Disturbance legacies and nutrient limitation influence interactions between grazers and algae in high elevation streams


  • Corresponding Editor: D. P. C. Peters.


Debate about control of interaction strength among species is fueled by variation in environmental contexts affecting food webs. We used extensive surveys and two field experiments to test the individual and interactive influences of variation in the assemblages and associated traits of grazers as shaped by the legacy of disturbance, nutrient limitation and the presence of top predators on the accrual of basal resources. We quantified hydrologic variation and streambed movement to describe the legacy of disturbance and sampled biota of 20 streams over five years in a high-elevation catchment in Colorado, USA. Grazer assemblages switched from caddisfly-dominated to mayfly-dominated as disturbance increased. We manipulated the composition of grazer assemblages and the availability of nutrients (N and P) within flow-through mesocosms assembled adjacent to 10 streams, and also deployed larger in-stream channels manipulating the presence of top predators (brook trout) in five streams varying in disturbance regimes. In both experiments we compared the rate of accrual of benthic algae and the strength of grazer-algal interactions among treatments. We observed no indirect effects of top predators on grazer mobility, grazer consumption of algae, or accrual of algal biomass (no trophic cascades). However, in both experiments accrual rates of algae yielded a unimodal pattern and grazer impacts on algae decreased with increasing disturbance, but only at ambient (limiting) nutrient conditions. When nutrients were amended in the mesocosm experiment, algal accrual was uniformly high and grazer impacts on algae were consistently low. Reduced algae accrual at high disturbance levels may be explained by direct effects of environmental harshness on algae, and at low disturbance by indirect effects on grazer traits (behaviors) rather than on grazer density. In more benign streams per capita and per unit biomass grazer impacts on algae were high and drift dispersal was low, both behaviors that reduced accrual of algae. We conclude that nutrient limitation and indirect effects of disturbance on accrual of algae mediated by grazer traits can be stronger than indirect effects of predators on algae, providing a new contribution to the debate about the influence of environmental context on the strength of food web interactions.


The food web concept may be simultaneously one of the simplest and most challenging in ecology. The readily comprehensible elementary school version (bird eats worm, cat eats bird) belies the complexity of natural food webs, often with a maze of direct and indirect interactions. Understanding patterns of variation of species interactions requires investigating how environmental context affects food web processes (Terborgh and Estes 2010). To that end, innovative empirical research over a range of environmental conditions is needed to test contemporary theory and predictions related to variation in specific food web functions, such as the strength of herbivore-plant interactions.

One fundamentally important question with regard to food webs is whether consumers exert predominant control over the distribution and abundance of primary producers (Power 1992a). Strong consumer control of plants has been demonstrated in a variety of terrestrial (Tilman and Pacala 1993), freshwater (Power 1992b) and marine (Paine 2002) habitats. However, variation in the strength of herbivore-plant interactions depends not only on the vulnerability of plant taxa to herbivory, but also on aspects of the environment including the disturbance regime, habitat complexity and nutrient availability (Hunter and Price 1992, Rosemond et al. 1993, Persson et al. 1996). A meta-analysis of experimental studies on herbivore-plant interactions in streams concluded that stream herbivores have stronger control of their resources than herbivores in other systems, contrary to the view that streams are regulated primarily by abiotic factors (Feminella and Hawkins 1995). However, lack of data on the roles of environmental factors, such as the disturbance regime, the predation regime and nutrient limitation, constrains generalizations about key factors explaining variation in the strength of grazer-algal interactions.

Empirical work is needed to test general predictions that features such as disturbance, environmental complexity, resource supply, the presence of invulnerable taxa, and trait-mediated effects influence the strength of top-down interactions (Oksanen et al. 1981, Abrams 1995, Power and Dietrich 2002, Bolker et al. 2003). Theories developed specifically for streams have considered the role of environmental heterogeneity and the dynamic nature as well as the importance of disturbance in streams (Townsend 1989). Recent conceptual frameworks challenge ecologists to explicitly incorporate changes in streams across the landscape (or “riverscape”) into our understanding of stream food webs (Power and Dietrich 2002, Woodward and Hildrew 2002). In this study we addressed relevant aspects from those models and general ecological theory to test how three fundamental characteristics of streams that vary across the riverscape: the amount and timing of flow-related disturbance, the availability of resources and the presence of top predators, affect the strength of herbivore-plant interactions.

The disturbance regime can influence herbivore-plant interactions directly and indirectly by shaping the composition and abundance of both consumers and primary producers (Peterson 1996). There is increasing recognition that attributes of intermediate consumers affect the regulation of energy flow as much as attributes of top predators or resources (Hunter and Price 1992, Power and Dietrich 2002). As a consequence, the influence of the disturbance regime on the composition of grazers may alter the susceptibility of the system to strong grazer-algal interactions. Some of the most striking changes in the composition of benthic macroinvertebrate communities are associated with changes in flow-related disturbance (floods and stream drying; Death and Winterbourn 1995, Townsend et al. 1997). Power and Dietrich (2002) argue that river disturbance most severely affects basal taxa, such as benthic grazers and algae, because they are generally rock-bound.

In streams of the Colorado Rocky Mountains, floods are predictable in timing but variable in magnitude among streams within and between watersheds (Poff 1996). We have shown previously that the extent of high-flow disturbance associated with increased bed movement can be highly variable within one high elevation snowmelt system (Peckarsky et al. 2014). Furthermore, differences in hydrologic and geomorphic characteristics among streams may be associated with variation in the assemblage of consumers, which thereby reflects the “legacy of disturbance.” Taxa like cased-caddisflies, whose movement is hampered by stony cases, may be less resilient and resistant to flow-related disturbances (Wootton et al. 1996, Lytle 2001, 2002). In contrast, highly mobile mayflies are more resilient to flow-related disturbance (Robinson and Minshall 1986, Boulton et al. 1988). Therefore, in this study we tested the expectation that the grazer assemblage of more disturbed streams would be dominated by mobile, disturbance-tolerant taxa whereas armored taxa, which are more vulnerable to disturbance, would predominate in more benign streams.

Another major influence on the strength of herbivore-plant interactions is resource availability. Stream algae are often limited by the availability of nutrients (Borchardt 1996), and consequently, nutrient limitation has the potential to affect the strength of grazer-algal interactions. The abundance of benthic algae reflects counteracting dynamic processes affecting accrual (e.g., nutrients) and loss (e.g., physical disturbance and grazing; Biggs 1996, Roll et al. 2005). Therefore, we tested the influence of nutrients on the strength of grazer-algal interactions across a range of disturbance regimes and associated grazer assemblages in recognition of the importance of considering the interaction between simultaneous influences of top-down and bottom-up controls on primary producers (Rosemond et al. 1993, Forrester et al. 1999, Flecker et al. 2002).

We have demonstrated previously that grazers can have strong top-down effects on distribution and abundance of algal biomass in the East River streams (Taylor et al. 2002, McIntosh et al. 2004). Furthermore, we have observed indirect effects of top predators on the biomass of algal resources via their influence on grazer behavior (behavioral trophic cascades), but only in simplified experimental food webs involving mobile grazers that are vulnerable to predation, and where limiting nutrients are augmented (Peckarsky and McIntosh 1998, Peckarsky et al. 2013). Importantly, we have not observed strong top-down control of predators on grazer density in our system (Peckarsky et al. 2008). Therefore, under more natural conditions we also tested whether variation in the legacy of disturbance influences the indirect effects of top predators on basal resources by modifying the abundance or behavior of primary consumers. Following the logic of the well-known trade-off between vulnerability to disturbance and to predation (Wootton et al. 1996), we expected trophic cascades to more likely occur in disturbed streams dominated by mobile grazers (mayflies) that have higher biomass-specific consumption and behaviors most responsive to predators (McIntosh et al. 2004, Álvarez and Peckarsky 2005). Trophic cascades were not expected in more benign streams where better-defended grazers (cased caddisflies) predominate, because they are less vulnerable to predatory fish (Allan 1981, Nyström et al. 2003) and less responsive to predators (McIntosh et al. 2004). Finally, based on previous experiments (Peckarsky et al. 2013), we suspected that nutrient limitation that prevails in all streams of this system (Moslemi 2010) would compromise our ability to detect trophic cascades under any disturbance regime.

Based on those expectations, our overall goals were to use surveys and field experiments to test the relative and interactive effects of the legacy of disturbance, nutrient limitation (bottom-up) and top predators (top-down) on the composition of grazer assemblages, the rate of accrual of benthic algae and the interaction strength between grazers and algae in multiple streams of one high elevation catchment. Specifically, we asked the following questions: (1) Does the composition of the grazer community switch from disturbance-intolerant to disturbance-tolerant taxa as the disturbance regime becomes more harsh? (2) How does nutrient limitation (bottom-up) affect the strength of interactions between grazers and benthic algae? (3) Does the trade-off between resistance/resilience to disturbance and vulnerability to predation of grazers affect variation in the strength of top-down interactions along a gradient of disturbance?


Description of study streams

We implemented surveys and experiments in the upper East River catchment near the Rocky Mountain Biological Laboratory (RMBL) in western Colorado, USA (38°57′0″ N, 106°58′59″ W; elevation 2900 m). Streams originate from snowmelt, lake-outlets or are spring-fed; their substrates consist of cobble and boulders underlain by gravel, and their canopies are generally open. Surveys conducted from 2006 to 2010 in 20 streams of this catchment showed that diatoms predominate in more disturbed streams, but bryophytes and filamentous algae occur in more hydrologically stable streams and during low-flow periods (Appendix A: Table A1; Álvarez and Peckarsky 2013). Primary consumers consist of algal grazers, predominantly mayflies (Ephemeroptera) and caddisflies (Trichoptera), and shredders (Plecoptera; Appendix A: Table A2). Top predators are salmonids, primarily brook trout (Salvelinus fontinalis), whereas stonefly (Plecoptera) larvae are the top predators in some streams that are fishless due to natural barriers to fish dispersal (Appendix A: Tables A2 and A3; Peckarsky et al. 2001).

In July 2006 we collected five benthic invertebrate samples with a modified box sampler and measured other variables in all 20 streams that could potentially explain mechanisms of effects of experiments, including nutrient and light levels, water temperatures, discharge, and algal biomass (Appendix A: Table A3). All streams are autotrophic, and have very low dissolved inorganic nutrients (N: 20–495 μg/L, P: <1–7 μg/L) and primary producers are mostly P-limited (Appendix A: Table A4; Moslemi 2010).

Multivariate disturbance index

We defined disturbance as events or environmental conditions that occur when stream discharge changes sufficiently to affect the stability or habitability of the streambed (Peckarsky et al. 2014). Accordingly, we used the first axis of a principal components analysis (PC axis 1) that best describes variation in parameters closely associated with streambed movement as a multivariate index of disturbance (Peckarsky et al. 2014). This index is based on the assumption that physical habitat disruption caused by the movement of streambed particles disturbs benthic organisms (Matthaei and Townsend 2000). To obtain variables for this index we installed stage-height loggers in 20 streams over the widest possible range of physical conditions in the catchment to measure hydrologic variables from 2004 to 2009 and also quantified inter-annual movement of rocks from geo-referenced photographs over one high-flow year (2007–2008). As indicated by a model selection procedure (Peckarsky et al. 2014), the best combination of variables included in the multivariate axis of disturbance (Appendix A: Table A3) were: (1) an empirically-derived index of streambed stability scaled by sizes of particles moved or covered, (2) maximum daily increase in discharge, and (3) a qualitative index of channel stability (Pfankuch 1975).

Mesocosm experiment testing effects of the legacy of disturbance and nutrients

In July 2007 a mesocosm experiment conducted adjacent to 10 of the 20 survey streams (five containing brook trout, and five fishless streams) tested for systematic variation over the disturbance gradient in the rates of algal accrual and the dynamic index of grazer-algal interaction strength (Berlow et al. 1999). We also manipulated nutrient levels to test for interactions between nutrient availability and disturbance legacy on response variables thereby comparing top-down and bottom-up effects following Flecker et al. (2002).

We constructed mesocosms from plastic buckets (diameter = 30 cm, area of bottom excluding drain = 600 cm2) supported by wooden frames on the stream bank (Appendix B: Figs. B1–B4). Four mesocosms were installed at each stream in a 2 × 2 factorial design creating four treatments randomly allocated to buckets: (1) grazers + nutrients, (2) grazers with ambient nutrients, (3) no grazers + nutrients and (4) controls (no grazers and ambient nutrients). Gravity delivered water from each stream, filtered to remove invertebrates, through a manifold to mesocosms, each containing 10 rocks collected from the stream. We scrubbed rocks with a soft-bristled toothbrush to remove most existing biofilm to standardize starting conditions. Because mesocosms were buffered from natural disturbances during the experiment, this experiment tested for the effects of the legacy of disturbance by adding grazer assemblages at natural densities and species compositions reflecting the lasting effects of disturbance regimes on invertebrate abundance and composition (Matthaei and Townsend 2000, Effenberger et al. 2011). Grazers were collected from each stream and placed in the mesocosms allocated to grazer treatments using species composition and abundance observed in 2006 and 2007 benthic samples (Fig. 1; Appendix A: Tables A5 and A6).

Figure 1.

Density of grazing mayflies (squares) and caddisflies (triangles) collected in (A) July 2006 and (B) July 2007 in streams of the upper East River drainage that varied according to a multivariate index of disturbance (Peckarsky et al. 2014), and the presence (red symbols) and absence (blue symbols) of brook trout. The density of mayflies increases (2006: F1,18 = 19.67, P < 0.001, R2 = 0.52; 2007: F1,8 = 26.00, P < 0.001, R2 = 0.77) and the density of caddisflies decreases (2006: F1,18 = 45.74, P < 0.001, R2 = 0.72; 2007: F1,8 = 13.14, P = 0.007, R2 = 0.62) with increasing legacy of disturbance.

For nutrient enrichment treatments we used a fifth bucket containing nutrient-diffusing pellets (plant fertilizer containing NH4 and PO4: continuous release bloom booster; Scotts Miracle-Gro Company, Marysville, Ohio, USA). Nutrient-enriched water was dripped into the two enriched-nutrient mesocosms at a rate of 7.2 ± 0.3 mL/s (mean ± 1 SE), maintaining nutrient levels at ~10× ambient NH4 (mean ± SE = 1.3 ± 0.3 raised to 15.2 ± 0.7 μg/L) and SRP (mean ± SE = 0.7 ± 0.1 raised to 9.8 ± 0.7 μg/L) throughout the experiment, as determined by analysis of water chemistry obtained at 4-d intervals (Moslemi 2010). Ammonium was immediately analyzed using a Turner Designs Aquafluor (Sunnyvale, California, USA) following methods of Taylor et al. (2007), and SRP was measured using a spectrophotometer (Shimadzu 1240) following the molybdate blue method (Murphy and Riley 1962). Ambient (i.e., nutrient-limited) stream water was dripped into control treatments at a similar rate.

After 14 d (13–27 July 2007), with daily filter cleaning, we counted, dried (24 h at 60°C) and bulk-weighed grazers by species (±0.01 mg). From each mesocosm we froze half of the rocks, and within 4 d scrubbed their entire surfaces using a soft-bristled toothbrush, filtered the scrubbate and extracted the filters in 90% buffered ETOH for 12–24 h. (The other five rocks were processed for stoichometric analyses of the epilithon [Moslemi 2010]). We estimated algal pigments using fluorometric methods (Arar and Collins 1997), and measured rock tracings using a leaf area meter (LI-COR Model LI 3100) to standardize estimates of algal biomass by area of streambed.

We analyzed the effects of the legacy of disturbance (as reflected by the grazer assemblage) and nutrient enrichment on accrual of algal biomass, total grazer abundance, and the strength of grazer-algal interactions using linear mixed effects models calculated with restricted maximum likelihood using lme in R (R Development Core Team 2012; version 2.15.0). First, we compared rates of algal accrual and density of total grazers between treatments with grazers present (enriched vs. ambient nutrients and across the disturbance legacy gradient). Second, we compared the dynamic index of grazer-algal interaction strength (DI) between nutrient treatments and across the disturbance legacy gradient (DI = (LN(G/R))/Bt, where G = biomass of chl a in the treatment with grazers, R = biomass of chl a in the treatment without grazers at comparable nutrient levels, B = abundance of the grazers (numbers or biomass at the end of the experiment), and t = duration of the experiment (14 d). Nutrients were included as fixed effects, and disturbance (PC axis 1 from Peckarsky et al. 2014) was included as covariate. Fully factorial models were run initially, and non-significant interaction terms were removed from the final models using model simplification. Non-linear regression in Sigma Plot (version 12.0, Systat Software) was used to fit a unimodal curve to algal accrual rate, which had a non-linear relationship with the multivariate disturbance index in mesocosms with ambient nutrients.

In-stream channel experiment testing effects of predators across a disturbance legacy gradient

A larger-scale in-stream channel experiment tested the effects of a free-ranging predatory fish on the accrual rate of algae and the abundance, foraging impact on algae and dispersal behavior of grazers at ambient (limited) nutrient levels. During a 4-week period of moderate flow in July 2010 we manipulated top predators (brook trout) using channels placed in five trout streams that varied along a gradient of disturbance legacy (Appendix A: Table A3). Other fish streams and all fishless streams used in this study were too small to deploy the channels. Channels are constructed of 1.0 cm diameter aquaculture mesh allowing colonization of algae and invertebrates from all sides and preventing fish from escaping or entering channels, but reducing light levels in the channels by 34%. Channels were attached to wooden frames and had two sides (each 3 m long × 0.5 m wide) separated by a wooden baffle, to enclose a fish in one side and exclude fish from the other (Appendix B: Figs. B5 and B9–B10).

Natural substrates from each stream (10 rocks of standardized particle size distributions) were scrubbed with a soft-bristled toothbrush to standardize starting biofilm and incubated in each stream in five stainless steel mesh baskets (30 cm wide × 20 cm long ×10 cm high with 1.0-cm mesh baffles; Appendix B: Figs. B6–B8). After 2 weeks we measured chl a accumulated in one basket from each stream to estimate starting algal biomass, and placed the other four baskets in the channels, two on each side at 1 and 2 m from the upstream end. Baskets served as discrete, standardized, area-delineated sampling units to measure the effects of the treatments on algal accrual, grazer density and the impact of grazers on algae in each stream. Natural stream substrates were placed around the baskets to fill the channel floor, anchor the channels and provide adjacent substrates from which invertebrates could colonize baskets.

We stocked one randomly selected side of each channel with one brook trout collected from the East River near the study streams (mean ± SE = 181 ± 4 mm SL; 66.3 ± 5.2 g wet weight); and the other side was left fishless. Flow visualizations using dye suggest that chemical cues from the enclosed fish do not cross the wooden baffle separating the sides (Appendix B: Figs. B11–B13). As an unavoidable consequence of our inability to deploy channels in fishless streams, the fishless side of each channel was exposed to unknown and variable background levels of fish cues, which could affect our ability to detect a fish effect in this experiment (Weissburg et al. 2014).

After 2 weeks of exposure to treatments we measured the abundance and composition of grazers and the algal biomass that colonized the baskets. Because treatments without grazers were not possible, we calculated the impact of grazers on algae in baskets using methods comparable to clearance rate coefficients reported by Dodson (1972). We used the per capita grazer impact (GI) on change in algal biomass (chl a) over the experiment to compare the strength of top-down control of algae and indirect effects of predators between the two sides of each channel (with and without fish) and over the disturbance gradient. GI = (LN(GI ) – LN(GF))/Nt, where GI = biomass of chl a at the beginning of the experiment, GF = biomass of chl a at the end of the experiment, N = abundance (density) of grazers, and t = duration of the experiment (14 d). This variable tested mechanisms of observed effects on algae related to grazer foraging performance (behavior).

Although the fish could consume grazers in these channels, previous models and empirical work indicate that the effects of fish at this scale (3 m long channels) most likely reflect predominantly predator-induced changes in grazer behavior (Englund 1997), which is also the principal effect on prey in these systems (McPeek and Peckarsky 1998, Peckarsky et al. 2008). Therefore, after 1 week of exposure to treatments, we estimated grazer drift in the water column to evaluate the trait-mediated indirect effects of predators or disturbance legacy on grazer dispersal. We measured nighttime drift density of grazers into and out of each side of the channels (day drift is negligible in fish streams; McIntosh et al. 2002) by fastening Wildco drift nets (aperture 30 cm wide, 365 μm mesh) to rebar immediately upstream and downstream of channels. We calculated the drift rates (per minute), immigration ratios (drift in/drift out) and drift propensity (drift density out/density in the cages on corresponding sides of the channels), which estimates the probability that an individual drifted (Rader 1997, Wilcox et al. 2008).

For linear relationships we performed homogeneity of slopes tests of an ANCOVA model to evaluate the independent effects of fish and the legacy of disturbance if there was no interaction between variables. Variables were log10-transformed as necessary to meet the assumptions of normality. Non-linear regression in Sigma Plot (version 12.0, Systat Software) was used to fit a unimodal curve to algal accrual rate which had a non-linear relationship with the multivariate disturbance index.


Natural patterns of variation in grazers reflect a gradient of disturbance legacy

As observed over all 20 streams surveyed (2006), and in the 10 streams used for experiments (2007), the composition and abundance of grazers varied significantly along the disturbance gradient (Fig. 1). The density of mayfly grazers increased and the density of caddisfly grazers decreased as the legacy of disturbance regimes became stronger. Those patterns were consistent over all years of the study (Appendix A: Tables A2 and A5–A6).

Mesocosm experiment testing effects of the legacy of disturbance and nutrients

Variation in algal accrual among treatments

The biomass of algae accrued on rocks in treatments containing natural assemblages of grazers varied with disturbance legacy and nutrients. In mesocosms with ambient (limited) nutrients algal accrual had a unimodal relationship with disturbance, increasing, then decreasing with disturbance legacy (Fig. 2A). However, a significant two-way interaction indicated that algal accrual was high and unaffected by disturbance legacy when nutrients were added (Fig. 2A). Streams with low disturbance were fishless and those at the high end of the disturbance gradient had fish, which precluded us from distinguishing the effects of presence of fish cues in the source water from the legacy of disturbance in this experiment. These confounding effects motivated the larger scale in-stream channel experiment to test directly the effects of the disturbance regime and top predators on the accrual of benthic algae (trophic cascade).

Figure 2.

Relationships between the multivariate index of disturbance and (A) the accrual of algal biomass on stones, (B) the density of total grazers, and (C) the per capita grazer-algal interaction strength (see Methods for formula) in mesocosms with enriched nutrients (filled symbols) or ambient (limited) nutrient conditions (open symbols) at the end of the 14 d experiment in July 2007. Source water for mesocosms adjacent to streams contained trout cues in some cases (red symbols) and others were fishless (blue symbols). (A) Unimodal relationship between algal accrual and disturbance under ambient nutrient conditions (nonlinear regression: F2,7 = 11.00, P < 0.01, R2 = 0.76); algal accrual was high and unaffected by disturbance when nutrients were added (nutrient effect: F1,12 = 16.63, P = 0.002; nutrient × disturbance interaction: F1,12 = 4.78, P = 0.049). (B) No effect of nutrients or disturbance legacy on grazer density (nutrient effect: F1,12 = 0.51, P = 0.49; disturbance effect: F1,12 = 0.382, P = 0.548). (C) Per capita grazer – algal interaction strength decreased with disturbance under ambient nutrient conditions, and was low and unaffected by disturbance with enriched nutrients (nutrient effect: F1,12 = 7.11, P = 0.021; nutrient × disturbance interaction: F1,12 = 7.59, P = 0.018). More negative grazer-algal interaction strengths indicate greater effects of grazers on algae (inverted y-axis). Results were similar with interaction strength calculated using biomass of grazers (nutrient effect: F1,8 = 10.00, P = 0.013; disturbance × nutrient interaction: F1,8 = 14.57, P = 0.005; Appendix A: Tables A5 and A6).

Grazer abundance and composition

While we stocked grazers in the mesocosms according to natural densities and composition (with mayflies increasing and caddisflies decreasing with increasing disturbance; Fig. 1B), the total grazer density (mayflies + caddisflies) did not vary systematically along the disturbance gradient (Fig. 2B; Appendix A: Tables A1 and A5–A6). Therefore, effects of grazers on the accrual of algae could only be ascribed to species composition and behavior and not to total grazer abundance.

Grazer-algal interaction strength

The dynamic index of grazer-algal interaction strength standardized by grazer density (per capita) also varied with nutrients and along the gradient of disturbance legacy (Fig. 2C). Interaction strength was uniformly low in mesocosms with added nutrients, independent of disturbance (Fig. 2C). A significant disturbance × nutrient interaction indicated that grazer-algal interaction strength decreased with increasing legacy of disturbance, but only at ambient nutrient levels. Analysis of interaction strength per unit grazer biomass gave the same result as the per capita effect (Appendix A: Tables A5 and A6).

In-stream channel experiment testing effects of predators across a disturbance legacy gradient

Variation in rates of algal accrual and grazer abundance

As observed in the mesocosm experiment in treatments with ambient nutrient levels (Fig. 2A), there was a unimodal relationship between the rate of algal accrual and the legacy of disturbance, with no differences between the sides of channels with and without a brook trout (Fig. 3A). The absence of a fish effect in the in-stream channel experiment suggests that disturbance regime rather than presence of fish in the source water explains the unimodal relationship between disturbance and algal accrual in both experiments.

Figure 3.

Relationships between the multivariate index of disturbance and (A) the rate of algal accrual, (B) the density of total grazers, (C) the per capita impact of grazers on algal biomass (see Methods for formula), and (D) the propensity of grazers to drift out of the channels = grazer drift density divided by grazer density in the channels. Red and blue symbols indicate fish and fishless sides of the channels, respectively, and regression lines are for combined fish and fishless data when there was no fish effect. (A) Unimodal relationship between disturbance and algal accrual (quadratic R2 = 0.57, F2,9 = 4.699, P = 0.051; no fish effect: paired t = 0.898, P = 0.420, 4 df). (B) Grazer density decreased with disturbance (F1,7 = 15.39, P = 0.006; no fish effect: F1,7 = 0.50, P = 0.501; no fish × disturbance interaction: homogeneity of slopes test: F1,6 = 0.13, P = 0.732). (C) Per capita impact of grazers on algal biomass decreased with disturbance: lower numbers indicate greater impacts of grazers on algae (inverted y-axis) (F1,7 = 52.58, P < 0.001; fish effect: F1,7 = 7.38, P = 0.030; no fish × disturbance interaction F1,6 = 0.877, P = 0.385). (D) Drift propensity of grazers increased with disturbance (F1,7 = 14.04, P = 0.007, no fish effect F1,7 = 1.34, P = 0.284; no fish × disturbance interaction F1,6 = 1.382, P = 0.284).

Unlike ambient patterns (Figs. 1, 2B; Appendix A: Tables A1 and A5–A6), grazer densities (98% mayflies) in the baskets declined in streams with increasing disturbance (Fig. 3B); but there was no fish effect on grazer densities and no fish × disturbance interaction.

Variation in grazer impact on algae across treatments

The foraging performance of grazers, as indicated by their per capita impact on algae, declined with increasing disturbance (Fig. 3C), consistent with the results of the mesocosm experiment in treatments with ambient nutrients (Fig. 2C). While there was a significant effect of fish on grazer impact, grazers had a higher per capita effect on the fish side of channels at any given level of disturbance (Fig. 3C), which runs counter to expectations of a trophic cascade, whereby predators should suppress grazer foraging behavior. Furthermore, this fish effect is small in magnitude compared to the effect of disturbance (log ratio effect sizes from the ANCOVA: 0.89 for fish, and 1.60 for the difference between least and most disturbed streams). Also, the disturbance effect on grazer impact did not depend on fish, as indicated by a non-significant homogeneity of slopes test of the ANCOVA model.

Variation in grazer drift propensity across treatments

The propensity of grazers to drift out of channels increased with increasing legacy of disturbance (Fig. 3D), with no difference between the fish and fishless sides of the channels (mean ± SE = fish sides: 1.84 ± 0.48; fishless sides: 2.91 ± 1.11). Also, there were no relationships between the immigration ratio of grazers (drift in/drift out of channels) and the density of grazers (R2 = 0.134) or presence of fish (mean ± SE = fish sides: 1.33 ± 0.55; fishless sides: 1.25 ± 0.32). Therefore, grazer movement patterns into and out of channels did not explain variation in grazer density observed within the channels.


Variation in the legacy of disturbance among streams of this study resulted in a switch from grazer communities dominated by disturbance-intolerant (caddisflies) to disturbance-tolerant taxa (mayflies), and a unimodal relationship between disturbance legacy and rates of algal accrual in both experiments. It is clear that variation in algal accrual rates across the disturbance gradient of these streams is not a function of total grazer density, which did not vary with disturbance legacy, but instead, is a function of grazer species composition and traits (grazing performance and dispersal behavior). Results of the experiments suggest that better foraging performance (both experiments) and increased grazer residence time (i.e., lower propensity to move [channel experiment]) of the grazer assemblages may explain low rates of algal accrual in more benign streams. While it is tempting to argue that direct effects of disturbance suppress algal accrual in more disturbed streams (Peterson 1996, Biggs et al. 1999, Segura et al. 2011, Álvarez and Peckarsky 2013), direct scouring was not possible in the mesocosms. However, a unimodal relationship between the multivariate axis of disturbance and the benthic algal biomass in the 10 streams used for that experiment (Appendix A: Table A4) suggests that the source pool of algal colonists may be limiting in streams at both the low and high ends of the disturbance regime. Therefore, disturbance has the potential to affect grazer-algal interaction strength not only by altering grazer assemblage composition and grazer traits (foraging and dispersal), but also by directly modifying the biomass of algae (Roll et al. 2005).

Observed variation in grazer assemblage composition followed the predicted response to disturbance (Fig. 1), but under ambient nutrient conditions all grazers were less effective at suppressing algal accrual under increasing disturbance (Figs. 2C and 3C), regardless of their vulnerability to predators. Thus, the disturbance-predation trade-off we hypothesized (following Wootton et al. 1996) does not universally control the strength of top-down interactions. Other studies have reported negative effects of disturbance on grazer performance (Matthaei and Townsend 2000, Wellnitz and Poff 2006, Effenberger et al. 2011), but potential mechanisms are unresolved. Possibilities include reduction in grazing efficiencies, compromised ability to access rich algal patches, reduction by disturbance of more palatable forms of benthic algae (Álvarez and Peckarsky 2014), or difficulty foraging at high current velocities (Hintz and Wellnitz 2013). Furthermore, declining grazer performance in more disturbed streams could be explained by increasing drift propensity. If grazers are more transient (move more frequently) in disturbed streams, and individual residence times in algal patches decrease, then per capita impacts on algal accrual may also decrease as observed by Roll et al. (2005) in a similar food web of a Sierra Nevada stream in California. This phenomenon is similar to the observation that slowed movements by consumer fronts of mobile grazers enable them to achieve high local densities, consequently overwhelming their resources (Silliman et al. 2013).

Nutrient limitation was a necessary condition for the effects of increasing disturbance legacy and associated grazer assemblages on both the unimodal pattern of accrual of benthic algae and the declining grazer impacts on algae. When we added nutrients to mesocosms, increased algal growth potential limited the magnitude of herbivore effects, thereby weakening the strength of the grazer-algal interactions (as in Taylor et al. 2002, Peckarsky et al. 2013). The rates of accrual of benthic algae reflect counteracting dynamic processes including nutrient supply, physical disturbance and grazing (Peterson 1996, Biggs et al. 1999). Consequently, nutrient enrichment may change the palatability of algae favoring filamentous forms and thereby prevent grazers from depressing benthic algal biomass, overriding the influence of both predators and disturbance on grazer-algal interactions (Rosemond et al. 1993, Persson et al. 1996, Roll et al. 2005).

Trophic cascades were not observed in this study, consistent with all previous observations in these mountain streams (McIntosh et al. 2004, Peckarsky et al. 2013, Àlvarez and Peckarsky 2014), which originally inspired our quest to understand the context-dependency of trophic interactions in this system. Much research on trophic cascades has been undertaken in streams, motivated by the iconic study of Power (1990); and evidence from subsequent studies may be fruitful in the search for a more synthetic and comprehensive understanding of trophic control. For example, the strength of cascades may vary temporally in the same system as revealed by long-term observations showing stronger cascades in northern California (USA) rivers during years following high-disturbance hydrological regimes that eliminate predator-resistant grazers (Power et al. 2008). Also, the strength of trophic cascades varies spatially among streams depending on their physical attributes and the composition of their biota (Bowlby and Roff 1986, Power 1992c, Nyström and McIntosh 2003). Finally, bottom-up controls like nutrient availability can cause variation in the strength of trophic cascades among streams (Rosemond et al. 1993, Biggs et al. 2000, Riley et al. 2004, Peckarsky et al. 2013). Clearly, an expanded spatial and temporal context both within and between systems is needed to achieve a more general understanding of the controls governing food-web processes, such as trophic cascades (Power and Dietrich 2002, Woodward et al. 2008).

In our study, the most mobile and early-colonizing grazer species (Baetis mayflies) predominated in the baskets of the channel experiment testing for the effects of disturbance regime on species interactions among predators, grazers and benthic algae. Thus, experimental conditions may have altered the composition and densities of grazer assemblages relative to natural streams (Fig.1; Appendix A: Table A2). However, Baetis is also the mayfly most vulnerable to fish predation (Allan 1981) and most responsive to chemical cues from trout (McIntosh et al. 2002); therefore, its predominance should have increased the probability of observing an effect of fish in the channels. That no cascade was observed adds to accumulated evidence indicating that “biomass cascades,” whereby top predators change densities or behavior of grazers thereby affecting mean biomass of algae, do not occur in this mountain stream system. In contrast, previous studies have demonstrated the existence of “cryptic cascades” (Tessier and Woodruff 2002) in these streams, whereby top predators indirectly affect the distribution (heterogeneity: McIntosh et al. 2004, Álvarez and Peckarsky 2005) and composition (Álvarez and Peckarsky 2014) of benthic algae, mediated by their effects on grazer behavior.

In contrast to classic studies of trophic cascades in rivers of Northern California, failure to detect “biomass cascades” in this study depended neither on the hydrological regime (Power et al. 2008), nor on the predominance of predator-resistant vs. disturbance-resistant grazers (Wootton et al. 1996, Power and Dietrich 2002). We propose two hypotheses to explain those inconsistencies: (1) “Cryptic cascades” should predominate in systems where non-consumptive effects of predators have greater effects than consumptive effects, such that predators have stronger effects on prey behavior than on prey density (McPeek and Peckarsky 1998). Although strong behavioral cascades have been observed in other streams (McIntosh and Townsend 1996), terrestrial (Schmitz et al. 2004) and marine systems (Trussell et al. 2006, Kimbro et al. 2014), predator-mediated effects on prey foraging behavior are more likely to alter the distribution (heterogeneity) and species composition of basal resources than their mean biomass (Abrams 2000). Therefore, “cryptic cascades” may be more common in systems where trait-mediated indirect effects are subtle but, nonetheless, outweigh consumptive effects of predators (McPeek and Peckarsky 1998). (2) “Biomass cascades” should be rare in disturbance-prone and nutrient limited systems, where direct effects of disturbance on grazer foraging behavior and basal resources are stronger than predator-mediated indirect effects on lower trophic levels, and algal growth potential limits the size of herbivore effects (Feminella and Hawkins 1995). Furthermore, ecosystem complexity and habitat heterogeneity may dampen the strength of “biomass” trophic cascades in some cases (Power 1992c, Letourneau et al. 2004, Woodward et al. 2008).

In summary, variation in the legacy of disturbance among streams in one high-elevation catchment strongly affected not only the composition of primary consumers, but also their traits, such as dispersal behavior, grazing performance, and consequently the rates of accrual of benthic algae biomass. Furthermore, strong direct and indirect effects of the legacy of disturbance on rates of accrual of algae and top-down effects of grazers on algae were diluted by large nutrient subsidies. Moreover, the trade-off between vulnerability to disturbance and predation, while reflected in the composition of the grazer assemblages along the disturbance gradient, did not result in predicted variation in the strength of trophic cascades. Notably, the effects of variation in the legacy of disturbance on the accrual of benthic algae biomass mediated by grazer composition and behavior were stronger than the indirect top-down effects of predators.


We thank Marge Penton, Wendy Brown, Steve Horn, Sandye and Oakley Adams, Marita Davison, Helen Warburton, and Maria Alp for help in the field and lab, and Rick Lindroth for use of his leaf area meter. Steve Horn constructed and engineered the mesocosm installation and the in-stream fish channels, with help from Bryan Horn. Discussions with John Orrock, Pete McIntyre, Evan Childress, Mary Power, Claudio Gratton, Tony Ives, Emily Stanley, Steve Carpenter, LeRoy Poff, and Rex Lowe and comments by Amy Rosemond and two anonymous reviewers improved this paper. Funding was received from the National Science Foundation (DEB 0516035) to B. L. Peckarsky and A. R. McIntosh, Doctoral Dissertation Improvement Grant (DEB-0710031) to J. M. Moslemi, the University of Canterbury and the Royal Society of New Zealand Marsden Fund (UoC0801) to A. R. McIntosh, and the University of Vigo and the Spanish Ministry of Science and Innovation through the National Program for Fundamental Research (CGL 2009-07904) to M. Àlvarez.

Supplemental Material

Ecological Archives

Appendices A and B are available online: http://dx.doi.org/10.1890/ES15-00236.1.sm