INTRODUCTION
Insectivores, particularly birds and bats, in natural (i.e. non-agricultural) tropical forests initiate top-down density- and trait-mediated trophic cascades that protect plants by reducing leaf damage from herbivorous arthropods (Dyer et al. Reference DYER, CARSON, LEIGH, Barbosa, Letourneau and Agrawal2012, Kalka et al. Reference KALKA, SMITH and KALKO2008, Mäntylä et al. Reference MÄNTYLÄ, KLEMOLA and LAAKSONEN2011, Mooney et al. Reference MOONEY, GRUNER, BARBER, VAN BAEL, PHILPOTT and GREENBERG2010, Schmitz et al. Reference SCHMITZ, KRIVAN and OVADIA2004). Trophic cascade strength varies considerably due to varying biotic and abiotic contexts (Agrawal et al. Reference AGRAWAL, ACKERLY, ADLER, ARNOLD, CÁCERES, DOAK, POST, HUDSON, MARON, MOONEY, POWER, SCHEMSKE, STACHOWICZ, STRAUSS, TURNER and WERNER2007). Most studies of context-dependence to date have focused on direct trophic linkages to one or more participants in the trophic cascade, for example, identity of predators (e.g. bats or birds); intraguild predation; and plant productivity (Borer et al. Reference BORER, SEABLOOM, SHURNIN, ANDERSON, BLANCHETTER and HALPERN2005, Kalka et al. Reference KALKA, SMITH and KALKO2008, Mooney et al. Reference MOONEY, GRUNER, BARBER, VAN BAEL, PHILPOTT and GREENBERG2010, Van Bael & Brawn Reference VAN BAEL and BRAWN2005, Vance-Chalcraft et al. Reference VANCE-CHALCRAFT, ROSENHEIM, VONESH, OSENBERG and SIH2007).
However, trophic cascades occur within larger biological communities, and participants interact with many other organisms not typically considered in research on trophic cascades (Kéfi et al. Reference KÉFI, BERLOW, WIETERS, NAVARRETE, PETCHEY, WOOD, BOIT, JOPPA, LAFFERTY, WILLIAMS, MARTINEZ, MENGE, BLANCHETTE, ILES and BROSE2012). For example, ungulates such as pigs (Suidae) and peccaries (Tayassuidae) may mediate trophic dynamics through disturbance to forest understorey (Queenborough et al. Reference QUEENBOROUGH, METZ, WIEGAND and VALENCIA2012). While these omnivores do consume some plant material, their greatest effects on plants are non-trophic, via disturbances such as trampling and rooting that alter microhabitats and microclimates and, potentially, arthropod predator abundance (Beck Reference BECK, Forget, Lambert, Hulme and Vander Wall2005, Ickes et al. Reference ICKES, DEWALT and APPANAH2001, Queenborough et al. Reference QUEENBOROUGH, METZ, WIEGAND and VALENCIA2012). Consequently, ungulates may reduce the abundance or foraging intensity of arthropod herbivores and leaf damage (Côté et al. Reference CÔTÉ, ROONEY, TREMBLAY, DUSSAULT and WALLER2004, Stewart Reference STEWART2001). Ungulates may also increase abundance of other vertebrate insectivores through their effects on litter depth, complexity and nutrient cycling. For example, peccaries increased the abundance of leaf-litter herpetofauna relative to experimental peccary exclosures at La Selva Biological Station, Costa Rica (Reider et al. Reference REIDER, CARSON and DONNELLY2013). Consequently, ungulates could weaken, negate or even reverse the direction of the trophic cascade through primarily non-trophic interactions. Thus, weaker trophic cascades in wet tropical forest understorey may be due to abundant ungulates rather than, or in addition to, lower productivity or a more diverse producer trophic level (Boege & Marquis Reference BOEGE and MARQUIS2006, Van Bael & Brawn Reference VAN BAEL and BRAWN2005, Van Bael et al. Reference VAN BAEL, BRAWN and ROBINSON2003).
The wild pig (Sus scrofa) and peccaries, including the omnivorous collared peccary (Pecari tajacu), have increased in abundance in some fragmented tropical forests, due at least in part to apex predator loss and increased food availability from nearby farms (Beck Reference BECK, Forget, Lambert, Hulme and Vander Wall2005, Ickes Reference ICKES2001, Michel & Sherry Reference MICHEL, SHERRY, Sudarshana, Nageswara-Rao and Sonegi2012, Terborgh & Estes Reference TERBORGH and ESTES2010, Timm Reference TIMM, McDade, Bawa, Hespenheide and Hartshorn1994). Here, we evaluate the non-trophic effects of the collared peccary on an insectivorous bird/bat–arthropod–understorey plant trophic cascade in natural (i.e. non-agricultural) tropical forest. We tested two interdependent hypotheses. First, insectivorous birds and bats cause trophic cascades by reducing the abundance or changing the behaviour of insect herbivores, thereby indirectly causing a decrease in plant damage. Second, the collared peccary weakens, negates, or reverses the direction of these cascades. We experimentally tested these two hypotheses by excluding insectivorous birds and bats from a diverse group of understorey plant species both within and outside of large peccary exclosures.
METHODS
Study site
We conducted this research at La Selva Biological Station, Costa Rica (10°26ʹN, 83°59ʹW, 35–130 m asl), a 1611-ha lowland rain-forest reserve receiving 4 m of rain annually, with a February–April dry season (Sanford et al. Reference SANFORD, PAABY, LUVALL, PHILLIPS, McDade, Bawa, Hespenheide and Hartshorn1994). The collared peccary is highly abundant at La Selva (Kuprewicz Reference KUPREWICZ2013), and most other large, terrestrial vertebrates are extirpated or rare (e.g. white-lipped peccary (Tayassu pecari), Baird's tapir (Tapirus bairdii); Timm Reference TIMM, McDade, Bawa, Hespenheide and Hartshorn1994).
Design of mammal and bird/bat exclosures
We established five paired 20 × 50-m fenced exclosure and control plots ( = five blocks) within old-growth forest in 1999. These exclosures allow entry to small and mid-sized mammals (e.g. Central American agouti, Dasyprocta punctata) while only excluding peccaries (Reider et al. Reference REIDER, CARSON and DONNELLY2013); hereafter we refer to these as peccary exclosures. We subdivided each peccary exclosure or control plot into 40, 5 m × 5-m subplots, following a split-split plot randomized complete block design. In June 2007, three of these subplots were randomly assigned to contain net-covered exclosures excluding birds and bats, and three as controls (N = 60 total). We constructed 8-m3 (2 × 2 × 2 m) bird/bat exclosures using plastic bird-control netting (3 × 3.5 cm mesh) stretched across four PVC poles tied back to nearby trees and staked to the ground (following Boege & Marquis Reference BOEGE and MARQUIS2006). These 8-m3 bird/bat exclosures are hereafter referred to as netted exclosures. This mesh size excludes all understorey insectivorous birds (except perhaps Myiornis atricapillus, a small canopy-subcanopy forager; T. Sherry unpubl. data), bats, and other mid-size and large insectivores (e.g. mouse opossums, Marmosa spp.), while allowing entry by small insectivorous lizards, frogs and all but the largest arthropods (Boege & Marquis Reference BOEGE and MARQUIS2006, Van Bael & Brawn Reference VAN BAEL and BRAWN2005). While the netting could provide additional substrate for orb-weaving spiders, we observed few orb-weavers using the netting, and removed those we did find. We did not use procedural controls (open-sided exclosures to control for net effects) since prior studies found that the chosen mesh type and size reduces light by <1% and has no impact on microclimatic variables such as temperature or rainfall (Marquis & Whelan Reference MARQUIS and WHELAN1994, Van Bael et al. Reference VAN BAEL, BRAWN and ROBINSON2003). All leaf litter accumulated on top of the netted exclosures was removed and placed inside the exclosures, and any tears in the netting were repaired at each visit. Accumulated leaf litter could reduce light and, consequently, photosynthetic rate between monthly removals, and netting tears (though rare) could allow birds and bats temporary access, thus weakening trophic cascade strength. However, because the netting effects should be consistent across peccary exclosure treatments, litter accumulation and temporary gaps in the netting should not influence the key question of how peccaries affect trophic cascades.
Each netted exclosure or control was centred on a single focal plant 0.5–2 m tall and possessing at least 20 leaves. This size range was determined by the size of the exclosures (2 × 2 × 2 m), and the need for a minimum of 20 leaves for leaf area and damage assessments. We included a total of 14 species from 12 families and five clades (Angiosperm Phylogeny Group III 2009; Appendix 1). Our peccary exclosure and control locations, coupled with our randomized sampling design in tropical rain-forest understorey, necessitated using more diverse shrub species than previous studies (Mäntylä et al. Reference MÄNTYLÄ, KLEMOLA and LAAKSONEN2011), but had the advantage of allowing us to evaluate whether trophic cascades occurred across phylogenetically diverse taxa. Focal plants were chosen using a stratified random sampling design; within each randomly selected 5 × 5-m subplot we chose a plant from one of the focal species selected for their relative frequency and appropriate size.
Arthropod surveys
We visually surveyed all leaves and stems of each focal plant for arthropods immediately prior to constructing netted exclosures (June 2007) and again in July 2007, October 2007 and April 2008. All arthropods were identified to order and family (where possible, e.g. Coleoptera, Hemiptera), and assigned to one of four foraging guilds (predator, phloem-feeding herbivore, chewing herbivore and omnivore; Triplehorn & Johnson Reference TRIPLEHORN and JOHNSON2004). We calculated arthropod densities as number of individuals per unit leaf area for each survey. We studied arthropod predators because of their potentially important role in the trophic cascade as intraguild predators (Mooney et al. Reference MOONEY, GRUNER, BARBER, VAN BAEL, PHILPOTT and GREENBERG2010, Vance-Chalcraft et al. Reference VANCE-CHALCRAFT, ROSENHEIM, VONESH, OSENBERG and SIH2007).
Leaf-damage and area assessments
We assessed plant damage by determining missing leaf area using a 0.25-mm2 transparent plastic grid, and measuring leaf length and width on 20 randomly selected leaves per focal plant, including both young and old leaves (Boege & Marquis Reference BOEGE and MARQUIS2006). Original, undamaged leaf area was calculated using leaf-area regression equations derived from leaf width and length for 100 undamaged leaves per species. We then estimated damage as per cent area missing using the ratio of total missing area to original, undamaged leaf area. We calculated total leaf area per plant by multiplying mean extant leaf area times the number of leaves on the plant. Total leaf area and per cent damage were calculated in June and October 2007 and April 2008. Finally, we visually surveyed each leaf for signs of mammalian browsing, particularly whether whole or partial leaves had been eaten or removed.
Data analysis
We separately analysed densities of all arthropods, predatory arthropods, chewing herbivores, phloem-feeding herbivores, and the six most common arthropod orders (Hemiptera, Araneae, Hymenoptera, Diptera, Lepidoptera and Orthoptera), as well as per cent leaf damage, using repeated-measures generalized linear mixed models (Proc GLIMMIX, SAS 9.3, SAS Institute, Cary, NC, USA). Arthropod density and leaf damage estimates were averaged across the three replicate netted exclosure or control subplots within each peccary exclosure and control plot, resulting in N = 20. Replicates were averaged to avoid pseudoreplication, but as replicates included multiple plant species, this prevented the inclusion of plant as a random effect in the analysis. Peccary exclosure treatment (α), bird/bat exclosure treatment (β), time (τ) and all two- and three-way interactions were fixed factors. Random effects included an intercept for each block (δ) and slopes for all two-, three- and four-way interactions between block, peccary exclosure, bird/bat exclosure and time. GLIMMIX calculates a single, composite error term (εijklmn) incorporating all whole plot and subplot error using random effects. Denominator degrees of freedom were calculated using the between-within method, which divides residual degrees of freedom into between-subject and within-subject components (SAS Institute 2009). The complete model was
\begin{eqnarray*}
\begin{array}{l}
y = a + \alpha _i + \beta _j + (\alpha \beta )_{ij} + \tau _k + (\alpha \tau )_{ik} + (\beta \tau )_{jk} \\
\qquad +\, (\alpha \beta \tau )_{ijk} + \delta _l + (a\delta )_{il} + (\beta \delta )_{jl} + (a\beta \delta )_{ijl} \\
\qquad +\, (a\tau \delta )_{jkl} + (\beta \tau \delta )_{ijk} + (\alpha \beta \tau \delta )_{ijkl} + \varepsilon _{ijklmn} , \\
\end{array}
\end{eqnarray*}where a represents the fixed-effect intercept.
We included initial density and initial leaf area as covariates for all arthropod density analyses. All two- and three-way interactions between covariates and bird/bat and peccary exclosure treatments were non-significant, and were removed from subsequent analyses. For the leaf damage analysis we included initial leaf area and initial leaf damage as covariates; mean predator and chewing herbivore densities were also included as covariates to test whether bird/bat- or peccary-exclosure effects persisted after controlling for arthropod densities (Shipley Reference SHIPLEY2009). Pseudo-information criteria (Pseudo-AICc and Pseudo-BIC) were used to select the optimal distribution and link (Dick Reference DICK2004). Significant bird/bat × peccary treatment interactions were identified using two-tailed tests of least squares means with Tukey–Kramer HSD adjustments for multiple comparisons (Littell et al. Reference LITTELL, MILLIKEN, STROUP, WOLFINGER and SCHABENBERGER2006).
RESULTS
Peccary exclusion effects
Peccary exclusion effects on all arthropods varied over time (peccary × time interaction), significantly increasing total arthropod density only in the first month post-exclosure (July 2007; Table 1, Figure 1a). Peccary exclusion effects on predators and spiders (Araneae) also varied over time (peccary × time interactions; Tables 1, 2). Peccary exclusion increased densities of arthropod predators – 94% of which were spiders – twofold in July and by half in October, but reduced arthropod predator densities by 42% in April (Figure 1b). Peccary exclusion increased mean Hymenoptera densities by 73% on plants both with and without birds and bats (mean ± SE, peccary exclosure: 3.7 ± 0.6 m−2, peccary control: 2.1 ± 0.3 m−2; Table 2). Peccary exclusion reduced Diptera densities by 31% on plants with birds and bats, but did not affect Diptera densities on plants without birds and bats (bird/bat × peccary interaction; Table 2, Figure 2). Peccary exclusion reduced per cent leaf damage by 27% on plants with birds and bats, but increased leaf damage by 57% on plants without birds and bats (bird/bat × peccary interaction; Table 1, Figure 3).
Table 1. Results of repeated measures generalized linear mixed models used to test responses of total arthropod density, predator density, phloem-feeding and chewing herbivore density, and leaf damage (per cent leaf area missing) in a factorial experiment with two levels of bird/bat treatment (exclosure and control), two levels of peccary treatment (exclosure and control), and four sampling periods over a 10-mo period (June, July, and October 2007 and April 2008) in tropical rain-forest understorey at La Selva Biological Station, Costa Rica. * P < 0.05; ** P < 0.01; *** P < 0.001; **** P ≤ 0.0001.
Table 2. Results of repeated measures generalized linear mixed models used to test responses of densities of the most common arthropod orders (Araneae, Diptera, Hemiptera, Hymenoptera, Lepidoptera and Orthoptera) in a factorial experiment with two levels of bird/bat treatment (exclosure and control), two levels of peccary treatment (exclosure and control), and four sampling periods over a 10-mo period (June, July and October 2007 and April 2008) in rain-forest understorey at La Selva Biological Station, Costa Rica. * P < 0.05; ** P < 0.01; *** P < 0.001; **** P ≤ 0.0001.
Figure 1. Effects of the significant two-way interaction between peccary exclosure and time on total arthropod density (a) and arthropod predator density (b) (mean ± SE) in June, July and October 2007 and April 2008 at La Selva Biological Station, Costa Rica. Significant treatment effects, adjusted for four monthly comparisons, are indicated as follows: * P < 0.05; *** P < 0.001.
Figure 2. Effects of the significant two-way interaction between bird/bat exclosure and peccary exclosure on Diptera density (mean ± SE) during June 2007–April 2008 at La Selva Biological Station, Costa Rica. Significant effects are indicated as follows: * P < 0.05.
Figure 3. Effects of bird/bat exclosure, peccary exclosure and their two-way interaction on per cent leaf area missing (mean ± SE) in June and October 2007 and April 2008 at La Selva Biological Station, Costa Rica (leaf damage was not assessed in July 2007). Effects of the bird/bat exclosure × peccary exclosure interaction on leaf damage averaged over the course of the experiment (a); and effects of bird/bat exclosure and peccary exclosure on leaf damage over time (b). Significant treatment effects are indicated as follows: * P < 0.05; ** P < 0.01; *** P < 0.001.
Bird and bat exclusion effects
Excluding birds and bats, with or without peccaries, increased total arthropod density by 52% (mean ± SE, bird/bat exclosure: 31.9% ± 2.5%, bird/bat control: 20.9% ± 1.6%; Table 1, Figure 4). Bird and bat exclusion effects on phloem-feeding herbivore and Hemiptera densities varied over time, and were unaffected by peccary presence or absence (bird/bat × time interaction; Tables 1, 2). Bird and bat exclusion doubled phloem-feeding herbivore density – 91% of which were in the order Hemiptera – only in the first month post-exclosure (July; Tables 1, 2; Figure 5). Bird and bat exclusion had no effects on arthropod predators or chewing herbivores (Table 1). However, bird and bat exclusion increased Diptera density by 28% without peccaries, but decreased Diptera density by 32% with peccaries (bird/bat × peccary interaction; Table 2, Figure 2).
Figure 4. Effects of the significant two-way interaction between bird/bat exclosure and time on total arthropod density (mean ± SE) in June, July and October 2007 and April 2008 at La Selva Biological Station, Costa Rica. Significant treatment effects, adjusted for four monthly comparisons, are indicated as follows: * P < 0.05; ** P < 0.01.
Figure 5. Effects of the significant two-way interaction between bird/bat exclosure and time on phloem-feeding herbivore density (mean ± SE) in June, July and October 2007 and April 2008 at La Selva Biological Station, Costa Rica. Significant treatment effects, adjusted for four monthly comparisons, are indicated as follows: ** P < 0.01.
Excluding birds and bats increased per cent leaf damage by 35% without peccaries, consistent with a trophic cascade (bird/bat × peccary interaction; Table 1, Figures 3a, b). However, bird and bat exclusion reduced leaf damage by 38% with peccaries, contrary to a trophic cascade (Figures 3a, b). Whereas per cent leaf damage increased by 49% during the experiment when birds and bats were excluded without peccaries, leaf damage only nominally increased when birds and bats were excluded with peccaries. Across all treatments, an increase in arthropod predator density led to reduced leaf damage (predators explained a significant proportion of the variation in leaf damage, F 1,14 = 6.5, P = 0.02) while chewing-herbivore density did not explain any variation in leaf damage (F 1,14 = 4.3, P = 0.06).
Temporal and seasonal effects
Densities of all arthropod guilds and all orders except Lepidoptera showed strong temporal and seasonal trends (time effects; Tables 1, 2). Total arthropod densities, across all bird/bat and peccary exclosure treatments, remained little-changed between June, July and October 2007 (wet season), then more than doubled by April 2008 (dry season; Figures 1a, 2). Arthropod predator density showed a similar temporal pattern, though predators increased more on plants with peccaries than without between October 2007 and April 2008 (Figure 1b). Conversely, phloem-feeding herbivore densities more than doubled between July and October 2007 on plants both with and without birds/bats, while herbivore densities nearly unchanged (without birds/bats) or slightly declined (with birds/bats) between October and April 2008 (Figure 5). Though per cent leaf damage increased over time on plants exposed to both birds/bats and peccaries, and on plants from which both birds/bats and peccaries were excluded, per cent leaf damage on the remaining plants (with birds/bats but without peccaries, and without birds/bats but with peccaries) remained nearly constant over the course of the experiment (Figure 3b). Thus time effects did not contribute to the leaf damage model (Table 1).
DISCUSSION
Results of this study supported both of our hypotheses. Insectivorous bird and bat exclusion increased arthropod densities, specifically of phloem-feeding herbivores, Hemiptera, and all arthropods combined. Bird and bat predation initiated a trophic cascade, in which insectivorous birds and bats protected plants by reducing leaf damage – but only within experimental peccary exclosures. This was not a density-mediated trophic cascade, because birds and bats did not reduce the density of chewing herbivores. Instead, the reduction of leaf damage by birds and bats only within peccary exclosures appears to be an example of a trait-mediated trophic cascade. In trait-mediated trophic cascades, the risk of predation can lead herbivores to reduce foraging activity in the presence of predators in order to increase vigilance (Schmitz et al. Reference SCHMITZ, KRIVAN and OVADIA2004). As a result of this trade-off, leaf damage would be reduced on plants accessible to birds and bats (specifically, those within the peccary exclosures in our study), even though chewing herbivore density was not affected by bird/bat presence.
Our finding that birds and bats reduced leaf damage within peccary exclosures is entirely consistent with previous work demonstrating that insectivorous birds and bats in tropical forests reduce the abundance of herbivores, thereby causing a decrease in plant damage (Gruner Reference GRUNER2004, Mäntylä et al. Reference MÄNTYLÄ, KLEMOLA and LAAKSONEN2011, Van Bael & Brawn Reference VAN BAEL and BRAWN2005, Van Bael et al. Reference VAN BAEL, BRAWN and ROBINSON2003). However, while cascades involving vertebrate insectivores have been shown in the understorey of more seasonal tropical forests (Beard et al. Reference BEARD, ESCHTRUTH, VOGT, VOGT and SCATENA2003, Boege & Marquis Reference BOEGE and MARQUIS2006, Kalka et al. Reference KALKA, SMITH and KALKO2008), to our knowledge ours is the first study to provide experimental evidence for such a cascade in the understorey of wet tropical forests such as La Selva. This finding contradicts theory suggesting that the low productivity and complex food webs of such wet forest understorey habitats alone would tend to weaken these trophic links (reviewed in Van Bael et al. Reference VAN BAEL, BRAWN and ROBINSON2003).
Our results also provide two lines of evidence for our second hypothesis, that peccaries not only weakened these trophic links but also negated the trophic cascade. First, contrary to the accumulation of leaf damage predicted by trophic cascade theory and observed on plants within bird/bat and peccary exclosures, plants within bird/bat exclosures with peccaries maintained nearly constant levels of leaf damage throughout the experiment. Consequently, excluding birds and bats actually decreased plant damage, relative to both control plants (with birds, bats and peccaries) and dual-exclosure plants (without birds, bats and peccaries). The greater leaf damage on plants with versus without birds and bats in peccary control plots could be attributed to consumption, trampling, and/or abrasion by peccaries, except for the fact that we never recorded a single event of peccary browsing on any of our focal plant species. However, direct effects of peccaries cannot explain the dramatically lower leaf damage on plants without birds and bats in peccary controls relative to exclosures, as peccaries were unable to access either set of plants.
Second, bird/bat exclosure significantly reduced Diptera density on plants with peccaries, contrary to the increase predicted by trophic cascade theory. Peccaries significantly increased Diptera density on plants with birds/bats, potentially due to elevated nutrient inputs from urine and faeces (Stewart Reference STEWART2001). However, Diptera density on plants with peccaries but without birds/bats was significantly lower than both nearby plants with birds/bats, and plants within dual-exclosures (without peccaries, birds or bats). This also cannot easily be explained by direct effects of peccaries, and further indicates peccary disruption of the bird/bat–arthropod–plant trophic cascade.
The finding that peccaries negate an insectivorous bird/bat-generated trophic cascade was clear and highly significant. To our knowledge, this is the first demonstration that the presence of an omnivorous ungulate can mediate a trophic cascade via strictly non-trophic means. Collared peccaries are highly abundant at La Selva (Kuprewicz Reference KUPREWICZ2013), thus their effects on the trophic cascade may be particularly pronounced at this site. However, collared peccaries are fairly common at other protected Neotropical sites (e.g. Barro Colorado Island; Wright et al. Reference WRIGHT, ZEBALLOS, DOMÍNGUEZ, GALLARDO, MORENO and IBÁÑEZ2000), and other ungulates such as wild pig (S. scrofa) and deer (Odocoileus spp.) are highly abundant in many tropical and temperate forests (Côté et al. Reference CÔTÉ, ROONEY, TREMBLAY, DUSSAULT and WALLER2004, Ickes Reference ICKES2001). Further studies are needed to investigate the effects of peccaries and other ungulates at a range of population densities and across sites.
Our results raise the question of what caused the declines in Diptera density and leaf damage inside these netted exclosures, but only in the presence of peccaries (i.e. outside peccary exclosures). It could not have been due to intraguild predation as typically defined in these types of trophic cascades, i.e. release of arthropod predators from predation and competition, as we explain here. Many similar studies have found that arthropod predators increase when birds and bats are excluded, and consequently suppress arthropod herbivores and reduce leaf damage (Vance-Chalcraft et al. Reference VANCE-CHALCRAFT, ROSENHEIM, VONESH, OSENBERG and SIH2007; but see Mooney et al. Reference MOONEY, GRUNER, BARBER, VAN BAEL, PHILPOTT and GREENBERG2010). Peccaries affected arthropod predator densities, but in order for intraguild predation to explain the reduced leaf damage on plants within bird/bat exclosures with peccaries, arthropod predator densities would have to be both (1) consistently greater with peccaries, and (2) greater within bird/bat exclosures than controls. Instead, we found that peccaries actually reduced arthropod predator densities in July and October 2007, and birds and bats had no effects on arthropod predator densities. While we cannot rule out intraguild predation altogether, the very small F value for the effect of bird/bat exclosure on arthropod predator density (F 1,14 = 0.32) makes a Type II error unlikely in this case.
While speculative, we suggest instead that the bird/bat exclosures provided both a disturbance- and predator-free refuge for small insectivorous herpetofauna (e.g. anoles and leaf-litter frogs) that were too small to be excluded by the netting. Leaf-litter amphibians and reptiles, including insectivorous Anolis (Norops) spp. anoles and Craugastor spp. frogs, are significantly more abundant with peccaries on the same peccary exclosure and control plots at La Selva as were used in our study (Reider et al. Reference REIDER, CARSON and DONNELLY2013). Moreover, we occasionally observed these herpetofauna within our bird/bat exclosures, and on our focal plants, during the study (N. Michel, pers. obs.). The bird/bat exclosures also excluded peccaries, with very few exceptions. The bird/bat exclosures may have served as a shelter that provided all the proposed benefits of recent peccary visitation (increased nutrient availability, litter mixing and rapid decomposition; Reider et al. Reference REIDER, CARSON and DONNELLY2013), without the risk of being trampled or even consumed by peccaries. Indeed, removing leaf litter from the top of the bird/bat exclosure netting and dispersing it within the cage monthly mixed the litter and may have increased decomposition rates, mimicking some peccary effects associated with greater herpetofaunal abundance (Reider et al. Reference REIDER, CARSON and DONNELLY2013). Furthermore, our netted exclosures would have also excluded many other ground-dwelling and avian predators of small amphibians and reptiles (e.g. raptors, coati Nasua narica, tinamous (Tinamidae)). Unfortunately, we have no quantitative data on herpetofauna in our study because we were not aware of how important this might be at the time we designed it.
If true, congregation of insectivorous amphibians and reptiles within bird/bat exclosures in peccary controls could result in a trait-mediated trophic cascade similar to that observed in the peccary exclosures due to the actions of birds and bats (Schmitz et al. Reference SCHMITZ, KRIVAN and OVADIA2004). If peccaries drove herpetofauna to accumulate within the netted exclosures, they could account for the observed declines in plant damage because both Anolis lizards and leaf-litter frogs are known to cause declines in insect herbivore abundance and leaf damage (Beard et al. Reference BEARD, ESCHTRUTH, VOGT, VOGT and SCATENA2003, Dial & Roughgarden Reference DIAL and ROUGHGARDEN1995). Although we also found no differences in chewing herbivore density between plants with versus without birds and bats in peccary controls, a foraging-predation risk trade-off, as described above, may have occurred here as well. If the predation risk imposed by abundant insectivorous herpetofauna within the bird/bat exclosures exceeded the predation risk outside the exclosures, leaf damage would be reduced within the netted exclosures – as we observed. Thus the peccaries, which are highly abundant at La Selva (Kuprewicz Reference KUPREWICZ2013) and have relatively small home ranges (38–305 ha in tropical forest; Keuroghlian et al. Reference KEUROGHLIAN, EATON and LONGLAND2004), may have repeatedly herded vulnerable small herpetofauna inside our netted exclosures, initiating a trait-mediated trophic cascade that reduced Diptera density and leaf damage. Further studies are needed to investigate herpetofauna densities within and outside bird/bat exclosures with and without peccaries.
Alternatively, it is possible that the netted bird/bat exclosures in the peccary control plots (but not within the peccary exclosures) attracted highly motile arthropod predators that were not detected during arthropod surveys. These arthropods may have sought shelter from disturbance within the exclosures as speculated above for herpetofauna, and in turn suppressed leaf damage. While this seems unlikely for volant arthropod predators (e.g. wasps, lady beetles) that can easily evade peccary disturbance, terrestrial arthropod predators (e.g. wolf spiders) could have sought refuge within the exclosures. Moreover, spiders that returned to the leaf litter during the arthropod surveys would not have been recorded. Other possible explanations, including a trait-mediated trophic cascade altering arthropod predator or herbivore behaviour or rapid leaf regrowth in the presence of peccaries – but only within netted exclosures – seem less likely. Further studies testing these or other hypothetical mechanisms should be conducted. But ultimately, the finding that peccaries indirectly negate a bird/bat–arthropod–plant trophic cascade is novel in and of itself, whatever the mechanism.
Effects of seasonality
Arthropod densities were highly seasonal, peaking in the late dry season. Arthropod densities peak in humid rain-forest understorey during the late dry season as arthropods seek moist refugia (Richards & Windsor Reference RICHARDS and WINDSOR2007). Moreover, as leaf-flushing peaks in February at the study site, the availability of young leaves preferred by arthropod herbivores – and, consequently, densities of herbivores and their predators – also peak in the late dry season (Coley & Barone Reference COLEY and BARONE1996).
The direction and strength of bird/bat- and peccary-exclosure treatment effects on arthropods also changed over time. Interestingly, bird/bat exclosure effects on total arthropod densities were strongest in July, a period of low plant productivity and low predator diversity (due to absence of Nearctic-Neotropical migrant birds), contrary to trophic cascade theory as well as previous studies (Philpott et al. Reference PHILPOTT, SOONG, LOWENSTEIN, PULIDO, LOPEZ, FLYNN and DECLERCK2009, Williams-Guillén et al. Reference WILLIAMS-GUILLÉN, PERFECTO and VANDERMEER2008; but see Borer et al. Reference BORER, SEABLOOM, SHURNIN, ANDERSON, BLANCHETTER and HALPERN2005). Instead, we suggest that bird/bat top-down effects may be strongest shortly after exclosure construction, as herbivorous and omnivorous arthropods seek refuge from predators within the novel structures. This top-down effect may dampen over time as arthropod and small vertebrate predators are attracted to the high prey densities within the exclosures, or as palatable leaf material declines in availability due to consumption or induced defences (Karban & Baldwin Reference KARBAN and BALDWIN1997, Vance-Chalcraft et al. Reference VANCE-CHALCRAFT, ROSENHEIM, VONESH, OSENBERG and SIH2007). Though rarely discussed, top-down effect strength peaks within a few weeks or months post-exclosure construction are apparent in many previous tropical vertebrate predator exclosure studies (Gruner Reference GRUNER2004, Johnson et al. Reference JOHNSON, KELLERMANN and STERCHO2010, Morrison & Lindell Reference MORRISON and LINDELL2012, Williams-Guillén et al. Reference WILLIAMS-GUILLÉN, PERFECTO and VANDERMEER2008).
Peccary exclosure treatment effects also varied seasonally. Peccaries reduced total arthropod densities in July and predatory arthropod densities in July and October, but increased predatory arthropod densities in April. We suggest that arthropods may be attracted to the peccary exclosures with their greater vegetation density (i.e. food, shelter; Michel Reference MICHEL2012) in the low-productivity wet season, whereas plant productivity may be greater in the peccary controls in the dry season due to nutrient inputs (i.e. urine, dung; Reider et al. Reference REIDER, CARSON and DONNELLY2013, Stewart Reference STEWART2001). However, further studies of the mechanism behind the seasonal reversal of peccary exclosure effects are needed.
Ultimately, seasonal fluctuations in arthropod densities did not cascade to the plant level: per cent leaf damage either consistently increased (bird/bat exclosure without peccaries, bird/bat control with peccaries) or remained mostly unchanged over the course of the experiment. Consequently, the trophic cascade observed on plants in bird/bat exclosures without peccaries, and its notable absence with peccaries, was unaffected by seasonality in arthropod densities.
Conclusions
Our conclusions strike a cautionary note regarding the use of bird netting to exclude insectivorous birds and bats. These exclosures do more than just exclude birds and bats; rather they exclude any vertebrate that cannot breach the bird-netting barrier as well as larger disturbance agents such as peccaries. The potential for insectivores other than birds, bats and herpetofauna to cause trophic cascades should be further explored in subsequent exclosure studies. We also suggest that there are natural analogues to our netted exclosures. Lianas can form dense impenetrable tangles (Schnitzer et al. Reference SCHNITZER, DALLING and CARSON2000) that may impede passage by peccaries and other vertebrates (e.g. raptors), thus potentially providing natural shelter from trampling and predation (Lambert et al. Reference LAMBERT, MALCOLM and ZIMMERMAN2006). Finally, our results show that ungulates can interact with insectivorous birds and bats in unexpected ways, with cascading consequences for arthropods and plants.
We stress that, regardless of the specific mechanism by which peccaries reduced leaf damage in the absence of birds and bats, the finding that an omnivorous ungulate can reverse a bird/bat–arthropod–plant trophic cascade is novel, and has important theoretical and conservation implications. Our work adds to the expanding literature showing that the magnitude – and even the direction – of species interactions such as trophic cascades depend on the biotic context (Agrawal et al. Reference AGRAWAL, ACKERLY, ADLER, ARNOLD, CÁCERES, DOAK, POST, HUDSON, MARON, MOONEY, POWER, SCHEMSKE, STACHOWICZ, STRAUSS, TURNER and WERNER2007). In our study, birds and bats protected plants from leaf damage only in the absence of peccaries. However, understorey insectivorous birds are experiencing population declines in many tropical forests, while at the same time large ungulates like peccaries are being hunted to extirpation (Harrison Reference HARRISON2011, Michel & Sherry Reference MICHEL, SHERRY, Sudarshana, Nageswara-Rao and Sonegi2012, Sodhi et al. Reference SODHI, ŞEKERCIOĞLU, BARLOW and ROBINSON2011). Unless gleaning bats are able to compensate for declining insectivorous bird populations, these concurrent trends are likely to expose understorey plants to increased arthropod herbivory pressure, as observed on plants within bird/bat and peccary exclosures in this study. Given the on-going changes in large ungulate, bird, bat and herpetofaunal populations in tropical forests (Peters et al. Reference PETERS, MALCOLM and ZIMMERMAN2006, Sodhi et al. Reference SODHI, ŞEKERCIOĞLU, BARLOW and ROBINSON2011, Terborgh & Estes Reference TERBORGH and ESTES2010, Whitfield et al. Reference WHITFIELD, BELL, PHILIPPI, SASA, BOLAÑOS, CHAVES, SAVAGE and DONNELLY2007), we urgently need to improve our understanding of the potentially complex interactions involving these important species and the potential consequences for tropical biodiversity.
ACKNOWLEDGEMENTS
This work was supported by an Organization for Tropical Studies Research Fellowship to N. Michel, a National Science Foundation (NSF) Doctoral Dissertation Improvement Grant to N. Michel and T. Sherry (DEB-1010952), NSF grant DEB-071743 to T. Sherry, and a LA Board of Regents Graduate Fellowship from Tulane University. We thank Rebecca Forkner for assistance with research design, Deedra McClearn and the staff of La Selva Biological Station for their support and assistance, Orlando Vargas Ramírez for plant identification assistance, and Julie Jackson Lewis and James Lewis for assistance in the field. Lee Dyer, Ben Hirsch, Egbert Leigh, Tim Nuttle, Sunshine Van Bael and several anonymous reviewers provided constructive comments that greatly improved this manuscript.
Appendix 1. List of numbers of individuals of each plant species used in the experiment, grouped by order (per APG III, 2009) for analysis. The number of plants of each species used in bird and bat exclosures and controls are listed in parentheses in the form (exclosure/control).
