1 Introduction

Predation is a key factor shaping transitions in community structure along the freshwater habitat gradient going from small temporal ponds to large permanent lakes (Wellborn et al. 1996). The transition between temporary habitats that contain few predators and permanent fishless habitats that contain large invertebrate predators, and the transition between permanent fishless and fish-containing lakes, have been well studied (Wellborn et al. 1996; Stoks and McPeek 2003a, b). Interestingly, there is also overwhelming evidence for a structuring role of fish predation within permanent fish-containing lakes, associated with water turbidity (Moss 1990; Scheffer et al. 1993; Scheffer 1998). Shallow fish-containing lakes can occur in two alternative stable states: a clear-water state with abundant submerged vegetation and a turbid-water state with high algal biomass (Scheffer et al. 1993; Scheffer 1998). The former state is the only one possibilility at low nutrient loads, whereas the latter predominates at high nutrient loads. Both states can occur at an intermediate nutrient level, and are stabilized by a suite of feedback mechanisms (Scheffer et al. 1993; Scheffer 1998). There is much evidence that the community composition of fish, macroinvertebrates and zooplankton may change dramatically with the turbidity state of the lake (reviewed by Scheffer 1998; Van de Bund and Van Donk 2002; Cottenie and De Meester 2003; Van de Meutter et al. 2005). Traditionally, these differences have been largely attributed to the sheltering property of aquatic vegetation in clear-water lakes (Timms and Moss 1984; Lammens 1989). However, given that many fish species rely on vision to detect their prey (Guthrie and Muntz 1993), it is plausible that water turbidity itself may drastically affect predator–prey interactions, and hence community structure.

The majority of studies that have dealt with predator–prey interactions in relation to water turbidity in lakes and ponds have focused on the top predator’s perspective. For example, Reid et al. (1999) showed that turbidity may reduce prey detection distance and hence predation rates by fish, although compensatory mechanisms such as altered foraging behaviour may exist, which partially counter these effects (Vøllestad et al. 1986). Water turbidity is, however, likely to not only influence the predator but also the prey. Prey that detects its food and predators visually may not only suffer from a lower foraging efficiency in turbid water, but also may encounter difficulties in judging the risk of predation in a situation in which both prey and predator detection are impaired. Abrahams and Kattenfeld (1997) hypothesized that in turbid systems, antipredator behaviour of the prey will become ineffective and will occur less frequently. Several studies found support for this hypothesis in fish–fish predator–prey systems (Abrahams and Kattenfeld 1997; Jacobsen and Perrow 1998; Jacobsen et al. 2004). From this hypothesis, Abrahams and Kattenfeld (1997) derived a second hypothesis that predator–prey interactions in turbid water will be primarily characterized by the direct effects of predator consumption of prey, while in clear-water systems, direct as well as indirect effects of the predator on the prey’s growth rate through behavioural modification will be important. Therefore, the turbidity state of a lake might not only influence the prey’s behaviour but also its key life history characteristics like growth and survival. To our knowledge, the second hypothesis of Abrahams and Kattenfeld (1997) has never been tested.

In this study, we looked at the effects of water turbidity on behaviour, growth and survival of Ischnura elegans (Van der Linden 1820), a damselfly that occurs over a wide range of permanent aquatic habitats in Belgium. I. elegans larvae are opportunistic intermediate predators in the littoral zone of ponds and lakes, and often suffer from predation by fish (Corbet 1999). In a first experiment, we evaluated the first prediction of Abrahams and Kattenfeld (1997) in an invertebrate prey species by testing whether Ischnura larvae relax their antipredator behaviour under risk of predation in turbid water compared to clear water. To get insight as to how former exposure to turbid water may shape this behavioural response, we used both field-caught larvae that originated from a clear pond and from a turbid pond, and reared half of the larvae from each pond in the laboratory in clear or turbid water prior to the experiment. In a complementary life history experiment, we investigated how these behavioural interactions translate into survival and growth rate of larvae in the field under clear-water and turbid conditions, thus explicitly evaluating the second hypothesis of Abrahams and Kattenfeld (1997). Larvae of I. elegans from the clear-water and the turbid pond were reared in the presence or absence of fish in enclosures set up in both a clear-water and a turbid pond. Since, in addition to turbidity, the presence of aquatic vegetation may impair the detection of prey and predators in natural systems (Crowder and Cooper 1982; Burks et al. 2001), we crossed the fish treatment with high and low density of vegetation in each pond.

2 Materials and methods

2.1 Behaviour experiment

2.1.1 Experimental design

The effect of water turbidity on behaviour of larvae of the damselfly I. elegans was studied in a common garden experiment in the laboratory. The experiment had a 2×2×2 design with all combinations of three different factors related to turbidity: the pond of origin (clear, turbid), rearing turbidity in the laboratory (clear, turbid) and actual turbidity during the behavioural trials. I. elegans larvae of instar F-1 and F-2 (with F-0 as final instar) were collected on 27 May 2003 in the nature reserve De Maten (Northeast Belgium). This reserve consists of 34 shallow ponds that exhibit a wide range in turbidity levels. Larvae were collected in two neighbouring ponds: turbid water pond 12, and clear-water pond 13 (Cottenie et al. 2001; Van de Meutter et al. 2005). They were brought to a climate room (20°C, L:D 16:8) and were placed in aerated 10-L buckets (40 larvae/pond; 10 larvae/bucket). Half of the larvae from each pond (20 individuals) were reared in water of their native pond; the other half were transplanted to buckets containing pond water with the other turbidity state. In each bucket, we placed ten stems (length 15 cm) of plastic aquatic vegetation (mimics of Ceratophyllum, Plantastics) to provide some structure to the habitat. We daily visually checked the water turbidity in the turbid water buckets, and added small amounts of bentonite clay on two occasions (day 8 and day 18) to maintain turbidity at the original level. Larvae were kept in the laboratory for 21 days and were daily fed Daphnia magna ad libitum.

From 16 to 20 June, we scored the behaviour of all larvae in random order once in clear water and once in turbid water, but never on the same day. Observations were done in 10-L aquaria (24×34×13 cm) in which two compartments were created: a small compartment (10×10.5 cm) attached to the wall of the aquarium, separated from a large compartment by transparent Plexiglas walls and a small piece of netting. Thus, both visual and olfactorial cues could be transmitted from one compartment to the other. Ten minutes before the start of an observation trial, a pumpkinseed Lepomis gibbosus was added to the small compartment to assure the presence of fish cue. As we were interested in the behaviour of the larvae in the presence of fish, fish and fish cue were present in all observation trials. At the start of the experiment, the fish was transferred to the large compartment, and 25 adult Daphnia sp. were added to the small compartment. Next, an I. elegans larva was gently released into the small compartment and was allowed to settle down for 1 min. After this acclimatization period, its behaviour was recorded for 10 min. We recorded six distinct behaviours (cf. Johansson et al. 2001): walking (frequency and duration), swimming (frequency and duration), orienting the head towards a Daphnia (frequency), advancing towards a Daphnia (frequency and duration), striking towards a Daphnia with the labial mask to catch it (frequency) and catching a Daphnia (frequency). Following the advice of Davies-Colley and Smith (2001), and given the fact that we focused on the visual aspects of turbidity, we did not measure turbidity in our experiments in NTU (nephelometric units) but with a Snell tube. A Snell tube consists of a small secchi disk that can be lowered in a long PVC tube filled with water. Snell depth is given as the mean of the measurements of the maximal depths (in cm) at which a black cross on the disk can be seen when lowering the disk (Ø 5 cm) in the tube and the depth at which the cross can be seen again when lifting the disk. Turbid water in our experiment had a Snell depth of 15 cm, corresponding to that of the turbid pond where the larvae originated from, and was created at the start of each day by adding bentonite clay to dechlorinated tap water. To maintain turbidity levels throughout the experiments, we stirred up the water between all behavioural trials.

2.1.2 Statistics

Prior to analysis, the original behavioural variables were log(Y+1)-transformed to eliminate positive relationships between means and variances of all variables. As the six behavioural variables are potentially highly correlated, we performed a principal component analysis (PCA), which extracts new orthogonal variables as a linear combination of the original variables. We obtained three new variables (PC1–PC3) with eigenvalues >1, which together explained 87.5% of the original variance in the behavioural dataset. PC1 (55.9% explained variance) had high positive loadings (>0.75) from duration and frequency of walking and is further referred to as +WALK. PC2 (22.4% of the explained variance) had strong negative loadings (<−0.75) from all variables related to foraging (advancing towards Daphnia, striking towards Daphnia, catching Daphnia) and is hereafter named −FORAGE. PC3 (9.2% explained variance) was highly positively correlated (>0.75) with duration and frequency of swimming (hereafter +SWIM). To evaluate the effect of origin of the larvae, rearing turbidity and actual turbidity during the experiment on the newly extracted behavioural variables, we performed a three-way repeated measures MANOVA. Origin (clear pond/turbid pond) and rearing turbidity (clear/turbid) were used as categorical predictors. For each of the three behavioural PCs, we treated the score of a given larvae in clear and in turbid water as repeated measures, giving in total three sets of repeated measures as dependent values. When significant, the rmMANOVA is followed by rmANOVAs for each PC variable.

2.2 Enclosure experiment

2.2.1 Experimental setup

To study the effects of fish and vegetation density on larval survival and growth rate of the damselfly I. elegans in ponds with a contrasting turbidity state, we set up an enclosure experiment in the same turbid (pond 12) and clear-water pond (pond 13) as the ones we sampled for I. elegans larvae in the behavioural experiment. To take into account possible acclimatization of the larvae to their native pond, we did a reciprocal transplant by rearing half of the larvae in their own pond and the other half in the pond with contrasting turbidity state. This resulted for each pond in eight different treatment combinations of origin (pond 12, pond 13), fish presence (absent, present) and vegetation density (sparse, abundant). Each treatment combination was replicated in three enclosures, giving a total of 24 enclosures in each pond. Within each individual enclosure, all larvae originated from the same pond. Enclosures were placed in three different blocks (A, B, C) of eight. Within each block, the eight different treatment combinations were randomly assigned to the enclosures.

Enclosures had a cylindrical frame (diameter 50 cm, height 90 cm) of chicken wire (mesh size 2 cm). Within this frame, we placed a cylindrical nylon bag (mesh size 1 mm) with the same size and shape. To prevent the bags from collapsing, we fixed them tightly to the frame at 30 cm and at 60 cm. A plastic dish (diameter 50 cm) served as the bottom. Each enclosure was attached to three 1.5-m PVC pipes that were drilled into the sediment. One block of eight enclosures was placed in both ponds on 31 July (block A), one on 1 August (block B) and one on 2 August (block C). One pumpkinseed Lepomis gibbosus (standard length 5.8–6.3 cm) was added to the fish treatments on 24 August. The diet of pumpkinseeds of this size class in “De Maten” consists largely of macroinvertebrates, containing mainly chironomids, ephemeropterans and zygopterans (Declerck et al. 2002). All pumpkinseeds originated from pond 27 of De Maten, a pond with a turbidity state intermediate to that of pond 12 and pond 13, and were inoculated to the enclosures directly after capture. The time lag of 3 weeks between installing the enclosures and applying the fish treatment served to permit colonization of prey organisms for the damselfly larvae and the fish.

The vegetation treatments were created by adding sago pondweed Potamogeton pectinatus. Sago pondweed is a common submerged macrophyte species in the shallow ponds of De Maten, and it occurs in both clear-water and turbid ponds. Sago pondweed was collected in pond 27, rinsed carefully on white plastic dishes to remove all large macroinvertebrates, and divided into small (30-g wet weight) and large (300-g wet weight) portions. On 24 August, we added a large portion to all enclosures of the high-density macrophyte treatment and a small portion to the low-density macrophyte treatment.

On 26 August, we added 20 damselfly larvae to each enclosure. The resulting density corresponds to natural damselfly densities in both ponds. Two species of damselfly, I. elegans and E. najas, make up more than 90% of the damselfly community in both ponds, and occur at a ratio of 3/1 in pond 13 and of 2/1 in pond 12. Therefore, enclosures with larvae originating from pond 13 consisted of 15 I. elegans and 5 E. najas, whereas larvae originating from pond 12 consisted of 13 I. elegans and 7 E. najas. Larvae were collected by sampling the littoral of both ponds and were placed in the enclosures on the same day. Only instars F-4 to F-1 larvae were used in the experiment. The experiment was run for 21 days. Afterwards, the entire content of each bag was preserved in 4% formaldehyde. Unfortunately, three enclosures in the turbid pond (treatment combinations: high vegetation–no fish-origin clear pond [2] and low vegetation–fish-origin turbid pond [1]) were damaged during the experiment. Data of these enclosures were excluded from all following analyses.

2.2.2 Growth rate and survival

For every enclosure, we quantified the growth rate of I. elegans as [ln(mean final dry mass)−ln(mean initial dry mass)]/21 day. As this metric is not an actual measure of individual growth rate of larvae, we will hereafter refer to it as the change in average body mass (CABM). Mean initial dry mass of the I. elegans that were inoculated in the enclosures was determined on an additional sample of 50 individuals from each pond. Mean final dry mass for an enclosure is the mean dry mass of I. elegans larvae that were retrieved from that enclosure after 21 days. Dry mass was determined with a Mettler-Toledo HL52 microbalans to the nearest 10 μg after drying larvae at 50°C for 48 h. Prior to drying, head width was measured with an Olympus SZX-ILLK200 stereomicroscope to the nearest 10 μm. Head width is an accurate measure of total body length of damselflies (Corbet 1999) and was measured to detect possible size selective foraging of the fish. Survival was calculated as the number of I. elegans larvae retrieved at the end of the experiment divided by the initial number of I. elegans larvae added to the enclosure (which was different depending on pond origin).

2.2.3 Statistics

To test for possible effects of experimental pond, fish, vegetation and pond of origin on the survival of I. elegans larvae, we performed a four-way ANOVA using the mixed procedure in SAS 8.02 (SAS institute 2001). Experimental pond, fish, vegetation and pond of origin were used as categorical predictors; survival was the dependent variable. In three treatment combinations with fish, changes in average body mass could only be estimated from a single replicate (the remaining cages having zero survival). This may yield unreliable estimates of larval growth. To circumvent this problem, we sacrificed the origin treatment by lumping the changes in average body mass data for larvae of a different pond origin. Effects of experimental pond, fish and vegetation on changes in average body mass of Ischnura larvae were analysed using a three-way ANOVA with fish, vegetation and experimental pond as categorical predictors, and CABM as the dependent variable. Since our experiment has an unreplicated randomized block design, ‘Block’ (three levels: A, B and C) was included in both models as a random main effect (Underwood 1997). In both analyses, however, no significant effect of Block was present (both P>0.12), and we pooled the MS of the Block factor with the error MS. Specific effects of interest were further explored using contrast analysis in comparison with least square means.

Apparent differences in changes in average body mass between larvae from enclosures with fish and without fish can originate either from differences in effective growth rate in the presence of fish or from size-selective foraging of the fish. To differentiate between both, we performed an ANCOVA on the final mass of the larvae, with head width as the covariate and fish treatment as the categorical factor. Because, larvae from the same enclosure are statistically dependent, we included the categorical factor “Enclosure” nested in the fish treatment and its interaction with head width (enclosure [fish]×head width) to correct for this. To obtain a linear relationship between the variables, final mass was log10-transformed. Head width is known to be constant per larval stage and independent of food regime (McPeek 1990). If final mass of I. elegans larvae still differs among fish treatments after correction for head width, then this mass difference is not caused by size-selective foraging of the fish and thus truly reflects differences in effective growth rate.

3 Results

3.1 Behaviour experiment

In general, I. elegans larval behaviour was clearly affected by the turbidity during the experiment (rmMANOVA: F 3,58=7.21, P<0.001), but not by rearing turbidity (F 3,58=2.09, P=0.11), nor by the origin of the larvae (F 3,58=0.15, P=0.94). Univariate rmANOVAs showed that larvae that originated from the turbid pond walked more when the water was turbid, but not those that originated from the clear-water pond (+WALK: turbidity during experiment × origin: F 1,126=4.29, P=0.040; Fig. 1). Walking depended also on the rearing turbidity of the larvae: larvae that were reared in turbid water walked more than larvae that were reared in clear water (F 1,126=10.38, P=0.002). Larvae also swam more in turbid water than in clear water, regardless of their origin or their rearing turbidity (+SWIM: F 1,126=5.57, P=0.02; other P>0.25; Fig. 1). Finally, foraging behaviour of the larvae was not modified by the turbidity during the experiment nor by the rearing turbidity state that they experienced (−FORAGE: all P>0.29, Fig. 1).

Fig. 1
figure 1

The behavioural principal components +WALK, −FORAGE and +SWIM of Ischnura larvae in the presence of fish in function of turbidity during the experiment, rearing turbidity and pond of origin. Left panels give values for larvae from the clear-water pond, right panels for larvae of the turbid pond. Means are given ±1SE

Full size image

3.2 Enclosure experiment

3.2.1 Survival

The results of the ANOVA on survival data of Ischnura larvae are summarized in Table 1. Survival of Ischnura larvae was determined by an interactive effect of fish and vegetation, which by itself depended on experimental pond (experimental pond × fish × vegetation, P=0.0005). In the clear-water pond, fish had an overall negative effect on survival of Ischnura, whereas in the turbid pond, fish had a negative effect on Ischnura survival at low vegetation density, but not at high vegetation density (Fig. 2). The lack of an effect of fish on Ischnura survival at high vegetation density in the turbid pond resulted from a combination of reduced survival in the absence of fish (contrast, turbid pond versus clear pond for the no-fish–high vegetation density treatment; P=0.0001) and a tendency for increased survival in the presence of fish (contrast, clear pond versus turbid pond for the fish–high vegetation density treatment: P=0.081). Ponds also differed in the way larvae of a different origin were affected by vegetation density (experimental pond × origin × vegetation, P=0.0014). Larvae originating from the clear-water pond were better at surviving at high vegetation density than at low vegetation density, whereas larvae originating from the turbid pond showed little difference. In the turbid pond, larvae from both origins were similarly affected by vegetation density.

Table 1 Results of the ANOVAs on survival and change in average body mass (CABM) of Ischnura larvae in the enclosure experiment
Full size table
Fig. 2
figure 2

Survival of Ischnura larvae of the enclosure experiment in the clear-water pond (upper panels) and in the turbid water pond (lower panels) in function of fish, vegetation density and pond of origin. Left panels show the results for the larvae that originated from the clear-water pond; right panels show the results for larvae originating from the turbid pond. Symbols represent the means of three replicate enclosures except for the ‘high vegetation-no fish-origin clear pond’ and the ‘low vegetation-fish-origin turbid pond’ treatment combinations in the turbid pond, the means of which were based on 1 and 2 enclosures, respectively. Means are given ±1SE

Full size image

3.2.2 Change in average body mass

The results of the ANOVA in the data for CABM of Ischnura larvae are summarized in Table 1. In both ponds, fish had a negative effect on CABM of Ischnura larvae at low vegetation density, but not at high vegetation density (fish × vegetation, P=0.037, Fig. 3). However, when corrected for head widths, only a weak negative trend of fish at low vegetation density on final mass of Ischnura was present (ANCOVA; fish: F 1,110=2.85, P=0.094; fish × head width: F 1,110= 2.04, P=0.16). This suggests that size-selective foraging of the fish may have contributed to the apparent differences in growth rate of the Ischnura. CABMs in the fish treatments tended to be slightly higher in the turbid than in the clear-water pond (experimental pond × fish, P=0.10). This was supported by a negative fish effect on final size-corrected body mass of larvae in the clear-water pond (ANCOVA; fish: F 1,174=4.21, P=0.042; fish × head width: F 1,174=6.14, P=0.014; Fig. 4a), but not in the turbid pond (ANCOVA; fish: F 1,110=3.05, P=0.083; fish×head width: F 1,110=2.04, P=0.16; Fig. 4b). This indicates lower effective growth of larger larvae in the presence of fish in the clear-water pond, but not in the turbid pond.

Fig. 3
figure 3

Change in average body mass of Ischnura larvae of the enclosure experiment in the clear-water pond (left) and in the turbid pond (right) in function of fish and vegetation density. For each mean, the number of enclosures followed by the total number of individuals on which the means are based are given. Means are given ±1SE

Full size image
Fig. 4
figure 4

Regressions of final head widths versus the logarithm of final mass for Ischnura larvae in the enclosure experiment. Regressions are given for the no-fish and the fish treatment, separately for (a) larvae of the clear-water pond and (b) for larvae of the turbid pond. Dashed lines represent the regressions for the no-fish treatment, solid lines represent the regression for the fish treatment

Full size image

4 Discussion

We showed that the invertebrate predator Ischnura changes its behaviour according to the turbidity of the water. Moreover, we found indications that growth and survival of Ischnura larvae were differently affected by the presence of fish and vegetation in a clear-water pond compared to a turbid water pond. Taken together, our results suggest that predator–prey interactions differ between clear-water ponds and turbid water ponds. We discuss these findings in more detail and confront them with the predictions of Abrahams and Kattenfeld (1997).

According to the first prediction by Abrahams and Kattenfeld (1997), antipredation behaviour of prey should be less pronounced in turbid water than in clear water. One of the most common antipredator behaviours in prey is the reduction of activity in the presence of predators (Lima 1998). In damselflies, foraging, walking and swimming behaviour have been shown to attract the attention of predators, leading to increased predation and mortality (Convey 1988; McPeek 1990; Baker and Smith 1997; Elkin and Baker 2000). By consequence, these behaviours are generally decreased in the presence of a predator (e.g. McPeek 1990; Stoks et al. 2003). Translated to our model system, the prediction of Abrahams and Kattenfeld (1997), that the presence of a fish in Ischnura should show more foraging, walking and swimming activity in turbid water than in clear water. Overall, our results are in accordance with this hypothesis for walking and swimming. Similarly, Heads (1985) observed that Ischnura larvae moved more in the dark than in the light, and suggested that this was because larvae felt safer under the cover of dark. In contrast with the first hypothesis, foraging activity did not differ between clear and turbid water. Although decreased foraging in the presence of fish has been observed in several species of damselflies (e.g. McPeek 1990; Stoks and Johansson 2000; Stoks and McPeek 2003b) this response is not always present in I. elegans (Slos and Stoks in review). This is consistent with the fact that Ischnura, a fast growing species, tries to avoid large reductions in growth rate in the presence of a predator as opposed to other damselfly genera like Enallagma (McPeek et al. 2001; McPeek 2004).

The behaviour of Ischnura in clear versus turbid water was to a large extent determined by its former experiences with turbidity. Larvae that were reared for 3 weeks in turbid water showed a higher walking activity than larvae that were reared in clear water. This probably indicates carry-over effects (Sih et al. 2003), where organisms that have adapted their behaviour to a particular risk environment may to some extent carry over this behaviour when confronted with a new environment, as was shown earlier in Ischnura (Slos and Stoks in review). Strikingly, pond origin still affected some responses to water turbidity, even though all larvae had been kept for 3 weeks in the laboratory. Larvae from the turbid pond walked more in turbid water than in clear water, but larvae from the clear pond showed no difference. We acknowledge that both ponds from where the larvae originated may differ in other aspects than turbidity and hence it cannot be claimed that the differences are related to the turbidity of the ponds. If turbidity is the factor involved, it is still unclear whether these origin-specific responses to turbidity stem from carry-over effects following acclimatization to the turbidity of the original pond, or have a genetic basis. Genetic differences, however, seem unlikely since both ponds are separated only by 10 m. To verify this, more detailed quantitative genetic research is needed.

Overall, our results of the enclosure experiment are compatible with the second hypothesis of Abrahams and Kattenfeld (1997): direct effects of fish on Ischnura (survival) were apparent in both the turbid and the clear-water pond, indirect effects of fish on Ischnura (CABM, effective growth rate) tended to be stronger in the clear-water pond than in the turbid pond. Survival of Ischnura at low vegetation density was strongly and negatively affected by the presence of fish in both ponds. This confirms our premise that pumpkinseed is an effective predator of Ischnura and can have severe negative effects on its survival, as shown in other studies (McPeek 1998). In the turbid but not in the clear-water pond, however, the fish effect tended to be relaxed when the vegetation density was high. Previous studies have shown a mitigating effect of vegetation on fish predation on macroinvertebrates and zooplankton (Crowder and Cooper 1982; Gilinsky 1984; Burks et al. 2001). Dense vegetation beds prevent visual detection of prey by fish and may also physically hinder the fish to catch their prey (Crowder and Cooper 1982; Burks et al. 2001). In our study, the mitigating effect of vegetation on fish predation tended to be strengthened when the protective effect of vegetation was combined with a reduced visibility due to water turbidity. In the turbid pond, high vegetation density had a negative effect on survival of Ischnura in the absence of a predator, possibly suggesting increased intraspecific interactions (Elkin and Baker 2000). Unfortunately, we do not have data on Ischnura behaviour under different combinations of macrophyte density and turbidity that could support this. Finally, larvae originating from the clear-water pond were better at surviving at high versus low vegetation density in the clear-water pond, whereas survival of Ischnura originating from the turbid pond benefited only weakly from high vegetation density. Possibly, larvae from the clear-water pond were better adapted at surviving in a clear-water and high vegetation density environment, since it is their natural environment.

Change in average body mass in the fish treatments tended to be higher in the turbid than in the clear-water pond, suggesting indirect effects of fish were mitigated in the turbid pond (Fig. 3). Accordingly, fish negatively affected size-corrected body mass in the clear pond, indicating lower effective growth rate in the presence of fish. However, fish did not affect body mass in the turbid pond (Fig. 4). We suggest that the fish-induced lower effective growth rate in clear but not in turbid water could result from the combination of two factors. First, damselfly larvae and fish partially compete for the same food sources like chironomids and cladocerans (Thompson 1978; Declerck et al. 2002), and food availability to damselfly larvae probably was lower in the presence of fish. In the turbid pond, prey densities may have remained more elevated than in the clear-water pond because of constrained visual predation by the fish. Second, in nature prey is often aggregated in patches (Diehl and Kornijów 1998), and foraging success may be linked to walking behaviour to find patches with prey (Baker 1982). In the presence of fish, Ischnura walked less in clear water compared to turbid water, which may result in reduced feeding rates and growth rates. The same two factors may also explain why CABMs were reduced in the presence of fish at low vegetation density, but not at high density. The sheltering effect of high vegetation density may have increased prey availability and possibly Ischnura mobility towards optimal foraging patches. The fact that the effect of fish on final corrected mass of Ischnura larvae in the low vegetation treatment was only weak, suggests that other mechanisms such as size-selective predation by fish could also have contributed to the observed pattern, however, the number of surviving individuals (9) in the low vegetation treatment may have been too low to detect the subtle differences for corrected final body mass.

In conclusion, our results confirm the importance of aquatic vegetation in shaping predator–prey interactions in clear-water versus turbid lakes, while also showing that turbidity itself may drastically affect predator–prey interactions by changing prey behaviour and possibly life history. More specifically, our results demonstrate that an invertebrate, Ischnura, shows a more risky higher activity in turbid water compared to clear water, and are compatible with the idea that fish–Ischnura interactions are primarily mediated by direct effects in turbid water, and increasingly by both direct and indirect effects in clear water. Further research should focus on the possible consequences of the observed patterns of altered life history for population dynamics and community structure in prey populations of clear-water and turbid lakes.