Context-dependent planktivory: interacting effects of turbidity and predation risk on adaptive foraging

Authors

  • Kevin L. Pangle,

    Corresponding author
    1. Department of Biology, Institute of Great Lakes Research, Central Michigan University, Brooks Hall 194, Mount Pleasant, Michigan 48859 USA
    Search for more papers by this author
  • Timothy D. Malinich,

    1. Department of Biology, Institute of Great Lakes Research, Central Michigan University, Brooks Hall 194, Mount Pleasant, Michigan 48859 USA
    2. Aquatic Ecology Laboratory, Department of Evolution, Ecology, and Organismal Biology, The Ohio State University, 1314 Kinnear Road, Columbus, Ohio 43212 USA
    Search for more papers by this author
  • David B. Bunnell,

    1. United States Geological Survey, Great Lakes Science Center, 1451 Green Road, Ann Arbor, Michigan 48105-2807 USA
    Search for more papers by this author
  • Dennis R. DeVries,

    1. Department of Fisheries and Allied Aquacultures, Auburn University, 311 Swingle Hall, Auburn, Alabama 36849 USA
    Search for more papers by this author
  • Stuart A. Ludsin

    1. Aquatic Ecology Laboratory, Department of Evolution, Ecology, and Organismal Biology, The Ohio State University, 1314 Kinnear Road, Columbus, Ohio 43212 USA
    Search for more papers by this author

  • Corresponding Editor: J. Drake.

Abstract

By shaping species interactions, adaptive phenotypic plasticity can profoundly influence ecosystems. Predicting such outcomes has proven difficult, however, owing in part to the dependence of plasticity on the environmental context. Of particular relevance are environmental factors that affect sensory performance in organisms in ways that alter the tradeoffs associated with adaptive phenotypic responses. We explored the influence of turbidity, which simultaneously and differentially affects the sensory performance of consumers at multiple trophic levels, on the indirect effect of a top predator (piscivorous fish) on a basal prey resource (zooplankton) that is mediated through changes in the plastic foraging behavior of an intermediate consumer (zooplanktivorous fish). We first generated theoretical predictions of the adaptive foraging response of a zooplanktivore across wide gradients of turbidity and predation risk by a piscivore. Our model predicted that predation risk can change the negative relationship between intermediate consumer foraging and turbidity into a humped-shaped (unimodal) one in which foraging is low in both clear and highly turbid conditions due to foraging-related risk and visual constraints, respectively. Consequently, the positive trait-mediated indirect effect (TMIE) of the top predator on the basal resource is predicted to peak at low turbidity and decline thereafter until it reaches an asymptote of zero at intermediate turbidity levels (when foraging equals that which is predicted when the top predator is absent). We used field observations and a laboratory experiment to test our model predictions. In support, we found humped-shaped relationships between planktivory and turbidity for several zooplanktivorous fishes from diverse freshwater ecosystems with predation risk. Further, our experiment demonstrated that predation risk reduced zooplanktivory by yellow perch (Perca flavescens) at a low turbidity, but had no effect on consumption at an intermediate turbidity. Together, our theoretical and empirical findings show how the environmental context can govern the strength of TMIEs by influencing consumer sensory performance and how these effects can become realized in nature over wide environmental gradients. Additionally, our hump-shaped foraging curve represents an important departure from the conventional view of turbidity's effect on planktivorous fishes, thus potentially requiring a reconceptualization of turbidity's impact on aquatic food-web interactions.

Introduction

Adaptive phenotypic plasticity can shape the nature and magnitude of species interactions within ecosystems, and in turn, have a profound influence on population demographics and dynamics, community composition, ecosystem structure and function, and evolutionary outcomes (see reviews by Werner and Peacor 2003, Ohgushi 2005, Schmitz et al. 2008). A major driver of adaptive phenotypic plasticity is predation risk, with many prey species responding to changes in perceived predation risk by modifying their behavior, morphology, physiology, or life history in ways that alter their interactions with the predator (reviewed in Lima 1998, Tollrian and Harvell 1999, Ohgushi 2005). While such adaptive phenotypic responses are expected to improve fitness (Tollrian and Harvell 1999), they also typically invoke costs to the prey that contribute to predator-prey interactions (i.e., the predator's nonconsumptive effect) in fundamentally different ways than direct consumption (Abrams 1995, Luttbeg and Schmitz 2000). Such predator-driven phenotypic responses in the reacting prey also can indirectly affect the broader community through altered food-web interactions (i.e., trait-mediated indirect effects [TMIEs]; reviewed in Bolker et al. 2003, Werner and Peacor 2003, Abrams 2007).

Accurately predicting the outcomes of predator-driven adaptive phenotypic plasticity can be difficult, however, owing in part to the strong dependence of plasticity on the environmental context (Agrawal et al. 2007). Although still rare relative to the hundreds of studies that have demonstrated the strong influence of adaptive phenotypic plasticity in ecosystems (reviewed by Werner and Peacor 2003, Miner et al. 2005, Ohgushi 2005), a handful of studies have demonstrated how environmental factors such as resource availability (Luttbeg et al. 2003, Hawlena et al. 2011), ecosystem productivity (Turner 2004, Werner and Peacor 2006), and landscape features (e.g., water depth, tidal flows, habitat complexity, canopy cover; Rothley and Dutton 2006, Grabowski et al. 2008, Trussell et al. 2008, Heithaus et al. 2009) can strongly influence nonconsumptive effects and TMIEs by altering the trade-offs associated with adaptive phenotypic responses.

A group of environmental factors of particular relevance are those that can influence sensory performance of consumers at intermediate trophic levels. For example, Carr and Lima (2010), Turner and Chislock (2010), and Large et al. (2011) provide examples of how fast winds, elevated pH, and high water flow velocity, respectively, can disrupt the ability of intermediate consumers (birds, freshwater snails, and marine hard clams, respectively) to perceive predator cues in the air or water, thus leading to a diminished behavioral response of these intermediate consumers to their predator(s). Specific to TMIEs, Kimbro (2012) demonstrated how the tidal regime in the northeastern Gulf of Mexico influenced the ability of snails to detect predator cues in the water, which mediated the use of cordgrass (Spartina alternifolia) by snails as a refuge from top predators (blue crab Callinectes sapidus and conch Melongena corona). Because cordgrass also happens to serve as a primary basal resource for snails (an intermediate consumer), its enhanced use as a refuge by snails during diurnal tidal inundation led to increased snail grazing on cordgrass relative to when tides were more mixed (Kimbro 2012).

Building on this body of research regarding context-dependent TMIE, we examined the influence of turbidity (cloudiness in water caused by suspended particles such as sediments and phytoplankton) on a TMIE. Historically, turbidity has been viewed as one of the most important regulators of species interactions, and in turn community structure, in aquatic ecosystems (Cuker 1993, Aksnes et al. 2004, Horppila and Liljendahl-Nurminen 2005). Further, owing to continued watershed development, which has increased the delivery of sediments and productivity-limiting nutrients to rivers and downstream (recipient) ecosystems, the importance of turbidity as a community structuring mechanism has increased and likely will continue to do so with sustained climate-driven increases precipitation and runoff (Nelson et al. 2009, Rabalais et al. 2009).

Herein, we sought to more fully define how turbidity controls the indirect effect of a top predator (piscivorous fish) on a basal prey resource (zooplankton) that is mediated through changes in the foraging behavior of an intermediate consumer (zooplanktivorous fish). Turbidity can affect the sensory performance of visual consumers by reducing the distance at which they can detect prey (reviewed in Utne-Palm 2002). For visual consumers at intermediate trophic levels, turbidity can thus theoretically hinder their efficiency in locating their basal prey resource, while simultaneously lessening their own risk to predation from visual consumers at higher trophic levels. In this way, turbidity has the potential to weaken (via visual constraints) or actually strengthen (via changes in foraging behavior resultant of reduced visual risk perception) the interaction between visual consumers at intermediate trophic levels and their basal prey resource. Previous studies with zooplanktivorous fishes support this mixed effect of turbidity on planktivory, with some experiments revealing negative effects of turbid water on consumption rates (e.g., Gardner 1981, Wellington et al. 2010) and others revealing positive effects (e.g., Boehlert and Morgan 1985, Gregory and Northcote 1993, Miner and Stein 1993).

What mechanisms are responsible for these discrepancies and how these mechanisms are realized in nature are open questions. We suspect, however, that turbidity and predation risk interact to modify the behavior of intermediate consumers. This expectation is driven by our synthesis of many unrelated studies, which in general show that (1) turbidity reduces foraging efficiency of visual consumers in the absence of predation risk (Gardner 1981, De Robertis et al. 2003, Wellington et al. 2010) and (2) visual consumers at intermediate trophic levels that are exposed to predation risk exhibit fewer anti-predator behaviors (Gregory 1993, Miner and Stein 1996, Abrahams and Kattenfeld 1997), as well as more aggressive foraging behavior (Lehtiniemi et al. 2005), in turbid versus clearer water. Based on these observations, we hypothesized that, in the presence of a predator, turbidity would lessen the risk of predation perceived by an intermediate consumer, and in turn, lead to increased consumption of the basal prey resource. Importantly, our hypothesis is based in part on the fact that the rate at which visual detection decays with increased turbidity has been shown to be greater for top predators (e.g., piscivorous fish) than for intermediate consumers (e.g., planktivorous fish) (De Robertis et al. 2003). From the perspective of TMIEs, our hypothesis implies that the positive indirect effect of the top predator on our basal resource, as mediated by the adaptive foraging behavior of our intermediate consumer, would be weakened with enhanced turbidity. Because foraging by our intermediate consumer would eventually decline as turbidity reached extremely high levels that compromise life functions (Newcombe and McDonald 1991), we also predicted a unimodal (hump-shaped) foraging response curve for intermediate consumers over a full gradient of turbidity and in the face of predation risk.

To test our hypotheses, we first formulated them in a mathematical framework, generating theoretical predictions of the adaptive foraging response of a visually-feeding intermediate consumer (i.e., a zooplanktivorous fish) across gradients of turbidity and predation risk. We then evaluated these theoretical predictions in nature by quantifying patterns in foraging of several age-0 zooplanktivorous fishes that were collected from different ecosystems: yellow perch (Perca flavescens) in Lake Erie and its embayments; black and white crappie (Pomoxis nigromaculatus and P. annularis, respectively) in 12 Ohio reservoirs; and bluegill sunfish (Lepomis macrochirus) in several Alabama pond ecosystems. Finally, to more fully evaluate the model predictions, given that all of our field observations occurred with predation risk being present, we conducted a laboratory experiment that quantified the foraging rate of age-0 yellow perch under two different levels of turbidity in both the presence and absence of predation risk. In combination, our findings (1) have important implications for how foraging interactions in response to turbidity are viewed by aquatic ecologists (and expressed in models) and (2) provide novel insights into the robust importance of context-dependent TMIEs in nature.

Model

Formulation and analysis

We used an optimality modeling framework developed by Werner and Anholt (1993) to predict the adaptive foraging response of our intermediate consumer and the corresponding TMIE of the top predator on the basal resource, under varying levels of predation risk and turbidity. The key adaptive behavior in our model is swimming speed, which is a behavior that simultaneously dictates foraging success (i.e., faster consumer swimming leads to higher encounter and consumption rates of its basal resource; Gerritsen and Strickler 1977) and predation risk (i.e., faster consumer swimming leads to higher rates of predator encounter and consumer mortality; Gerritsen and Strickler 1977, Werner and Anholt 1993). Thus, in our model, our intermediate consumer would modify its swimming speed (our proxy for foraging rate on the basal resource; Werner and Anholt 1993) in a way that minimized its ratio of mortality (μ) over growth (g). This μ/g approach has been shown to accurately predict planktivorous fish foraging (Gilliam and Fraser 1987, Jensen et al. 2006), particularly for age-0 individuals (Giske and Aksnes 1992) whose survival is strongly linked to g (Miller et al. 1988). While the μ/g approach ignores the potential effect of time horizons (e.g., onset of winter, shift from pelagic to benthic life) in predictions of optimal behavior (Mangel and Clark 1988), optimal swimming speed models that have considered such time horizons produced qualitatively similar predictions to those obtained using this simpler (μ/g) approach (Werner and Anholt 1993).

In our model, mortality rate (μ) of our intermediate consumer was characterized as a function of swimming speed (s), using the equation:

display math

The constant q represents mortality independent of s and is constrained to positive values in all of our model derivations. The coefficient m and exponent x relate s to μ. In the context of our model, m is the rate at which top predators are encountered given s, assuming that the direction of predator and intermediate consumer movement and their spatial distribution are random, whereas x represents changes in detectability with s (Werner and Anholt 1993). If x = 1, encounter with a predator is directly proportional to s. When x > 1 or if 0 < x < 1, then predator encounter rate accelerates or decelerates, respectively, with increases in intermediate consumer s. This depiction of s-dependent μ has proven to be a good proxy for wide range of consumer swimming speeds in three-dimensional (3-D) space (Gerritsen and Strickler 1977).

In general, increased turbidity (T) reduces intermediate consumer μ by lowering the reaction distance (R) of the predator. The dependence of R on T is best characterized by an exponential decay or hyperbolic relationship (Utne-Palm 2002); thus, we represent R using the term eaT, where the exponent a represents the exponential rate of decay of the predator's R with increasing T. In 3-D space, the rate at which predators encounter prey is proportional to the product of R2 and predator density (P), assuming that R of the predator is constant in all directions (Gerristen and Strickler 1977). To incorporate the effect of T into Eq. 1, we substituted m with the term (eaT)2P.

We assumed g to be directly proportional to foraging rate (f), which was characterized as a function of s using a Type-II functional response equation (Holling 1959):

display math

where the coefficients z and h represents zooplankton (basal resource) density and handling time, respectively. In so doing, we assumed that an increase in s is not energetically costly, based on the insignificant amount of energy expended by planktivorous fish on swimming relative to other metabolic demands (Klumb et al. 2003). As with our predator, increases in T reduce the R of our intermediate consumer. Thus, the effect of T was incorporated into Eq. 2 by multiplying zs in the numerator and zhs in the denominator by (ebT)2, where b represents the exponential rate of decay of our intermediate consumer's R with increasing T. Importantly, the exponents a and b represent the same phenomenon, differing only in that they characterize T's effect on the visual detection of the prey by our top predator and intermediate consumer, respectively. Because increased turbidity has a greater negative effect on the visual capabilities of piscivores than planktivores (De Robertis et al. 2003), we assume a > b in our model.

To find s*, the s that minimizes μ:g, as well as to quantify how s* varies with P and T, we differentiated the quotient of Eqs. 1 and 2, here simplified as

display math

with respect to s, set the derivative equal to zero, and solved for s. We then derived solutions across a range of x values, starting with x = 1 (i.e., encounter with a predator is directly proportional to s). With x = 1, s* is given by:

display math

Eq. 4 above shows that an increase in T will always lead to an increase in s*. Turbidity-induced increases in s, however, do not necessarily equate to increases in f (using g as its proxy) because T also has a negative effect on f through its reduction of the intermediate consumer's R. For example, fish may increase their s just enough to exactly offset the reduction in f due to diminished visibility. To determine the net effect of T on f, we substituted s* for s in Eq. 2, obtaining the equation:

display math

Eq. 5 provides several interesting predictions related to optimal foraging, many of which have been considered by Werner and Anholt (1993). Here, we focus on the novel and most germane prediction to this study, which is the relationship among foraging, turbidity, and predation risk. We found that, as T increases, f(s*) is also predicted to increase, when a > b.

Results

Below, we discuss the relationship between T and intermediate consumer f, as well as its sensitivity to changes in the predation-related parameters, P and x (see Fig. 1). In so doing, we also discuss the prevalence of TMIEs.

Figure 1.

Relationship between turbidity, T and the optimal foraging rate, f(s*), of our intermediate consumer, given different probabilities of being captured by an encountered predator (x): (A) x = 0.5, (B) x = 1, and (C) x = 2. In each panel, predator density (P) varies from low (P = 0) to high (P = 5). Values of other parameters common to all panels are: mortality independent of swimming speed, q = 0.4; zooplankton (basal prey resource) density, z = 10; zooplankton prey handling time, h = 0.1; the exponential rate of decay of our top predator's reactive distance (R) with increasing T, a = 0.2; the exponential rate of decay of our intermediate consumer's R with increasing T, b = 0.05; maximum swimming speed, smax = 50; and T ranges from 0 to 40. The range of parameter values were chosen to illustrate different possible shapes of the curve.

When we consider the scenario in which piscivorous predators are absent (P = 0; solid lines in Fig. 1), our zooplanktivorous intermediate consumer swims at its maximum s regardless of T. In addition, its consumption of the basal prey resource declines with increases in T (owing to reductions in R) and the TMIE on the basal resource is nil (because P = 0).

When a predation risk is added to the model (P > 0), evidence for context-dependent TMIEs exists. With no T or low levels of T (i.e., turbidity as a refuge is nil or minimal, respectively), the intermediate consumer's s is relatively low, as is its consumption of the basal resource (f) (i.e., predator-driven TMIEs become evident, which are maximal at T = 0). As P increases, reductions in f become more severe, as does the magnitude of the TMIE. As T increases from low to intermediate levels, our intermediate consumer increases its s (and hence, its f on the basal prey resource increases), which more than offsets the negative effect of T on intermediate consumer's R. Concomitantly, the importance of TMIEs decreases, as T reduces predation risk. At some intermediate level of T, however, the intermediate consumer achieves an s that corresponds to its maximum s, and its consumption of the basal resource (f) becomes the same as that predicted when predation risk is absent (see areas of convergence between dashed and solid lines in Fig. 1). Functionally, TMIEs also become non-existent at this point. Beyond this point, T diminishes the ability of the intermediate consumer to detect the basal resource, but has no additional influence on s. Thus, f declines precipitously as T increases beyond intermediate levels (descending solid lines in Fig. 1), with the importance of the TMIEs remaining nil.

Overall, our model predicts a hump-shaped (unimodal) relationship between intermediate consumer consumption of a basal prey resource and turbidity when the risk of predation is present, whereas foraging by a zooplanktivorous intermediate consumer is predicted to be negatively related to turbidity in the absence of predation risk. Likewise, in the presence of predation risk, our model predicts the positive TMIE of the top predator on the basal resource to peak with no turbidity and decline thereafter, until it reaches an asymptote of zero at intermediate levels of turbidity (i.e., when intermediate consumer foraging becomes the same as that predicted when the top predator is absent).

Thus far, we have focused on model predictions when the encounter rate with a predator is directly proportional to s (i.e., x = 1; Fig. 1B). However, the same general pattern is predicted if 0 < x < 1 (i.e., predator encounter decelerates with increasing intermediate consumer s; see Fig. 1A) or if x > 1 (i.e., predator encounter decelerates with increasing intermediate consumer s; see Fig. 1C). The primary inference from this similarity, as it pertains to our study, is that the humped-shaped relationship between turbidity and foraging is robust to curvature changes in the relationship between s and μ.

Field Observations

Methods and analysis

We explored patterns in nature to test our theoretical predictions of turbidity-dependent TMIEs. Specifically, we quantified consumption rates of age-0 zooplanktivores (i.e., the intermediate consumer) in the presence of predation risk in several, vastly different ecosystem types: (1) yellow perch from Lake Erie and its embayments; (2) black and white crappie from 12 Ohio reservoirs; and (3) bluegill sunfish from three Alabama ponds. We chose these systems, because each presented a wide-ranging gradient in turbidity (Table 1) and each differed from the others in terms of geographic location, size, and community assemblage. In turn, we could then make more general inferences based on similarities and differences found among these systems. Because each of these ecosystems contained fish predators (Lake Erie: Carreón-Martinez 2012; OH reservoirs: Bunnell 2002; AL ponds: Partridge and DeVries 1999), we predicted a hump-shaped relationship between consumption rate and turbidity (see Model section).

Table 1. Summary of field collections of age-0 zooplanktivores.Thumbnail image of

Sampling and laboratory methods regarding field collections are described briefly here, and in detail in Appendix A. Sampling targeted zooplanktivorous age-0 fish (<50 mm total length, TL; Table 1) that resided in the upper 3 m of the water column. Stomach content analysis was performed on individuals (Appendix A), with foraging rate calculated as the total number of prey in stomachs. At the time of collection, ambient crustacean zooplankton density (number of individuals/m3; the basal resource) was measured, as was Secchi depth (i.e., water clarity; nearest 1 cm), which is strongly inversely related to turbidity (e.g., Shoup and Wahl 2009: R2 = 0.99).

To test whether a hump-shaped pattern existed in stomach content with respect to turbidity, we used quantile regression, an approach that can capture changes in the boundaries of data distributions (Scharf et al. 1998) and has previously been used to interpret diet data (Scharf et al. 1998, Pinnegar et al. 2003, Menard et al. 2006). Stomach content data are notoriously variable and difficult to analyze, in part due to factors unrelated to the interaction between a consumer and its prey, including post-collection digestion and collection-induced regurgitation (Bowen 1996), which can depress values and are difficult to account for. We dealt with these putative hidden biases by focusing on how turbidity constrained the upper limit of stomach content. We characterized the 0.99 quantile of stomach content as a function of turbidity (i.e., Secchi depth), using the function inline image in its linearized form: ln(y + 1) = β1ln(X) + β2X + ln(ε), where y and X are stomach contents and turbidity, respectively, and β1 and β2 are fitted parameters. This equation can capture humped-shaped relationships in ecological data (Cade and Guo 2000) and has some tractable qualities: (1) β1 must be greater than 0 and β2 less than 0 to produce a convex (hump-shaped) relationship; and (2) −β12 is equal to the X value that maximizes (or minimizes) y. Also, this function assumes that consumer stomach content approaches zero as turbidity becomes very high, which is biologically reasonable given that extreme turbidity can inhibit the ability of an individual to carry out its normal life functions (Newcombe and McDonald 1991). Quantile regressions were run using the rq command in the QUANTREG package in R (R Development Core Team 2008).

Prior to quantile regression analysis, we examined two factors, individual consumer size and ambient zooplankton (basal resource) density, both of which could confound the effect of turbidity on stomach content. Using linear regression, we found that for all three consumers, individual TL was positively related (all R2 > 0.25) to the number of prey items in individual stomachs. To account for this size effect, we relativized individual stomach content by dividing the number of prey consumed by the TL of fish (hereafter referred to as relative stomach content). We found ambient zooplankton density and relative stomach content to be unrelated (all R2 < 0.02) for all species and therefore did not account for it in our analyses.

To estimate the probability of our observed consumer foraging patterns arising by chance, we used a randomization approach. We first shuffled the data such that Secchi depth values were randomly assigned to a randomly-selected relative stomach content value (i.e., complete re-sampling without replacement) and then ran the quantile regression on the randomized data. Doing so allowed us to determine if the fitted relationship (1) was humped-shaped and (2) peaked at a Secchi value less than the median Secchi value. The two-step process was repeated 9,999 times for each species. The number of times that the two conditions were satisfied was divided by the total number of runs to estimate the probability that the humped-shaped relationship occurred by chance alone.

Results

Remarkably similar foraging patterns were observed for our three fish consumers, all supporting model predictions. For Lake Erie yellow perch, relative stomach content exhibited a humped-shaped relationship with Secchi depth (β1 = 0.54, β2 = −0.01; Fig. 2A), peaking at an intermediate Secchi depth (45 cm) and at a foraging level that was more than double of that observed at high levels of water clarity (i.e., deep Secchi depths). In addition, this peak Secchi depth was less than the median value among sampling events, indicating that, for the majority of the 70 sampling events, an increase in turbidity would be expected to increase foraging rate. Humped-shaped relationships also were observed in the relative stomach content of bluegill (β1 = 1.11, β2 = −0.02; Fig. 2B) and crappies (β1 = 0.50, β2 = −0.01; Fig. 2C), peaking at Secchi depths of 64 and 58 cm, respectively. As with yellow perch, the Secchi depth at which the peak was observed was less than the median value for the 61 bluegill and 35 crappie sampling events. Our randomization tests showed that the probability of our hump-shaped patterns arising by chance alone was low for yellow perch (p = 0.04) and bluegill (p = 0.002), but somewhat higher for crappie (p = 0.21). Although we have focused on analysis of the 0.99 quantile, non-random humped-shaped relationships were also evident in analyses of the 0.90 and 0.95 quantiles (Appendix B), indicating the robustness of this pattern.

Figure 2.

Hexagonally-binned field observations of relative gut content of planktivorous age-0 fish in relation to Secchi depth (inversely proportional to turbidity). Gray scale reflects number of observations within each bin (see legend). Observations include (A) yellow perch in Lake Erie, (B) white and black crappie in Ohio reservoirs, and (C) bluegill in Alabama ponds (see Appendix A for more details). For each system, the relationship between Secchi depth and the upper limit of relative gut content (0.99 quantile) was described using a quantile regression model (ln(y + 1) = β1ln(X) + β2X), shown by the lines. Fitted parameter values, β1 and β2 are described in the Field Observations: Results section.

Experiment

Methods and analysis

While intermediate consumer foraging patterns in the field aligned with our hump-shaped modeling prediction, our model also predicted a range of turbidities in which predation risk and turbidity would strongly interact to drive intermediate consumer foraging and the importance of TMIEs (see Fig. 1). Unfortunately, this full prediction could not be tested in the field, owing to the constant presence of predators in our study systems. To help better link our model predictions with our field results, we conducted a 2 × 2 factorial laboratory experiment in which foraging rates of age-0 yellow perch (intermediate consumer) on a basal resource (planktonic Artemia nauplii) were quantified under two conditions of turbidity (0 and 38 NTU) and two conditions of predation risk (absence and presence of predator cues). The 38 NTU turbidity treatment approximates the level at which foraging peaked for our Lake Erie yellow perch (Shoup and Wahl 2009). Thus, our experimental treatments allowed for a contrast between when TMIEs would be expected to be maximal (i.e., predation risk present, clear water) versus minimal (i.e., predation risk present, turbid water). Put another way, we expected our intermediate consumer to forage more rapidly in the turbid- versus clear-water treatment in the presence of predation risk, whereas its foraging rate was expected to be less in the turbid- versus clear-water treatment in the absence of predation risk.

We conducted the experiment in 16 20-L glass aquaria (47 × 24 × 29 cm) that were filled with dechlorinated city water maintained at 25°C (room temperature). The sides of the aquaria were covered with black plastic to limit outside disturbance, with the aquaria being lit from above by 64-watt, full spectrum, fluorescent lights. Four aquaria were randomly assigned to each treatment combination. The turbid-water treatment was achieved by suspending bentonite clay (100 mg/L) in aquarium water, whereas aquaria designated to the clear-water treatment received no bentonite clay. Clays are a primary contributor to turbid environments inhabited by yellow perch (Wellington et al. 2010) and generally scatter, rather than absorb, visible light (Kirk 1994) such that light intensity at the mid-level depth in the aquaria was relatively similar between clear and turbid treatments (30 and 26 μmol·m−2·s−1, respectively).

Predation risk was simulated using both visual and chemical cues (Mikheev et al. 2006). Because adult yellow perch are cannibalistic in many north temperate systems (Tarby 1974), we placed a water-filled (but sealed), clear plastic bag that contained an adult yellow perch into aquaria to provide a visual cue of predation risk. A similar-sized, water-filled bag without an adult yellow perch was placed into aquaria without predation risk. Our chemical cue of predation risk consisted of adding predator kairomones to aquaria water. These kairomones were created by incubating 36 adult yellow perch (mean wet mass ± 1 SD: 17.8 ± 1.76 g) in a 500-L cylindrical tank and feeding them 100 age-0 yellow perch—which were always completely consumed within 2 min—24 hours prior to the experiment. Water (5 L) from the predator tank was filtered through 64-um sieve and transferred into each “predator cue” aquarium 15 min prior to the experimental yellow perch being added. Experimental kairomone concentration reflected a density of adult yellow perch of 0.02 adults/L, which is consistent with maximum total piscivore densities observed in western Lake Erie (Legler 2008). “No predator cue” aquaria received 5 L of filtered water from a 500-L cylindrical tank that did not contain predators.

The 48 age-0 yellow perch used in the experiment were obtained from an aquaculture facility (Mill Creek Perch Farm, Marysville, Ohio, USA) 2 months prior to the experiment, maintained in the laboratory at room temperature under flow-through conditions, and fed a daily ad libitum diet of Artemia nauplii. Three yellow perch (mean ± SD: 39.5 ± 2.3 mm TL) were transferred into each aquarium 15 min prior to the start of the experiment. Experimental fish were starved for 48 hours prior to the experiment to empty digestive tracts.

The experiment was initiated when 1-day old Artemia nauplii (∼100 individuals/L, ∼400 μm TL) were added across the surface of each aquarium. The density and size of zooplankton used in this experiment is within the range of conditions experienced by yellow perch in nature (Reichert et al. 2010). The yellow perch were allowed to forage for 15 min before the experiment was terminated by simultaneously adding Alka-Seltzer tablets to all tanks to euthanize fish via carbon dioxide asphyxiation. Individuals were immediately frozen at −80°C to stop digestion of nauplii prey for later stomach content analysis. Remaining nauplii in the aquaria were filtered out and preserved in 95% EtOH.

Yellow perch stomach contents were extracted and counted under a dissecting microscope. Foraging rates were calculated as the total number of nauplii consumed divided by the allotted foraging time (15 min). Counts of unconsumed nauplii remaining in the aquaria indicate that the relative depletion of prey during the experiment was low (mean ± SD: 31± 16% SD), and thus not a confounding factor in our estimates of foraging rate (Båmstedt et al. 2000).

We used a generalized linear mixed model (GLMM) to quantify the effects of turbidity, predation risk, and their interaction on intermediate consumer foraging rate. Because a significant interaction was found (see below), simple effects within treatments also were evaluated (Stangor 2011). Data were not normally distributed, but rather had a Poisson-like distribution, which is typical of encounter-related measurements (Gerritsen and Strickler 1977). Subsequently, we used quasi-Poisson error distribution in the GLMM that allowed for overdispersion (Zuur et al. 2009). We included the effect of individual aquaria as a random factor in the model. However, we removed it from the final analysis, finding it non-significant (p = 0.42). Statistical models were run using proc GENMOD in SAS 9.1 (SAS Institute 2004).

Results

Foraging rates strongly varied among the experimental treatments (Fig. 3), resulting in a significant interaction (F1,42 = 5.41, p = 0.02) between turbidity and predator cue effects that confounded the main effects (turbidity: F1,42 = 1.93, p = 0.16 ; predator cue: F1,42 = 0.90, p = 0.34). Causes underlying the interaction were evident in the analysis of simple effects and generally supported our model predictions with regard to the existence of predator-driven TMIEs. When predator cues were present, the mean foraging rate on the basal resource by our intermediate consumer in turbid conditions was greater (2.1-fold) than that in clear conditions (F1,42 = 5.55, p = 0.02). By contrast, when predator cues were absent, the mean foraging rate did not differ between turbidity treatments (F1,42 = 0.51, p = 0.48). Further, in clear water, mean foraging rate in the absence of predator cues was significantly greater (2.3-fold) than in the presence of cues (F1,42 = 4.41, p = 0.04). In turbid conditions, however, mean foraging rate did not differ between predator cue treatments (F1,42 = 1.08, p = 0.30).

Figure 3.

Experimental results of mean foraging rate of age-0 yellow perch maintained in either clear or turbid water (nephelometric turbidity units, NTU) in which predator cues were either present (solid circles) or absent (open triangles). Values are means with ±SE.

Discussion

Herein, we sought to understand how turbidity interacts with predation risk to shape interactions between zooplanktivorous fishes, visually-foraging consumers located at an intermediate position in the food web, and their basal resource (zooplankton). Using a mathematical model, we predicted that foraging by our intermediate consumer would decline with increasing turbidity in the absence of predation risk, which is in line with the conventional view of turbidity's effect on visual foragers (reviewed in Utne-Palm 2002). Our model, however, predicted a hump-shaped (unimodal) relationship between turbidity and foraging by our intermediate consumer in the presence of a visual-feeding predator, which has been previously considered (Gregory and Northcote 1993) but not explicitly tested. Concomitantly, our model provided evidence for context-dependency in the importance of TMIEs across a turbidity gradient, with TMIEs decreasing in importance in a non-linear fashion as turbidity increases.

We also provided empirical support for our theoretical predictions. Independent field observations of age-0 yellow perch, crappie, and bluegill foraging in the presence of predation risk showed a humped-shaped relationship between turbidity and consumption on zooplankton. The fact that these relationships were found in ecosystems (Lake Erie, Ohio tributary reservoirs, and Alabama ponds, respectively) of varying size, food-web composition, and limnological characteristics (Appendix A) suggests that our findings are robust to a wide range of aquatic ecosystems. Ideally, our field studies would have included observations from natural systems absent of predators. Not being possible, we conducted an experiment to help fill this gap. It demonstrated a strong, interactive effect of turbidity and predation risk on foraging by a visual planktivore (age-0 yellow perch), with greater foraging of this intermediate consumer occurring in moderately turbid water versus clear water in the presence of predator cues. Although previous studies have shown an interactive effect of turbidity and predation risk on the behavior of planktivorous fish (Gregory 1993, Miner and Stein 1996, Abrahams and Kattenfeld 1997), including foraging-related behavior (Lehtiniemi et al. 2005), none to our knowledge had actually measured changes in interaction strength (i.e., foraging rate), thus necessitating our experiment. Further, previous studies often have attributed these turbidity-driven changes in the behavior of intermediate consumers to their inability to sense predators at high turbidity (Abrahams et al. 2007). Recent studies have failed to support this hypothesis by showing that fish can still sense predators in turbid water through the use of chemical cues (Lehtiniemi et al. 2005, Leahy et al. 2011). Our study reveals a mechanism in which behavioral responses to turbidity can still be adaptive even when fish can sense predators in turbid water.

In interpreting our experimental and field results, we assume that changes in planktivore foraging occurred due to a change in their behavior, as predicted by our model. An alternative explanation is that increased turbidity affected the behavior of zooplankton prey, making individuals somehow more susceptible to predation. For example, the Artemia nauplii used in our experiment are known to respond to the presence of planktivorous fish by moving vertically downward (McKelvey and Forward 1995) and perhaps this light-dependent avoidance response is weaker in turbid versus clear water. Such an explanation is unlikely to be the primary cause of our experimental results because (1) turbidity-related changes in foraging rates only occurred in the presence of piscivore cues and (2) downward migration would not be an effective deterrent of predation in our shallow aquaria. We also feel that induced downward migration is an unlikely explanation of our field patterns, as (1) our observations generally occurred at sites so shallow that migration was prohibitive and (2) they indicate that migration was unrelated to turbidity (see Appendix C for additional supporting analyses of field data). Another alternative explanation for increased foraging with increased turbidity is that turbid water causes zooplankton prey to contrast with its background more than in clear water thus making individuals more conspicuous (“added contrast hypothesis”, reviewed in Utne-Palm 2002). This hypothesis, however, has not been demonstrated empirically (Utne-Palm 1999), and in fact, much evidence exists to counter it (e.g., Barrett et al. 1992). Our experimental findings also do not lend support for the added contrast hypothesis, as no increase in foraging was observed in our turbid, no-predator cue treatment.

Previous studies of turbidity's effect on foraging by visually feeding zooplanktivorous fish have shown mixed results, with some finding negative effects (e.g., Gardner 1981, Wellington et al. 2010) and others finding positive effects (e.g., Boehlert and Morgan 1985, Gregory and Northcote 1993, Miner and Stein 1993). Gregory and Northcote (1993) provided some empirical data to help understand how both sets of findings are valid, suggesting that turbidity interacts with predation risk to mediate fish consumer foraging. These authors even generated hump-shaped relationships between intermediate consumer foraging and turbidity in the laboratory that were quite similar to our own. Importantly, however, Gregory and Northcote (1993) did not explicitly contrast intermediate consumer foraging in the presence and absence of predation risk. Because their hump-shaped curves were derived in the absence of predation risk, Gregory and Northcote (1993) assumed that their hatchery-reared fish consumer (Chinook salmon Oncorhynchus tshawyascha) was still perceiving risk and behaving accordingly in the laboratory. Our study clearly demonstrates the danger of such an assumption, as our hatchery-reared intermediate consumer responded differently in the presence versus absence of predator cues. Likewise, the behavioral response of Wellington et al.'s (2010) intermediate consumer (age-0 yellow perch) to turbidity in their experiment, which lacked a predator cue (similar to Gregory and Northcote 1993), was more akin to the foraging response of yellow perch both in our predator-free experimental treatment and in our theoretical model without predation risk (i.e., foraging declined with increasing turbidity). Regardless, our study certainly validates Gregory and Northcote's (1993) initial hypothesis that moderate levels of turbidity can maximize foraging by intermediate consumers in the presence of predation risk.

The observed effects of predator presence in low turbidity were non-trivial, and the ability of predation risk to reverse the effect of turbidity on visual foragers will certainly require us to re-conceptualize the broader impact of turbidity on aquatic systems. For example, studies that have modeled the influence of turbidity at the population or community level have almost exclusively represented foraging as a decreasing function of turbidity (but see Harvey and Railsback 2009). Replacing this function with a hump-shaped one would likely greatly affect model predictions, as such non-linearity has been shown to have profound effects on ensuing predator-prey and community dynamics (Peacor and Werner 2004). Also, the positive effect of turbidity on planktivore foraging, as demonstrated herein, may contribute to the apparent positive effects of turbid conditions on fish recruitment. For example, age-0 fish inhabiting turbid river plumes in lake and coastal marine systems generally exhibit a foraging, growth, and survival advantage relative to fish outside of the plume (Grimes and Kingsford 1996, Reichert et al. 2010). Although this positive effect often has been attributed to enhanced resources in the plume due to the relatively high nutrient concentrations (Grimes and Kingsford 1996, Garcia-Isarch et al. 2006), a recent study has found more support for a predator-driven mechanism that includes behaviorally-driven increases in foraging (Reichert et al. 2010). In addition, our results help explain zooplankton community patterns in turbid reservoirs, which exhibit as strong or even stronger impacts of planktivory than zooplankton communities in clear-water systems (Schulze 2011).

A clear strength of our model is that its predictions of a hump-shaped relationship are robust to changes in parameters values, as long as some natural conditions are met (e.g., increased turbidity has a greater negative effect on the visual capabilities of piscivores than planktivores). A downside to our approach, however, emanates from its simplicity, which was necessary to derive general analytical solutions that could allow for such broad predictions. Thus, while our model is able to predict the general shape of the risk-turbidity-foraging relationship, which we argue has relevance to a variety of aquatic ecosystems, we caution against its use for generating explicit predictions such as specific turbidity levels at which foraging peaks. Instead, our model may serve as a foundation for the creation of future models that incorporate greater complexity and details, as well as more specific predictions.

Our modeling, field observational, and experimental evidence have implications for understanding the relative importance of TMIEs in ecosystems that are dominated by visual consumers residing at intermediate positions in the food web. While our study focused on turbidity, other factors that are likely to influence TMIEs in a similar way include shade (as a result of canopy cover, for example), habitat complexity (e.g., macrophyte density), and fog. For example, Rothley and Dutton (2006) demonstrated experimentally how canopy cover-induced shade could modify TMIEs in an old field community by interacting with perceived predation risk by the intermediate consumer. Specifically, these authors showed that, in unshaded (i.e., more risky) conditions, their intermediate consumer (the grasshopper, Melanoplus sanguinipes) foraged less, and in turn, had a weaker overall impact on the basal resource (forb plants) than in shaded (i.e., less risky) conditions. Similarly, Trussell et al. (2008) showed that in the presence of a top predator (green crab Carcinus maenas), the plastic foraging behavior of an intermediate consumer (the snail, Nucella lapillus) on a basal resource (either barnacles, Semibalanus balanoides, or mussels, Mytilus edulis) depended strongly on habitat complexity (mussel vs. barnacle shells). Complex habitat provided a refuge for the intermediate consumer, thereby reducing the costs (risks) associated with foraging and simultaneously magnifying the interaction between the intermediate consumer and a basal resource (also see Grabowski et al. 2008 for similar findings involving TMIEs on oyster reefs).

Few previous studies, however, have considered the full natural range of such environmental factors, thus prohibiting generalization over large environmental gradients. Indeed, most experiments relevant to the context dependence of TMIEs have been conducted in a factorial design in which two or three levels of the factor are employed (e.g., Grabowski et al. 2008, Trussell et al. 2008). Although these experiments are important in demonstrating the potential of context dependence of TMIEs, our results and others (e.g., Large et al. 2011) indicate that consideration of TMIEs over broader ranges may reveal important non-linear relationships between environmental factor and species interactions.

Our study adds to a growing collection of others showing that explicit consideration for context-dependence in adaptive foraging and resultant TMIEs, particularly as it relates to factors that can be extremely variable over time and in space, is paramount. This need seems especially relevant as predator-prey interactions appear to depend more on risk-related environmental factors than more conventional factors (e.g., prey density), which was apparent in our field observations and those by Heithaus et al. (2009) in a nearshore marine community. While numerous environmental factors are likely to influence the tradeoffs associated with antipredator responses (e.g., resource availability, Luttbeg et al. 2003; ecosystem productivity, Turner 2004, Werner and Peacor 2006; landscape features, Heithaus et al. 2009), we see a particular need for research focused on factors that can affect sensory performance. Some research already has been conducted in this arena with regard to effectors on visual (e.g., shade, Rothley and Dutton 2006) and chemical (e.g., wind, Carr and Lima 2010; pH, Turner and Chislock 2010; water flow velocity, Large et al. 2011; tidal regime, Kimbro 2012) perception by consumers at a single trophic level. Our study suggests how such factors simultaneously and differentially influence the sensory performance of consumers at other trophic levels may also be important to accurately predicting NCEs and TMIEs. Because adaptive phenotypic plasticity can shape interactions across multiple levels of ecological organization (see reviews by Werner and Peacor 2003, Miner et al. 2005, Ohgushi 2005), we strongly support the view of others (e.g., Relyea 2004, Peacor and Werner 2004, Agrawal et al. 2007) who condone research aimed at identifying the mechanisms that underlie context dependence in predator-driven adaptive phenotypic plasticity. Only through such studies will we begin to move closer to understanding how species interactions, and in turn communities, are structured.

Acknowledgments

We thank both past and present members of the Aquatic Ecology Lab at The Ohio State University (OSU), for assistance with the experiment and OH reservoir crappie and Lake Erie yellow perch field collections. J. Mion and E. Roseman provided some of the Lake Erie yellow perch and Lake Erie water quality data and W. Lynch supplied yellow perch for our experiment. Several biologists from the Ohio DNR–Division of Wildlife also assisted with field collections in Ohio reservoirs. W thanks personnel at Auburn University's Ireland Center for help with the Auburn pond and Walker County State Fishing Lake sample collection and processing. Tammy DeVries, in particular, processed all of the Auburn zooplankton data. We also thank J. Drake, Y. Lou, W. Pangle, S. Peacor, R. Tien, and two anonymous reviewers for providing helpful comments on previous versions of this manuscript. Support for this manuscript was provided by (1) the Great Lakes Fishery Commission's Fishery Research Program (to SAL), (2) a Federal Aid in Sport Fish Restoration F-69-P project (to R. A. Stein), administered jointly by the U.S. Fish and Wildlife Service and Ohio DNR–Division of Wildlife, (3) an OSU Undergraduate Research Office Summer Fellowship to TDM, (4) Federal Aid in Fish Restoration Project F-40-R administered by the Alabama Department of Conservation and Natural Resources, Division of Wildlife and Freshwater Fisheries (to DRD), and (5) National Science Foundation grants DEB-9108986 and DEB-9410323 (to DRD). This manuscript is Contributions 1719 and 25 of the USGS Great Lakes Science Center and Central Michigan University Institute of Great Lakes Research, respectively. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Supplemental Material

Appendix A

Descriptions are provided for field collections of yellow perch (Perca flavescens), black and white crappie (Pomoxis nigromaculatus and P. annularis, respectively), and bluegill sunfish (Lepomis macrochirus).

Yellow perch

Daytime collections of limnetic age-0 yellow perch occurred weekly during late April through June in Ohio waters of Lake Erie at three locations: Sandusky Bay during 1994–1998 (n = 6 sites), Maumee Bay during 1996 (n = 6 sites), and near the mouth of the Huron River during 1996 (n = 3 sites). With exception of April–May 1994–1995, at which time surface-towed 500-μm ichthyoplankton nets (0.75-m diameter; average speed (SPD) = 2.4 m/s) were used, surface neuston nets (1 m × 2 m) were deployed to increase sampling volume. We initially equipped neuston nets with 500-μm mesh (SPD = ∼1.0 m/s), increasing to 1000-μm (SPD = ∼1.3 m/s) and 1800-μm (SPD = ∼1.5 m/s) mesh as fish became larger and faster (Mion et al. 1998). The duration of all tows ranged 1.5 to 6 min. Water volume sampled was estimated with a General Oceanics flowmeter (Miami, Florida, USA) mounted in the mouth of nets. Upon collection, all fish were preserved in 95% EtOH.

After each tow, temperature (nearest 1°C) and Secchi disk transparency (nearest 1 cm) were measured and zooplankton was collected using vertical hauls with a 0.30-m diameter plankton net. One or two hauls were conducted at each site within each location (i.e., 3–12 hauls/location/day). In the Maumee and Sandusky bays, zooplankton nets were equipped with 153-μm mesh. The net used at the Huron site had 63-μm mesh. The use of different mesh sizes in these exact nets has been shown to have no effect on the estimation of abundance for the zooplankton taxa primarily consumed by age-0 yellow perch used in this study (i.e., crustacean zooplankton larger than rotifers and copepod nauplii; Mack et al. 2012). All zooplankton was preserved in 70% ethanol (EtOH) until laboratory analysis.

Zooplankton abundances were calculated using the methods of Stahl and Stein (1994). Cladocerans were identified to genus and copepods were categorized as calanoids, cyclopoids, or nauplii. To estimate abundance, zooplankters were rinsed into a round counting dish consisting of 16 sections. For each sample, >50 individuals of an abundant taxon from at least 1/8 (i.e., two sections) of a mixed subsample were counted. For rare taxa (i.e., taxa in which <50 individuals were counted in 1/8 of the subsample), if the extrapolated number of individual per taxon was >25, we counted remaining sections until either 50 individuals were counted or we processed the entire sample. If the extrapolated number was <25 after counting the first 1/8 of the subsample, the abundance of the taxon was extrapolated from the count in the initial 1/8 subsample (Stahl and Stein 1994).

When possible, we processed the diets of 10 non-empty yellow perch/site/date. We counted and measured all zooplankton the entire digestive tract under a dissecting microscope (2.5× magnification). Cladocerans were identified to genus, and copepods were categorized as calanoids, cyclopoids, or nauplii. All yellow perch were measured (nearest 0.1 mm TL) prior to analysis under a dissecting microscope with an image analysis system.

Crappie

Daytime collections of limnetic larval white and black crappie occurred weekly during May through June, 1998–2000, in 12 Ohio reservoirs. Samples were collected with 1 m × 2 m neuston nets (500-μm mesh) at the water column surface (n = 3 sites/reservoir; near the inflow, the middle, and the dam). At least two replicate, 5-min tows (SPD = ∼1 m/s) were conducted. Water volume sampled was estimated with a General Oceanics flowmeter (Miami, Florida, USA) mounted in the mouth of nets and all fish were preserved in 95% EtOH upon collection. Secchi disk transparency, temperature, and zooplankton were measured at each site and with yellow perch (diets and lengths) and zooplankton processed in the same manner as described for Lake Erie yellow perch above.

Bluegill

Daytime collections of limnetic larval bluegill occurred weekly during April through October 1994 in two ponds (3.8–4.5 ha) on the E.W. Shell Fisheries Center North Auburn Unit at Auburn University and in 2001 from Walker County State Fishing Lake, Jasper, Alabama (WCL). Collections were made with a bow-mounted ichthyoplankton net (0.5-m diameter, 500-μm mesh) pushed at 1.0–2.0 m/s. Water volume sampled and towing speed were estimated with a General Oceanics flowmeter (Miami, Florida, USA) mounted in the mouth of the net. Three replicate samples (2–3 min duration each) were collected in the Auburn ponds and four replicate samples (4 min duration each) were collected from WCL on each sampling date; all samples were preserved in 95% EtOH upon collection.

At each site on each sampling day, water temperature profiles were recorded as described above. Three zooplankton samples (four in WCL) were collected using a 2-m tube sampler (7.5-m diameter; DeVries and Stein 1991), filtered through 54-μm mesh netting and preserved in 10% sucrose formalin solution (Auburn ponds) or 95% EtOH (WCL). Zooplankton were identified using a dissecting microscope (to genus for cladocerans or as nauplii, calanoids, or cyclopoids for copepods) and enumerated (in measured subsamples when necessary) until at least 200 individuals from the most abundant taxon were counted or until the entire sample was counted.

Bluegill diets were quantified in biweekly samples (n = 5 larvae per 1-mm size class; n = 15 larvae/date in WCL). All prey in fish digestive tracts were removed and identified to genus for cladocerans or as nauplii, and as calanoids, or cyclopoids for copepods. All larvae were measured in an identical fashion as yellow perch and bluegill.

Appendix B

Table B1. Parameter values from quantile regression analysis that related turbidity (X) and stomach content (y), using the equation ln(y + 1) = β1ln(X) + β2X + ln(ε) at the 0.90 and 0.95 quantiles.Thumbnail image of

Appendix C

In Lake Erie, our sampling sites were located in unstratified waters <5 m deep, thus constraining zooplankton prey to areas accessible to larval yellow perch, regardless of turbidity (Cole 1978, Patterson 1979). In Alabama ponds and in Walker County State Fishing Lake, zooplankton were sampled from only the top 2 m of the water column such that only zooplankton that directly overlapped with bluegill larvae were included in the zooplankton density estimate (Appendix A). Further, because turbidity and zooplankton densities were not significantly correlated (p = 0.07) in these ponds and actually trended in a negative direction (more zooplankton in clearer water), the overlap between zooplankton and bluegill larvae likely does not increase with increased turbidity.

In contrast to Lake Erie and Alabama systems, some sampling sites for the Ohio reservoirs were deep (up to 37 m) and zooplankton densities were estimated from net tows of the entire water column, thus creating the possibility for our results to be confounded by turbidity-dependent zooplankton vertical migration. To address this issue, we re-ran our quantile regression analysis using only Ohio reservoir sites that were <3 m deep (17 sites, 445 fish), which is the range of depths in which age-0 crappie concentrate in these systems (Arend 2002). We found the same relationship using this subset compared to the full dataset: a humped-shaped curve (β1 = 0.50, β2 = −0.01, peak at Secchi depth of 42 cm) with a peak less than the median Secchi depth (74 cm). Therefore, overall, evidence does not support the alternative explanation that that increased turbidity affected the behavior of zooplankton, making them more susceptible to predation. But, we do acknowledge that this mechanism may occur in some ecosystems, both freshwater and marine, and in turn, may further increase the consumption of zooplankton by visual planktivores in turbid conditions.

Ancillary