Oecologia (2008) 157:641–651 DOI 10.1007/s00442-008-1097-8
C O M M U N I T Y E C O L O G Y - O RI G I N A L P A P E R
EVects of stream predator richness on the prey community and ecosystem attributes Erika Nilsson · Karin Olsson · Anders Persson · Per Nyström · Gustav Svensson · Ulf Nilsson
Received: 25 June 2007 / Accepted: 4 June 2008 / Published online: 3 July 2008 © Springer-Verlag 2008
Abstract It is important to understand the role that diVerent predators can have to be able to predict how changes in the predator assemblage may aVect the prey community and ecosystem attributes. We tested the eVects of diVerent stream predators on macroinvertebrates and ecosystem attributes, in terms of benthic algal biomass and accumulation of detritus, in artiWcial stream channels. Predator richness was manipulated from zero to three predators, using two Wsh and one crayWsh species, while density was kept equal (n = 6) in all treatments with predators. Predators diVered in their foraging strategies (benthic vs. drift feeding Wsh and omnivorous crayWsh) but had overlapping food preferences. We found eVects of both predator species richness and identity, but the direction of eVects diVered depending on the response variable. While there was no eVect on macroinvertebrate biomass, diversity of predatory macroinvertebrates decreased with increasing predator species richness, which suggests complementarity between predators for this functional feeding group. Moreover, the accumulation of detritus was aVected by both predator species richness and predator identity. Increasing predator species richness decreased detritus accumulation and presence of the benthic Wsh resulted in the lowest amounts of detritus. Predator identity (the benthic Wsh), but not predator species richness had a positive eVect on benthic algal biomass. Furthermore, the results indicate indirect negative eVects between the two ecosystem attributes, with
Communicated by Volkmar Wolters. E. Nilsson (&) · K. Olsson · A. Persson · P. Nyström · G. Svensson · U. Nilsson Department of Ecology/Limnology, Lund University, Ecology Building, 223 62 Lund, Sweden e-mail:
[email protected]
a negative correlation between the amount of detritus and algal biomass. Hence, interactions between diVerent predators directly aVected stream community structure, while predator identity had the strongest impact on ecosystem attributes. Keywords Prey diversity · Periphyton biomass · Detritus · Open system
Introduction The increased rate of species loss has fuelled studies on the eVects of biodiversity on the structure and function of ecosystems. Many studies have found a positive relationship between the diversity of a group of functionally related species and the services these species perform, such as the relationship between species richness of plants and primary production in grassland savannahs (Hector et al. 1999; Tilman et al. 2001), and between species richness of Wlter feeders and decomposers and resource use (Cardinale et al. 2002; Jonsson et al. 2002). Two main mechanisms have been attributed to this pattern of increasing function with richness: the selection or sampling eVect, where the probability of selecting a productive species increases with species diversity, i.e. species identity is important and not diversity per se, and the complementarity eVect, which can be in the form of either niche diversiWcation or facilitation between the species present, hence species numbers (diversity) are important (Huston 1997; Petchey 2000). These mechanisms can be active at the same time depending on the response variable under study (Cardinale et al. 2006). However, despite the fact that distinguishing between them is essential for gaining a mechanistic understanding of the consequences of species loss on various services provided
123
642
by ecosystems, relatively few studies have tried to separate the eVects of identity from the eVects of diversity. The positive relationship between diversity and function found for e.g. plants does not have its correspondence in predator diversity eVects on prey diversity and ecosystem attributes (but see Schmitz 2007). Predator density is known to aVect both density and composition of prey populations (Kerfoot and Sih 1987) directly through consumption and indirectly through trait-mediated interactions (Kotler and Holt 1989; Werner and Peacor 2003), which ultimately may aVect ecosystems through trophic cascades (DuVy et al. 2007). Predators may even alter the physical habitat altogether, thereby promoting ecosystem processes (Jones et al. 1997). However, there is no clear pattern of how predator diversity aVects systems with several trophic levels (DuVy et al. 2007). Interactions between predators can change the outcome of predator–prey interactions (Sih et al. 1998) and potentially also any eVects on ecosystem attributes (Ives et al. 2005; DuVy et al. 2007) in three principle ways. First, predatory species that have diVerent foraging strategies may aVect diVerent parts of the food web (Chalcraft and Resetarits 2003). Hence, predators with diVerent foraging strategies can, in combination, have emergent eVects on lower trophic levels and reduce or enhance prey predation risk (Sih et al. 1998; Eklöv and VanKooten 2001; Nyström et al. 2001). Indeed, increasing predator diversity has been shown to alter trophic relationships and induce trophic cascades both in terrestrial and marine systems (Finke and Denno 2004; Bruno and O’Connor 2005; Byrnes et al. 2006). Second, some studies suggest that predators are substitutable and have the same eVects individually as in combinations (Nyström and McIntosh 2003; Vance-Chalcraft et al. 2004). Third, both interference interactions between predators and omnivory may in fact reduce risk of predation for the prey species (Finke and Denno 2004, 2005; Nilsson et al. 2006), which may reduce the strength of trophic cascades. Omnivores can alter energy routes in the food web and thus aVect other predatory species and trophic structure (Usio and Townsend 2002). By categorising predator hunting mode, predator habitat domain and prey habitat domain, Schmitz (2007) found that it is the extent of predator similarity together with a narrow or broad prey habitat domain that determines the outcome of interactions and thus, potential eVects on ecosystem attributes. This study aims at furthering our knowledge of the importance of predators and predator diversity in stream food webs. Using all possible combinations of three dominant and naturally coexisting stream predators, we tested for the eVects of predator species richness and predator species identity on: (1) macroinvertebrate biomass and diversity; and (2) two ecosystem attributes, benthic algal biomass and detritus accumulation. In our study system the
123
Oecologia (2008) 157:641–651
three most common and dominant predators are two species of Wsh (stone loach, Barbatula barbatula, and brown trout, Salmo trutta) and omnivorous crayWsh (signal crayWsh, Pacifastacus leniusculus) (Eklöv et al. 1998; Degerman et al. 2007). This combination of a salmonid, a benthic Wsh and an omnivorous crayWsh is a common stream predator assemblage in many parts of the world (Prenda et al. 1997; Carter et al. 2004; Elliott 2006; Olsson et al. 2006; Degerman et al. 2007). The three predators overlap considerably in diet, but have diVerent hunting modes and habitat domain. Hence, we expected them to generate diVerent direct and indirect eVects of predation, diVerent rates of bioturbation and, hence, detritus accumulation. Moreover, we were interested in the eVects of both species identity and richness in order to evaluate if individual predators have similar eVects, i.e. are they redundant, or if eVects from diVerent predators are complementary.
Materials and methods The experiments were conducted in June and July 2005, in an open system with artiWcial stream channels, where prey could continuously colonise in order to incorporate the processes of emigration and immigration that are known to be important in stream ecosystems (Allan 1983; Wooster and Sih 1995). Predator richness was manipulated to include all possible combinations to control for a potential selection eVect (Huston 1997; Giller et al. 2004). Mesocosm experiments are an important tool in understanding ecological processes and have been widely used in both predator–prey and food web dynamics as well as for studying competition and predation (Benton et al. 2007 and references therein). The small scale of these experiments should be taken into consideration since this can have confounding eVects (see e.g. Englund and Olsson 1996; Englund 1997, 2005), therefore results from these types of studies need to be interpreted and extrapolated with care (Carpenter 1996). However, results from previous studies using stream mesocosms have corroborated with results from Weld studies in natural streams (see e.g. McIntosh and Townsend 1996; Nyström and McIntosh 2003). Furthermore, we used a set up with a Xow-through system that mimics the dynamics of a natural stream, limiting the potential biases that an enclosed system contributes to, e.g. low oxygen levels, accumulation of waste products and overestimated predation rates due to a homogeneous habitat. Natural history The Weld mesocosm experiments were conducted next to an old mill channel, in the River Bråån, a stream with a total
Oecologia (2008) 157:641–651
drainage area of 165 km2, ranging 120 km through mostly agricultural landscape. We focused our study on three characteristic predators in this type of stream in southern Sweden; stone loach (Barbatula barbatula L.), brown trout (Salmo trutta L.) and signal crayWsh (Pacifastacus leniusculus Dana) that all live in habitats similar to that created in the experimental stream channels (Eklöv et al. 1999; Englund and Krupa 2000; Nilsson and Persson 2005). These three predator types dominate the predator assemblage in streams in southern Sweden (Degerman et al. 2007). They have diVerent habitat domain (Schmitz 2007), either associated with the bottom substrate or the water column, and hunting mode (Schmitz 2007)—active or sit and pursue (Maitland 1965; Nyström 1999)—but they have overlapping prey preferences (Maitland 1965; E. Nilsson, unpublished data). Thus, they have partly overlapping niches with respect to prey but diVer in the way that they hunt and capture them, which suggests that the diVerent top predators could have either complementary or substitutable eVects on the prey community depending on predator combinations (Dahl 1998; Stenroth and Nyström 2003; Schmitz 2007) and on ecosystem attributes (Chalcraft and Resetarits 2003; DuVy et al. 2007). The stone loach is a nocturnal forager (Fischer 2004) that searches the bottom substrate for macroinvertebrates (e.g. chironomids), thereby disturbing the bottom surface. Brown trout feeds on drifting invertebrates of both terrestrial and aquatic origin, but are visually oriented foragers that cause little disturbance to the sediment. Adult trout (>15 cm) also feed on Wsh and crayWsh. The signal crayWsh is omnivorous and feeds on algae, detritus and macroinvertebrates. Its foraging physically disturbs the substrate and it has been recognised as an ecosystem engineer (Usio and Townsend 2002). Adult crayWsh can occasionally also capture and feed on Wsh (Guan and Wiles 1997). Experimental design We used 16 artiWcial stream channels (PVC; length 2.0 m, width 0.4 m, depth 0.2 m) as experimental units, placed in two rows of eight, all with their own water inlet and outlet (diameter 50 mm). To minimise the inXuence of direct terrestrial input all channels were covered with a mesh lid (5 mm mesh size) to prevent arthropods and leaves from subsidising the stream ecosystem (Nakano et al. 1999; Kawaguchi et al. 2003) and to prevent predators from escaping. Water was moved from the mill channel using a siphon and by taking advantage of a higher water level in the mill channel. The discharge into each stream channel was 0.5 l/ s. All hose inlets were mounted with mesh baskets (mesh size 10 mm) to prevent them from clogging and avoid Wsh and debris from entering the stream channels. Channel outlets had a 4-mm mesh to prevent predators from escaping
643
but allowed macroinvertebrates to exit. Each channel was Wlled with 40 l of streambed gravel and four larger cobbles. We also added two glazed ceramic tiles (81 cm2 each) to each end of the channel in order to measure the response of periphyton and accumulation of detritus. To avoid intraguild predation we used 0+ brown trout (51 mm § 4.0) [total length (TL) § 1 SD] and 1+ signal crayWsh (44.4 mm § 4.2), while the stone loach were on their second or third year (72.8 mm § 5.0). One day prior to the start of each of the experimental trials Wsh were captured using electroWshing in the River Bråån, where signal crayWsh are also present. Since juvenile crayWsh are diYcult to catch, we obtained juveniles from the Simontorp pond, a commercial hatchery pond (see Stenroth et al. 2006 for further details). The experiment was designed as a randomised block with four levels of predator species richness (zero to three predators in diVerent combinations). Due to logistic reasons it was replicated twice over time for a total of four replicates per treatment. Treatments with predators had equal densities of six predators in both single and multiple combinations, which is within the natural range in Swedish streams (Swedish Electro Fishing Register, www.fiskeriverket.se). On the start dates we randomly assigned the eight diVerent treatments to the stream channels (eight treatments £ two replicates/time block). Experimental procedure The time blocks spanned 26 days (9 June–5 July, 7 July–2 August). A week before the Wrst trial period water circulation was started to initiate invertebrate colonisation. One day before each trial period we added two standardised kick samples (2 £ 30 s) of macroinvertebrates to all stream channels in order to ensure prey presence at the start of the experiment. Before all predators were added to the channels they were measured (nearest mm, TL) and weighed (nearest 0.1 g). The channels were checked daily for dead predators and Xooding due to clogged nets. We replaced dead predators in order to keep the same predator density throughout the experiment. During the Wrst trial period no predators were replaced but during the second trial period four stone loach and seven signal crayWsh were replaced. However, since we were unable to detect that two crayWsh were missing from one replicate with brown trout + crayWsh, in the Wrst trial period, this replicate was omitted from further analysis. To quantify predator eVects on invertebrate activity we used active-fauna traps that are modiWed plastic beakers with a mesh (1.0 mm) funnel entrance and a removable mesh lid (1.0 mm) at the other end (see Nyström et al. 2001 for further details). Three traps were placed in each stream channel for both day- (1000–1700 hours) and nighttime activity (2200–0700 hours) once for each trial period before
123
644
ending the experiments. At the end of each trial period we sampled the macroinvertebrate fauna by manually collecting all macroinvertebrates from 15 l of substrate from each channel. All invertebrates were preserved in 70% ethanol. Macroinvertebrates were examined under a binocular microscope in the laboratory, counted and determined to at least family level, and assigned to a functional feeding group (FFG); Wltering collector, gathering collector, predator, scraper and shredder according to Nilsson (1996, 1997). The macroinvertebrates were later dried at 60°C for 24 h and then combusted in a muZe furnace at 460°C for 1 h to enable calculation of ash-free dry weight (AFDW) from each replicate channel. However, shredders and grazers were only present in very few stream channels making it impossible to calculate diversity indexes for them because there were so few species, which is why they were omitted from further analysis. Detritus that had gathered on the tiles was collected from the tiles in a 250-m mesh hand-held net, after which the tiles were put into plastic bags and stored in a freezer together with the detritus. In the laboratory, detritus was dried at 60°C for at least 4 h and combusted in a muZe furnace at 460°C for at least 4 h to enable calculation of AFDW (g), while periphyton was easily removed from the glazed ceramic tiles with paper towels, extracted in ethanol for 12 h in darkness at 20°C and analysed for chlorophyll a (g/cm2) to estimate biomass using a spectrophotometer (for more detailed methods see Jespersen and ChristoVersen 1987). We measured water temperature every 4 h (6times a day) to include both day- and nighttime temperatures, with ten temperature data loggers (HOBO Water Temp Pro loggers) randomly dispersed between channels, which averaged (mean § 1 SD) 18.2 § 2°C (range 14.1–21.4°C) for trial period 1, and 20.5 § 2°C (range 17.7–23.9°C) for trial period 2. Oxygen concentrations (8.25 § 0.1 mg/l) and pH (7.0 § 0.01) varied little between channels. Data analyses To examine the eVects of predator species richness, predator identity and trial period we used two nested multivariate ANOVAs (MANOVAs) for the diVerent categories of response variables, where predator identity was nested within predator species richness. For invertebrates we examined eVects on biomass, diversity and activity, while standing stocks of detritus and algal biomass (chlorophyll a) were used as variables to test diVerences in ecosystem attributes. All possible interactions with trial period were tested in initial models and non-signiWcant interactions were omitted and the analysis rerun, but data from the diVerent trial periods were never pooled. The nested MANOVAs present general trends in the data, but we conducted individual nested ANOVAs for the macroinverte-
123
Oecologia (2008) 157:641–651
brate and ecosystem variables to examine eVects on each of the variables. SigniWcant eVects of predator species richness were followed by Tukey’s post hoc tests to reveal among which levels of predator species richness there were any diVerences. Trial period was signiWcant for some of the tested invertebrate variables, i.e. seasonality in macroinvertebrate composition, where species emerge at diVerent times during summer. Thus, diversity and biomass were higher during the Wrst trial period but diVerences between trial periods were not what we intended to test and will not be discussed further. Furthermore, we used a Pearson correlation on the ecosystem attributes to examine any correlation between periphyton biomass and detritus accumulation. The Kolmogorov–Smirnov test of normality with Lilliefors’ signiWcance correction was used to check if data were normally distributed. If not, it was log10 (x + 1) transformed. All statistical analyses were conducted in SPSS 15.0 for Windows XP, where the Satterthwaite correction factor is applied by default to deal with unbalanced MANOVA designs. We used the reciprocal form of the Simpson diversity index to calculate invertebrate diversity, both total diversity and diversity within FFG. The reciprocal form of the diversity index was used in order to receive high index numbers that correspond with high diversity. It is one of the most commonly used diversity indexes available and it is useful when samples are dominated by one or a few species (Magurran 1988).
Results EVects on the prey assemblage The nested MANOVA on invertebrate variables (diversity, biomass and activity) revealed signiWcant eVects of predator species richness and time, but not predator identity [nested MANOVA: time, F11,9 = 13.227, P < 0.001; predator species richness, F32.8,27 = 2.372, P = 0.010, predator identity (predator species richness), F43.0,36 = 0.999, P = 0.501]. The interaction between time and predator species richness was non-signiWcant (F32.8,27 = 1.740, P = 0.066). The dominating group of invertebrates by weight was the Wltering collectors, which contributed to more than 50% of the AFDW in all treatments, followed by gathering collectors and predators (Fig. 1). There was, however, no eVect of either predator richness treatment on biomass of any of the FFGs (Fig. 1) (Table 1). The Simpson reciprocal diversity index over all macroinvertebrates ranged between 1 and 2.5, which indicates quite low diversity caused by dominance of a few abundant taxa. Diversity in the FFGs was highest for the predatory invertebrates (range 1–7) and lowest for the Wltering collectors
Oecologia (2008) 157:641–651
0.60 FFG biomass (g) ± 1 S.E.
Fig. 1 Mean invertebrate biomass (§1 SE) of the Wve functional feeding groups (FFG) for the diVerent levels of predator species richness. There were neither any diVerences in biomass between treatments nor any eVect of predator identity. However, there was an eVect of trial period, with higher biomass for the second trial period
645 Filtering collector Gathering collector Predator
Trial 2
Trial 1
0.50 0.40 0.30 0.20 0.10 0 0
1
(range 1–2.5) that was dominated in numbers by Chironomidae (Fig. 2). The nested ANOVA on total invertebrate diversity was signiWcant for eVects of predator species richness and time (Table 1). The diversity in the treatment with one predator was lower than for the no-predator treatment (post hoc: P = 0.029). For the FFG there were signiWcant diVerences in between levels of predator species richness for diversity of the predatory invertebrates (Fig. 2) (Table 1), where the treatment with all three predator species had a lower diversity index than treatments with zero (Tukey’s post hoc: P = 0.042) and one predator species (Tukey’s post hoc: P = 0.050). However, there were no eVects of predator species richness on diversity of the Wltering and gathering collectors (Fig. 2) (Table 1). Activity of benthic macroinvertebrates was assessed in traps (all FFGs pooled, but dominated by gathering collectors) and invertebrates were less active during the day (nested ANOVA: daytime activity F3,19 = 11.638; P < 0.001) in all treatments with predators compared to the control (Tukey’s post hoc test: P1pred = 0.001, P2pred < 0.001, P3pred = 0.001), while there was no trend in activity during the night (nested ANOVA, nighttime activity F3,19 = 0.727; P = 0.548).
2 3 0 Predator species richness
1
2
3
aVected by predator species richness and predator identity [nested MANOVA: time, F2,21 = 0.530, P = 0.596; predator species richness, F6,42 = 6.080, P < 0.001; predator identity (predator species richness), F8,42 = 2.156, P = 0.051] (Fig. 4a, b). The standing stock of detritus was signiWcantly inXuenced by predator species richness (nested ANOVA: F3,22 = 17.169; P < 0.001), where all treatments with predators had lower amounts of detritus than the control (post hoc: P1pred < 0.001, P2pred < 0.001, P3pred < 0.001). EVects of predator identity were also signiWcant (nested ANOVA: F4,22 = 2.924; P = 0.044). Thus, presence of predators aVected ecosystem attributes and within the treatments with one and two predators their identity aVected the amount of detritus in the experimental stream channels (Fig. 3a). The biomass of periphyton (chlorophyll a g/cm2) was not aVected by predator richness (nested ANOVA: F3,22 = 2.201; P = 0.117), but there was a tendency that predator identity within richness levels aVected the biomass of periphyton (nested ANOVA: F4,22 = 2.589; P = 0.065) (Fig. 3b). Furthermore, there was a negative correlation between periphyton biomass and the amount of detritus (Pearson correlation: r = ¡0.524, P = 0.002, n = 31) (Fig. 4).
Detritus and periphyton biomass Discussion In the treatment with no predators present (control), at least 10 times more detritus had accumulated on the ceramic tiles than in treatments with one to three predator species (Fig. 3a). There was high variation in the biomass of periphyton on ceramic tiles (Fig. 3b), but treatments that had high amounts of detritus, the control and crayWsh treatment, had low periphyton biomass (Fig. 3a, b), while the other treatments had twice as much algal biomass as the control (Fig. 3b). Detritus and periphyton biomass were
The results indicate that stream predators can have both complementary and individual eVects depending on the variable studied. Increased top predator species richness resulted in lower diversity of predatory prey, indicating complementarity, while there was no eVect of either predator species richness or predator identity on prey biomass. There were general eVects of predator species richness on ecosystem attributes, but there were no diVerences between
123
646 Table 1 Results from individual nested ANOVAs for the diVerent invertebrate variables. FC Filtering collector, GC gathering collector, P predator
Oecologia (2008) 157:641–651
18.620
0.007
1.596
FC diversity
7.422
1
7.422
3.214
0.089
GC diversity
7.216
1
7.216
18.888