The difference in roach distribution between the two areas became apparent 1 day after the treatment. Echograms were generated by HADAS from data recorded by an EY-M 7 khz echosounder, each covering 2 m traversed in 3 min (2 : 3 2 : 33 h) when near-shore roach daytime aggregations started to disintegrate (see Fig. 2). pushed down to the deeper strata (Fig. 5) and cut short from the half-metre-deep outflow. This happened, for example, after the treatment on 9 August 1996, in the north-western end of the experimental lake. The number of roach leaving the lake declined from pretreatment values of 6 7 fish day )1 (up to fish h )1 in the middle of the night) to 1 2 fish day )1 (up to 1 fish h )1 in the middle of the night), while the numbers of roach entering the lake were unaffected (Gliwicz et al., 1998). Secondly, although after each treatment nearly 3 roach intestines were inspected from trawl samples taken in both areas of the lake, no difference in feeding intensity was observed in any body-size category (Gliwicz et al., 1998). Variability in the roach diet was unusually high (Fig. 6), and the mean prey selectivity index was very similar for all five major prey species, an unusual observation. For example, the index was nearly the same for the two Daphnia species (details in Wiśniowska, 1999), although different body sizes should have translated Depth (m) into higher vulnerability to predation by roach (Fig. 7). Thirdly, the difference between the treatment and reference areas in the abundance and mean body sizes of a dominant zooplankton prey, Daphnia cucullata, was never very great nor long-lasting. Such a difference could only be detected a day or two following the treatment (details in Gliwicz & Dawidowicz, 21). Fourthly, neither the density nor the reproduction in zooplankton prey differed distinctly between the lakes. Two Daphnia species were examined thoroughly for these effects (Fig. 8). The only significant difference that was detected was a slightly higher D. hyalina density in the experimental lake compared with the two reference lakes, after several treatments with alarm substance in July and August Otherwise, the densities of all three D. hyalina populations were constant and similar in all lakes. The similarity was even more striking in a smaller prey, D. cucullata, which is an order of magnitude more abundant than D. hyalina in all lakes (details in Gliwicz et al., 2). Species-specific population-density thresholds? The constant population density of both Daphnia species in the three lakes, and the fixed density difference between the two species, gave us a hint to the nature of the disparity between top-down and bottom-up impacts. Population densities remained constant, although the intensity of reproduction differed greatly among the lakes and months with different food levels (Gliwicz et al., 2). It was also uniform along the long axis of the experimental lake and highly akin to those observed in 14 neighbouring lakes showing a wide range in food levels (assessed as chlorophyll a concentrations in the size fraction <5 lm; details in Gliwicz, 21). These observations are not unique. Other coexisting Daphnia species also have been found with fixed or species-specific density levels in many lakes (e.g. Kasprzak, Lathrop & Carpenter, 1999). However, the phenomenon has not attracted much attention in the literature and plausible answers to the questions have not been proposed until recently. One possible answer is that the population density of a given cladoceran species is fixed from the top-down by fish predation at a speciesspecific population-density threshold level, irrespective Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

7 232 Z. Maciej Gliwicz Fig. 5 Example of fright response of wild roach in the evening (2 : : 54 h). Mean depth of the roach population along the long axis of an experimental lake from its north-western to south-eastern end (13 18 m on distance scale) 1 day before the treatment with alarm substance (31 July) and the day after (2 August 1997). The mean depth of the roach population was greater at the north-western end of the lake before, on 31 July, reflecting the persisting effect of previous treatment on 11 July. Data were generated by HADAS from 2 to 1 m depth echos recorded with a SIMRAD EY-M 7 khz echosounder along a standard transect from the north-western to the south-eastern corner of the lake (squares, )1 SD), and on reverse (circles, +1 SD) when fish were already much closer to the surface in the fading light of dusk (details in Gliwicz & Dawidowicz, 21). Fig. 6 Example of high variability in food content in five individual roach of the same body size (8 1 cm) from the same pelagicseine trawl sample from the north-western part of an experimental lake treated with alarm substance. Fish were caught between 22 : and 22 : 3 h on 31 July Diverse multi-specific diet is shown on the left; uniform, single-species diet on the right. Food diversity in individuals 1 and 2 was even greater than can be seen, as both guts had cyclopoid copepods, Daphnia hyalina and Leptodora kindtii in numbers too small to be visible on the scale shown (details in Wiśniowska, 1999). Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

8 Top-down versus bottom-up effects Ivlev Vanderploeg & Scavia Mean Bosmina 1 D. cucullata D. hyalina Leptodora Chaoborus Roach body length (cm) Fig. 7 Food selectivity index for individual roach as a function of body length, with Ivlev (1961) scatter plots on the left and Vanderploeg & Scavia (1979) scatter plots and mean values on the right (n ¼ 264 for each prey category). Five major food categories were distinguished: Bosmina (three species combined), Daphnia cucullata, D. hyalina, Leptodora kindtii and larvae of the phantom midge, Chaoborus flavicans. All 264 intestines were taken from roach trawl-sampled on July and 3 July)4 August 1997 (details in Wiśniowska, 1999). Ind.1 4 m 2 ( 1 m) MAY JUL JUN AUG SEP 1996 Fig. 8 Mean population densities of Daphnia cucullata (open circles) and D. hyalina (filled circles) in an experimental lake (solid line) and two neighbouring reference lakes (dashed and dotted lines) throughout Note the logarithmic scale to show order of magnitude differences between the population densities of D. cucullata (body size at first maturation ¼.58 mm) and D. hyalina (.8 mm). For all densities starting from mid June (i.e. excluding the period of spring increase), 99% confidence intervals are shown as dotted area (details in Gliwicz et al., 2). Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

9 234 Z. Maciej Gliwicz of the level of food limitation, the rate of somatic growth of an individual, the reproductive effort in the population, and the maximum rate of population increase at a given food level. A possible mechanism for this phenomenon has been suggested elsewhere (Gliwicz, 21). It relates to the way in which dominant planktivorous fish assess the density of alternate prey. The assessment depends on the reactive distance of the fish, that is the distance at which a foraging fish can see the prey item. If the reactive distance differs for two prey species, the fish may perceive no difference in densities when the more conspicuous prey is far less abundant than the less conspicuous prey. For example, if the reactive distance for one species is twice as great as for the other, as is the case for D. hyalina and D. cucullata (Sliwowska, 2), the water volume in which the Daphnia can be seen by an individual fish, the so-called reactive field volume (i.e. a sphere with a radius equal to the reactive distance; Wetterer & Bishop, 1985), would be up to an order of magnitude greater for the larger Daphnia species (Fig. 9). This difference should be reflected in different densities of the two prey species in the lake, as was actually found in a range of lakes and various seasons (Fig. 8). Moreover, at the 1 : 1 ratio of the species-specific densities of the two Daphnia species, the selectivity index for the two alternate prey items should not differ, because fish would shift from one prey to the other as soon as the perceived density difference deviates from 1 : 1, corresponding to a real density ratio of 1 : 1 (Fig. 9). This ratio was found in our gut-content data set for roach (Fig. 7), suggesting that prey vulnerability is not only generated exclusively by the properties of an individual, but also by the properties of a population. Prey choice was thus not only related to the profitability of a single prey item, but also to the rewards resulting from the density of the prey population (planktivorous fish section). Planktivorous fish: selective or general predators? Are planktivorous fish selective or generalist predators? Our data of roach gut-contents would seem to show that, although the fish are selective, they should also be considered generalist predators. They are selective in that they would consume the prey species Fig. 9 Diagrammatic representation of the difference in reactive distance (i.e. the distance at which foraging fish would see prey) for two prey categories, which results in relative reactive field volumes for, and thus relative prey density assessment by, a foraging planktivore such as roach. As a 2 : 1 difference in the radius of the two spheres gives a 1 : 1 difference in volume, foraging fish would assess densities of two prey categories as equal when relative prey densities differ 1-fold, as they do in case of the two Daphnia species shown in Fig. 8 (details in Gliwicz, 21). whose individuals are most conspicuous and most rewarding. They are also, however, generalist predators that tend to feed upon the most abundant prey, shifting to the prey category that is most rewarding as a result of both the properties of an individual prey item and the density of the prey population. The two predatory behaviours are not mutually exclusive. On the contrary, they must be combined and co-ordinated with each other in every decision concerning prey choice, regardless of whether the subject of choice is a prey individual or a prey population. For example, fish may choose a prey item based on size, as in the apparent size model of O Brien, Slade & Viniard (1976), in which a planktivorous fish is assumed to select prey actively, pursuing whichever prey appears largest. In addition, two feeding modes must be compromised in a decision to switch from one prey population to another, such as choosing to feed on a prey category Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

10 Top-down versus bottom-up effects 235 that has just been found to be more rewarding, as in the model of Murdoch (1969) and Murdoch, Avery & Smyth (1975). More rewarding prey may be the prey category (species or single ontogenetic stage) that is relatively more abundant or offers a higher net energy gain, given the energy or time invested in a successful individual encounter and or found by the fish to be most efficient to capture. Each of the three reasons should be an equally valid justification for switching from prey item A to prey item B as soon as B becomes more rewarding. Most experimental evidence showing a switch to more rewarding prey comes from laboratory studies examining predator switching between different patches of prey or between different feeding habitats, rather than between prey categories in a homogeneous mixture of different prey, the real situation encountered by a planktivorous fish in the field. Field observations focus on the switch between habitats of different food profitability (e.g. Werner, Mittelbach & Hall, 1981), or different risk to predation (e.g. Hall et al., 1979; Gliwicz & Jachner, 1992), rather than on the switch from one food category to another in response to a change in their relative abundance (Murdoch & Bence, 1987). Although it is well known that planktivorous fish will switch from one zooplankton species to another on a seasonal (Eggers, 1982) or daily basis (Hall et al., 1979), the importance of prey relative abundance has mostly been ignored in the quest for understanding the phenomenon of prey switching and of food selectivity in planktivorous fishes in general. The focus was on prey relative body size, and the question of prey abundance was confined to the importance of overall prey density, and an increase in density that would enhance selectivity for more conspicuous prey. The phenomenon that selectivity is increased via an increase in the overall density of prey has been known since the pioneering work of Ivlev (1961), and was experimentally explored by Werner & Hall (1974). These authors allowed prey categories, different D. magna instars, to differ in body size and overall abundance, while keeping the relative abundance of the prey categories constant. Relative prey density was often touched on in theoretical approaches (Gerritsen & Strickler, 1977; Eggers, 1982; Wetterer & Bishop, 1985; Giske, Huse & Fiksen, 1998), but was ignored in experimental and field studies on planktivorous fish. Experimental and field studies examined selectivity in response to prey body-size and overall prey abundance. Some of the recent studies focused on effects of body size and zooplankton abundance in regard to the functional response (e.g. Johnston & Mathias, 1994). An exception is Luo, Brandt & Klebasko (1996), who were able to predict the size frequencies of zooplankton prey in anchovy stomachs from the ambient zooplankton body-size frequencies found in the habitat (mid- Chesapeake Bay). The mutual importance of prey body size and prey population density as two determinants of food selectivity in a typical planktivore has recently gained attention, following the realisation that a lake with an indigenous fish fauna has species-specific population-density thresholds for each cladoceran prey category. The threshold density is inversely related to the individual susceptibility of each cladoceran species to predation, which is most strongly related to body size at first reproduction (species-specific population-density thresholds section). The gut contents of roach from our experimental lake showed high variability in individual roach diets (Fig. 6) and in the selectivity index for different prey categories (Fig. 7) probably because of frequent switching among prey categories. Part of this variability may be an effect of switching on a daily basis, especially when the switch is to or from phantom midge larvae (Chaoborus spp.), which were frequently the sole prey found in the roach guts (Fig. 6). Such a switch may require a shift between two different habitats, the cladoceran-rich epilimnion and the deeper strata where Chaoborus can be encountered on the evening forays offshore, before light intensity becomes too low to allow foraging roach to detect their prey (Fig. 2). The behaviour of dailyswitching may also be behind the high variability of the selectivity index within a narrow size category of fish, a majority with values close to either )1 or +1 (Fig. 7). This suggests that an individual roach may prey upon small-bodied Bosmina or D. cucullata on one evening, but on larger D. hyalina on the evening after. This possibility is also reflected by the dominance of different prey categories in similar-sized roach guts from the same trawl sample (Fig. 6). However, a significant part of the high diet variability appears to result from more frequent switching, rather than from the daily (whole-evening) shift Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

11 236 Z. Maciej Gliwicz between different prey categories. In 6% of all 264 roach inspected, the diet diversity expressed as a prey Shannon-Wiener index (based on species contribution to the total food volume) was above.4, which corresponds with the diet of fish number 4 in Fig. 6. The diverse food composition of an individual roach would suggest that most individuals switch from one prey category to another many times within a single feeding session offshore. In its evening thrust towards the middle of the lake, where zooplankton prey is more abundant, a foraging roach may slow down to pick up a number of prey of one category, and then move forward again as soon as a local swarm has been wiped out. It can do so again with another prey category once that other prey has been assessed as more rewarding. Since the pioneering work of Ivlev (1961), Hrbáček (1962), and Brooks & Dodson (1965), the effect of predation by planktivorous fish has been assumed to be selective. Our analyses show that this assumption is correct, but only in the sense that each different body-sized species has a different population-density threshold that results from a different relative reactive field volume. Thus, once the relative proportions of coexisting species have been fixed by body-size dependent mortality, the effect of predation is not selective anymore. On the contrary, the force of fish predation appears to be a strong stabilising factor accounting for constant relative densities of different prey species throughout the seasons and from one habitat to another (Fig. 8). There are opposing forces that stem from the race between individuals of each population to grow and mature soonest. This is the reason for each population to show a reproductive rate as high as possible within the constraints set by temperature and food levels, whereas birth rates in the population and the rate of population increase are controlled by either time or resources (the time and resource limitation of Schoener, 1973). It thus appears that the availability of resources controls the rate of each population increase. Regardless of the rate of increase, the density of each population would eventually be fixed by a mortality rate resulting from fish predation and fish switching from one prey item to another depending on their relative densities (species-specific population-density thresholds section). This conclusion is in agreement with a notion expressed a long time ago (Elton, 1927): that general predators feed most heavily upon the most abundant species until their abundance is reduced, and that the predator switches the great proportion of its attack to another prey which has become the most abundant (Murdoch, 1969). Or as we should say more precisely being aware of the effect of body size relatively the most abundant (Gliwicz, 21). For example, it may be speculated that that the prey-switching behaviour was the reason why the selectivity for Bosmina was found to be slightly higher than for other prey in our experimental lake (Fig. 7). Unlike the other prey species, the Bosmina population may have been just in the phase of density increase beyond its species-specific threshold level at the time of our fieldwork on the lake. This may also be the reason why the food selectivity for a specific prey was neither found to be similar among individual fish, nor significantly different for different prey species, even those representing extremes in body size. High variability of the selectivity index for different prey categories is often a source of frustration for researchers analysing gut contents; they prefer finding high values for conspicuous and low values for less conspicuous prey species (e.g. Bohl, 1982). Clear differences in selectivity values, which are probably less common than published accounts imply, could be interpreted as a sign that a change in the dominant diet is being witnessed, the majority of fish switching from one prey to another which has become the most abundant (Murdoch, 1969), just as could be the case of Bosmina in our experimental lake. Zooplankton in the presence and absence of fish By switching from one zooplankton prey to another, planktivorous fish would hold the density of each species below the carrying capacity (K). Each density increase would be followed by a shift in fish diet from the most rewarding prey in the past to the most rewarding prey in the present situation (Fig. 1, top). The most conspicuous prey (the large-bodied and thus competitively superior species) would be held at the lowest density, corresponding to its low relative density resulting from the high vulnerability of individuals at maturation (large body size at first reproduction). Low abundance would allow for higher food levels, and thus for the coexistence of Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

12 Top-down versus bottom-up effects 237 Population density (% K ) Population density (ind.l 1 ) Elapsed time (days) 6 other species, including small-bodied cladocerans, rotifers and ciliates. This coexistence may last at least until the fish impact has been removed. For example, in Smyslov Pond, one of Hrbácek s famous fishponds in Bohemia, large-bodied D. pulicaria was found to monopolise resources for 9 days in the absence of fish predation (Fig. 1, bottom). The Smyslov Pond example shows that, in the absence of fish, Daphnia density can be controlled from the bottom up and held at the equilibrium level of the carrying capacity of the habitat. Algal food resources would be effectively controlled from the top down, well below their high potential, until fish feed again on Daphnia. With fish predation restored, 9 Elapsed time (days from 1 April 1972) Fig. 1 Top panel: Diagrammatic representation of a typical change in population density of a planktonic herbivore, such as Daphnia or Bosmina in the absence and presence of planktivorous fish. Bottom panel: example of a real density change of a Daphnia population in the absence of fish impact in Smyslov Pond in High numbers of Daphnia pulicaria accompanied by smaller numbers of D. galeata (solid line) lasted in equilibrium for 9 days with low levels of small edible algae (dotted line) and high levels of mineral resources, until extermination of D. pulicaria by fish (day 1) allowed other zooplankton taxa to form a typical multispecies zooplankton community and algae to form a bloom of 7 lg chlorophyll L )1, typical of Smyslov Pond (after Fott et al., 1974; and Fott, Desortova & Hrbáček, 198). Chorophyll (µg L 1 ) Daphnia density would decrease, chlorophyll concentration increase, and ecological space become available to other cladocerans and rotifers that are inferior competitors for resources. This situation appears to be typical of ponds and lakes, where the impact of fish allows for the coexistence of many species with similar ecological niches, including congeneric species such as D. hyalina and D. cucullata (species-specific population-density thresholds section). Concurring with Hutchinson s (1961) paradox of the plankton (e.g. Ghilarov, 1984), diverse plankton assemblages have often been accounted for by non-equilibrium effects based on the intermediate disturbance hypothesis, or justified by different abilities to partition resources (reviewed by Rothhaupt, 199). A diverse community of planktonic herbivores would also be seen following any long-lasting phase of clear-water resulting from a temporary relaxation in fish activity and periodic single-daphnia-species monopolising resources. This situation has been observed in many lakes and is well known as Daphnia summer decline which usually follows a spring clear-water phase (Sommer, 1989; Hülsmann & Voigt, 22). The only natural habitats in which the clear-water phase lasts as long as in Smyslov Pond, are those where fish are absent, and a competitively superior large-bodied phyllopode such as D. pulicaria or Artemia franciscana monopolise resources, holding them at an equilibrium level below the threshold food concentration needed for other species to grow and reproduce (Gliwicz, in press). In such fishless habitats, where water remains clear in spite of high nutrient loads, phytoplankton would be suppressed from the top down by the competitively superior herbivore species, whose high population density in turn is restrained from the bottom up by food availability. The absence of predation allows an individual to allocate all its efforts to the competition for resources, as interspecific competition gives way to intraspecific competition. In the fertile habitat of the Great Salt Lake, Utah, the diverse phytoplankton is held at an extremely low biomass (1 lg chlorophyll L )1 ) by an efficient herbivore, Artemia. When Artemia is removed experimentally or has retreated naturally to diapause, mineral resources are immediately monopolised by the most effective green algae, Dunaliella viridis, leading to a concentration of 3 lg chlorophyll L )1 (Gliwicz, in press). Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

13 238 Z. Maciej Gliwicz Such situations observed in fishless habitats suggest that population densities of both phyto- and zooplankton and body size of zooplankton species at first reproduction may be controlled from the bottom up as long as the top-down impact is effective. High zooplankton densities cannot last, however, after the impact has been removed. As soon as fish are introduced, herbivore population density and body size becomes fixed by top-down forces again, and bottom-up controls become restricted to rates at which density or body size can be restored to levels that would be fixed by fish predation. Conclusions The species-specific population-density thresholds in cladocerans, the similar values for the selectivity index in roach, and the contrast between zooplankton in the presence and absence of fish, all show that an impact from the top-down can control zooplankton biomass, individual body size and population density. In contrast, bottom-up forces influence assimilation rate, individual growth rate and reproduction (Fig. 11). The nature of the impacts from top down and bottom up hence is distinctly different also at the population level. Although in contrast to the individual level a common currency can be conveniently defined at the population level, the disparity of the two entities are equally great at both levels. This conclusion is supported by our data, at least as regards the herbivorous zooplankton. Fish predation would primarily determine the population density in a herbivore such as Daphnia. It would do so regardless of the somatic growth rate of individual animals, and the population reproduction rate, which are both independent of top-down effects. These rates are bottom-up controlled. The different nature of this bottom-up control is best reflected in the notion of the functional response, the processes of food assimilation, individual growth, and population increase, which are all controlled by food level. The rate at which food resources are being produced is not the critical factor, although some people would assume that low food level would not necessarily be equal with food limitation in animals such as Daphnia because even at low food levels food production may be high enough to support high feeding rates and fast individual growth (an anonymous review, pers. comm.). This would be the case as long as the topdown impact of predators was effective. Its removal would allow a single competitively superior species to monopolise resources at an equilibrium level held near the food-threshold level, as in the fish-free habitats of alpine and saline lakes (zooplankton in the presence and absence of fish section). The same reasoning is probably valid for the other trophic levels in the food web, both primary producers and predators. BOTTOM-UP: TOP-DOWN: P A Assimilation (A) Individual growth rate Rate of reproduction CONTROL OF STATE VARIABLES Biomass Individual body size Population density Fig. 11 Diagrammatic representation of the different nature of bottom-up (food limitation) and top-down impacts (predation) on zooplankton, with its abundance (biomass) controlled from the top down by planktivorous fish (right), and process rates (energy carbon flow) controlled from the bottom up by phytoplankton food availability (left). Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,

14 Top-down versus bottom-up effects 239 CONTROL OF RATES Assimilation (A) Individual growth rate Rate of reproduction CONTROL OF STATE VARIABLES Biomass Individual body size Population density Fig. 12 Diagrammatic representation of the different nature of bottom-up and top-down impacts on planktivorous fish, with their abundance (biomass) controlled from the top down by piscivores (right), and process rates (energy carbon flow) controlled from the bottom up by zooplankton prey availability (left). The phytoplankton biomass would depend on the top-down impact of grazing imposed by herbivores rather than by the trophic state of the habitat. This effect would be most apparent in the absence of fish, as low phytoplankton abundance is comparable in fishless habitats regardless of their potential, from ultra-oligotrophic mountain lakes to nutrient-rich saline lakes of hydraulically locked lowlands. The low-biomass multispecies phytoplankton of these habitats would last as long as the top-down impact of an effective herbivore persists. Its removal would allow single algal species to monopolise resources at an equilibrium level with mineral resources kept low, as in the fish-free habitat of the Great Salt Lake (Section 6). These mechanisms could also explain why our efforts at mediating roach feeding behaviour in the experimental lake were unsuccessful (modifying fish feeding behaviour section). We attempted to reduce roach feeding rate, not roach density or biomass, that is a rate, not a state variable. Although treatment with alarm substance could possibly affect roach density in the long term, a short-term increase of zooplankton density cannot be expected if the above scenario is correct. Thus, with the reasoning from Fig. 11 applied to the trophic level above (Fig. 12). I hypothesise that only the state (biomass and population density) of planktivorous fish affects the strength of top-down control, not the rate at which the fish reproduce, grow, or feed. The same effect might account for the fragility of effective top-down control: the spring clear-water phase is usually a short phenomenon and can, if it lasts longer, be abruptly terminated as seen in Smyslov Pond. The effect may also be the reason why topdown effects are gradually weakened from the top to the bottom of the food web as suggested by McQueen et al. (1989). Experiments run in the Plankton Towers at the Max-Planck Institute in Plön, Germany, showed that the top-down effects on roach can be very strong, but also that they are only transitory: fish frightened with alarm substance were more reluctant to feed in the daylight than the reference fish, but the initial difference in food abundance in the evening (Daphnia density) vanished overnight because fish fed in the dark (Gliwicz et al., 21). Ó 22 Blackwell Science Ltd, Freshwater Biology, 47,