SEQUENTIAL PREDATION IN A COMPLE X LIFE-HISTORY: INTERACTIONS AMONG EGG, LARVAL, AND POST-M ETAMORPHIC PREDATORS OF THE EAST AFRICAN TREEFROG, Hyperolius spinigularis By JAMES RICHARD VONESH A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2003
Copyright 2003 by James Richard Vonesh
To Sophia and Savannah.
ACKNOWLEDGMENTS This research could not have been accomplished without the assistance and support of a large number of people. I thank my field assistants, G. Matthews, B. Munizi, S. Mtunguja, A. Kajiru, and S. Balcomb, D. Sutherlin, M. Pauley, and E. Harper for assistance with data collection; C. Sawe, V. Pohjonen of ANR, and K. Howell, C. Msuya, and R. Senzota of the University of Dar es Salaam, for logistical support in the field; V. Clausnitzer for odonate identification; W. Mathis for dipteran identification; and my dissertation committee C. Osenberg (chair), B. Bolker, K. Sieving, L. Chapman, H. Lillywhite, C. St. Mary, and the SOB lab group for valuable feedback though out the entire PhD process, I am a better scientist and person because of our interactions these past years. I thank Tanzanian COSTECH (Permit # 2001-274) and the Division for Forestry and Beekeeping for permission to conduct research and the Tanzania Weather Service, NIMR, Amani, for rainfall data. Financial support was provided by an EPA STAR Fellowship, NSF DDIG DEB-9911965 and the Lincoln Park Zoo. Finally, I thank my family for their encouragement, support, and patience. iv
TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT.......................................................................................................................xi CHAPTER 1 INTRODUCTION........................................................................................................1 2 EGG PREDATION AND PREDATOR-INDUCED HATCHING PLASTICITY IN THE AFRICAN TREEFROG, Hyperolius spinigularis..............................................8 Introduction...................................................................................................................8 Methods......................................................................................................................10 Study Site.............................................................................................................10 Hyperolius Abundance, Reproduction, and Egg Survival...................................11 Hatching Plasticity Experiments: Afrixalus fornasini.........................................14 Hatching Plasticity Experiment: Typopsilopa Fly Larvae...................................15 Results.........................................................................................................................16 Hyperolius spinigularis Phenology.....................................................................16 Oviposition Characteristics and Clutch Success.................................................17 Sources of Egg Mortality.....................................................................................19 Egg Predator Effects on Larval Input into the Pond............................................20 Egg Predator Effects on Larval Traits.................................................................21 Discussion...................................................................................................................22 Clutch Success.....................................................................................................22 Egg-stage Predator Effects on Larval-stage Density...........................................24 Egg-stage Predator Effects on Hatching and Larval Traits.................................26 Consequences in Subsequent Life-stages............................................................28 3 CARRY OVER OF PREDATOR EFFECTS ACROSS THREE LIFE-HISTORY STAGES IN AN AFRICAN TREEFROG.................................................................42 Introduction.................................................................................................................42 Methods and Materials...............................................................................................44 v
Site Information...................................................................................................44 Natural History of Hyperolius spinigularis and its Predators.............................44 Experiment 1: Effects of Density on Larval Performance and Postmetamorphic Traits...................................................................................46 Experiment 2: Size-selective Post-metamorphic Predation................................48 Results.........................................................................................................................49 Experiment 1: Survival as a Function of Initial Density.....................................49 Experiment 1: Duration of the Larval Stage as a Function of Initial Density.....52 Experiment 1: Size at Metamorphosis as a Function of Initial Density..............54 Experiment 2: Size-specific Metamorph Predation.............................................55 Discussion...................................................................................................................55 4 CONSEQUENCES OF PREDATOR-INDUCED HATCHING PLASTICITY IN AN AFRICAN TREEFROG......................................................................................66 Introduction.................................................................................................................66 Materials and Methods...............................................................................................68 Site and System Information...............................................................................68 Site information............................................................................................68 System information......................................................................................69 Aquatic larval-stage predators......................................................................70 Consequences of Egg-Stage Predator Effects for Growth and Survival to Metamorphosis.................................................................................................71 Experimental design: density and size/age effects.......................................71 Establishing the egg-predation mediated density effect: N-EP and N+EP1, N+EP2........................................................................................................73 Establishing predator-induced size effect: S-EP and S+EP.............................76 Integrating Patterns of Larval Growth, and Densityand Size-Specific Risk to Predict Larval Mortality...................................................................................77 Larval growth...............................................................................................77 Effect of density on risk: the functional response........................................77 Effect of larval size on predation risk..........................................................79 Combining size and density specific risk.....................................................81 Simulating larval growth and mortality.......................................................82 Results.........................................................................................................................83 Aquatic Predators................................................................................................83 Larval Survival....................................................................................................83 Proportion of Surviving Larvae that Reach Metamorphosis...............................84 Mass at Metamorphosis.......................................................................................85 Growth Rates.......................................................................................................86 Density and Size-Specific Risk...........................................................................86 Comparison of Observed versus Simulated Larval Survival..............................87 Discussion...................................................................................................................88 vi
5 MULTI-PREDATOR EFFECTS ACROSS LIFE-HISTORY STAGES: NON-ADDITIVITY OF EGGAND LARVAL-STAGE PREDATION IN AN AFRICAN TREEFROG...........................................................................................103 Introduction...............................................................................................................103 Materials and Methods.............................................................................................105 Site Information.................................................................................................105 Experimental design..........................................................................................106 Detecting Multiple Predator Effects..................................................................107 Results.......................................................................................................................109 Discussion.................................................................................................................109 6 SUMMARY..............................................................................................................115 LIST OF REFERENCES.................................................................................................120 BIOGRAPHICAL SKETCH...........................................................................................131 vii
LIST OF TABLES Table page 2-1 Results from likelihood ratio test for each term in the logistic regression model......................................................................................................................29 2-2 Summary of effects from logistic regression on clutch success for environmental and clutch parameters....................................................................30 3-1 Summary of model fits describing larval-stage survival as a function of initial larval density..........................................................................................................60 3-2 Summary of model fits describing size at metamorphosis as a function of initial density....................................................................................................................61 4-1 Experimental design...............................................................................................94 4-2 Results of analysis for effects of density, predator, and size treatments on Hyperolius spinigularis larval survival, proportion of survivors to metamorphose, and metamorph mass....................................................................95 4-3 Model parameter estimates (and 95% confidence limits) decribing larval growth, the functional response of T. basilaris larvae preying upon recently hatched H. spinigularis larvae, and size-specific predation risk of H. spinigularis larvae to late instar T. basilaris.............................................................................................96 5-1 Experimental design.............................................................................................112 viii
LIST OF FIGURES Figure page 1-1 A typical amphibian life-cycle illustrating sequential predator effects on different life-history stages.......................................................................................................7 2-1 Life history of H. spinigularis..................................................................................31 2-2 A. Afrixalus fornasini adult female..........................................................................32 2-3 Typopsilopa sp. dipteran predator............................................................................33 2-4 A. Amani Pond, Amani Nature Reserve, East Usambara Mountians, NE Tanzania. B. Cup placed under H. spinigularis clutch to catch hatchlings...............................34 2-5 Rainfall and seasonal breeding activity of H. spinigularis at Amani Pond.............35 2-6 Sources of H. spinigularis clutch mortality.............................................................36 2-7 Embryonic survival for clutches in different fate categories...................................37 2-8 Effect of egg stage predators on the input of tadpoles over 10 d intervals through the study period........................................................................................................38 2-9 Results from the experiment to test for Afrixalus-induced effects on the timing of hatching and traits of hatchlings...............................................................................39 2-10 Results from the experiment to test for Typopsilopa-induced effects on the timing of hatching and traits of hatchlings..........................................................................40 2-11 A comparison of a predator-induced early hatched H. spinigularis larva and a larva from an undisturbed clutch..............................................................................41 3-1 Relationship between H. spinigularis larval-stage survival and initial density in the presence and absence of larval-stage predators..................................................62 3-2 Relationship between the length of the larval-stage (days to metamorphosis) for H. spinigularis and initial density in the presence and absence of aquatic predators...................................................................................................................63 3-3 Results from experiment 1 showing metamorph mass (g) at eight initial densities in the presence and absence of larval-stage predators..............................................64 ix
3-4 Proportional survival of small and large metamorph size classes............................65 4-1 Aquatic predators.....................................................................................................97 4-2 Results from the tank experiment.............................................................................99 4-3 Parameterization of the model................................................................................101 4-4 Larval survival: comparison of experimental and simulation results....................102 5-1 Testing for a non-additive MPE Effects of egg-stage predation (via Afrixalus fornasini) and larval-stage predation (via dragonfly larvae)..................................113 5-2 Examining the mechanisms Effects of the density and size/age effects of Afrixalus egg predation on aquatic larval survival in the presence of dragonflies..114 x
Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SEQUENTIAL PREDATION IN A COMPLEX LIFE-HISTORY: INTERACTIONS AMONG EGG, LARVAL, AND POST-METAMORPHIC PREDATORS ON THE EAST AFRICAN TREEFROG, Hyperolius spinigularis By James Richard Vonesh December 2003 Chair: Craig W. Osenberg Major Department: Zoology Most prey are vulnerable to more than one species of predator and the aggregate effects of multiple predators on a shared prey are often less than would be expected from their independent effects. This result, called risk reduction, typically arises from to direct predator-predator interactions (e.g., intraguild predation). However, the prevalence of risk reduction may reflect a bias in the types of systems we have studied. While almost all studies have examined prey with complex life cycles, so far ecologists have examined only the effects of predators of a single life-history stage. Predator-predator interactions (and thus risk reduction) are likely within a life-stage, since predators overlap in space and time. For prey with complex life histories, shared predators will not overlap in space and time. Sequential predators attack different life-stages in different habitats and will therefore be unlikely to interact directly. However, these predators may still interact indirectly, if early predator effects on prey density or traits alter subsequent predator-prey xi
interactions. Such indirect effects acting across stages and may be common in nature and important aspects of many food webs. This dissertation examines the densityand trait-mediated effects of arboreal eggand aquatic larval-stage predators of the African treefrog Hyperolius spinigularis. Through field experiments informed by field observations, I quantified both the overall multi-predator effects of egg and larval predators and the contribution of the densityand size-mediated mechanisms. I also examined the effects on eggand larval-stage predators on size at metamorphosis and size-selective post-metamorphic predation. My results showed the effects of egg-stage and larval-stage predators were not independent significantly more Hyperolius survived than predicted from predators separate effects. Both the density effect and (surprisingly) the reduced size effect of egg-stage predators decreased the effectiveness of larval-stage predators. Similarly, predator effects early in the life cycle increased size at metamorphosis and larger metamorphs were more likely to survive encounters with post-metamorphic predators. Thus, I observed multiple predator effects that resulted in risk reduction; however, risk reduction arose in the absence of the direct predator-predator interactions mechanisms reported in previous studies. xii
CHAPTER 1 INTRODUCTION Understanding the dynamics of communities is a fundamental goal of ecology. Because of the inherent complexity of natural communities, many studies have taken a reductionist approach, focusing on interactions of species pairs. It was believed that by assembling all possible pairwise interactions, the workings of the entire system could be predicted. In order for this approach to be successful, the nature of each interaction must remain essentially the same regardless of which other species are added to the system (Wooton 1994). However, if pairwise interactions are modified by the presence of other species in the community, these simple models of species interactions may be unable to describe and predict the behavior of more complex communities. In these cases, new higher order terms are required to successfully model community dynamics (Wooton 1993, 1994; Billick and Case 1994). Studies in the last decade suggest that such higher order interactions (HOIs) and indirect effects are ubiquitous and are frequently important in determining the community dynamics. For example, a recent review of experimental studies involving multiple predators reported that multiple predators had emergent effects on prey in 18 of 28 studies (Sih et al. 1998). However, while it is clear that HOIs are common in natural communities, ecologists know little about when to expect them, what form they will take under different conditions, or what their long-term implications are for population dynamics. There are two generic types of emergent multiple predator effects (MPEs) that can be defined based upon their deviation from a simple additive model of predator effects 1
2 (i.e., predator effects are independent): prey risk reduction (when the observed prey mortality rate is less than the rate predicted from an additive model) and risk enhancement (when the observed mortality rate is greater than that predicted from the additive model). Risk reduction for the prey species may result from interference (or predation) between predators (Hurd and Eisenberg 1990, Fauth 1990, Rosenheim et al. 1993), which reduces at least one predators feeding rate on the prey. Risk enhancement may result when prey exhibit conflicting responses to different predators. In this scenario the prey response to predator A increases its vulnerability to predator B, and vice versa (Martin et al. 1989). To date, the literature suggests that risk reduction is the more common phenomenon. Sih et al. (1998) found that only five studies of the 28 reviewed reported risk enhancement, whereas 13 reported risk reduction (eight found no, or trivial, MPEs). However, it is unclear from these studies what factors give rise to risk enhancement or reduction. Given that even simple natural communities typically involve multiple predators feeding on most prey, understanding emergent predator effects, and specifically predicting when risk enhancement or reduction will occur, is critical to understanding how predation acts to structure communities (Wilbur and Fauth 1990). Sih et al. (1998) argued that one of the most important factors in determining whether risk enhancement or reduction is likely to occur is the degree of overlap between predators in foraging habitat. In cases where different predators use the same habitat, interactions between them may be common, resulting in opportunities for predator interference or intraguild predation. Both of these types of interactions tend to result in risk reduction. Interestingly, all 28 studies reviewed by Sih et al. (1998) were performed within single habitats (e.g., ponds or streams). The few cases of risk enhancement arose
3 when there were differences in small-scale patterns of microhabitat use (e.g., tops vs. bottoms of boulders), but in general predators overlapped in their habitat use and consumed prey within the same habitat. This likely accounts for the high preponderance of risk reduction observed in these studies. Of the studies reviewed by Sih et al. (1998), 24 focused on prey species that undergo metamorphosis and change habitats ontogenetically. As a result, these species potentially interact with an entirely different suite of predators at different points in their life history (Fig. 1-1). Even species that do not metamorphose can exhibit dramatic ontogenetic habitat shifts (e.g., Werner and Gilliam 1984, Mittelbach and Osenberg 1993). None of the studies Sih et al. (1998) reviewed looked for interactions among predators of different stages. Thus, the preponderance of studies that detected risk reduction MPEs may reflect a bias in how the studies were conducted. Predators that prey upon different life history stages of the same species may not overlap in foraging habitat and therefore have little opportunity to interfere with each other. There is still great potential for emergent MPEs in such cases, although they may be more likely to lead to risk enhancement. For example, the presence of aquatic predators may reduce the foraging activity of amphibian larvae, which may then metamorphose at a smaller size, making them more vulnerable to size-selective terrestrial predators. To my knowledge, no study has examined whether stage-specific predation in complex life histories results in emergent MPEs. Given the common occurrence of habitat shifts and their importance to ecological interactions (e.g., Werner 1986, Osenberg et al. 1994), this represents a major void in the study of multi-predator effects.
4 The complexity of natural communities has posed a considerable challenge to experimental studies of multiple predator effects, especially when there are MPEs and when communities are speciose. The tropics are not only species-rich, but the rich predator assemblages of the tropics have led to the suggestion that predation plays a greater role in structuring tropical communities than temperate communities (Paine 1966, Janzen 1970, Connell 1978). Thus, MPEs may be more common in the tropics and may pose a greater challenge to the development of theory that can be used to understand the dynamics of tropical food webs. Despite the hypothesized role of predators in tropical communities, relatively few studies have examined predation in tropical aquatic systems, and most of these studies have relied on a correlative approach (Gascon 1991, 1992, Hero et al. 1998, Azevedo-Ramos et al. 1999). A more experimental approach to studies of predation, akin to those being conducted in temperate regions (e.g., Soluk 1993, Wooton 1994, Morin 1995), is needed to determine the role of predators in tropical systems. In the chapters that follow, I examine the effects of and interactions among sequential stage-specific predators of the African reed frog, Hyperolius spinigularis. This is a model system for examining the effects of MPEs that arise from stage-specific predation because the life history of the prey allows the effects of terrestrial stage and aquatic-stage predators to be isolated and manipulated. Hyperolius spinigularis oviposits on vegetation above water. Upon hatching, tadpoles drop into the pool, where they remain until they metamorphose into terrestrial juveniles. Hence the early life history consists of two stages (an arboreal egg stage and an aquatic larval stage) that are vulnerable to different suites of predators. The egg stage is vulnerable to predation by other treefrogs and by dipteran larvae. However, when predators attack a clutch, not all
5 frog embryos within a clutch are consumed. Those that are not consumed may hatch at a smaller size, increasing their vulnerability to size selective aquatic-stage predators (e.g., Warkentin 1995, Vonesh 2000). As a result, survivors of egg-stage predators may be more vulnerable to larval predation than tadpoles from uninfested clutches. This scenario may result in risk enhancement. In Chapter 2, I examine effects of egg-stage predators on H. spinigularis clutches at Amani Pond, Amani Nature Reserve, in the East Usambara Mountians of Tanzania. Through a 10-month field study in which I examine H. spinigularis reproduction and egg-stage survival, I evaluate which predators have the greatest effect on egg-stage survival and estimate how much egg-stage predation reduces the input of H. spinigularis input to the aquatic habitat. This provides me with an estimate of the density-mediated effects of egg-stage predators. I also evaluate effects of egg predators on traits of surviving larvae. Through laboratory experiments I examine the effects of egg predators on time, developmental stage, and size at hatching. In Chapter 3, I focus primarily on the density effect (as described in Chapter 1) of egg-stage and larval-stage predators. I ask how reductions in larval density (e.g., which also reduce larval competition for resources) due to predators affect the size at metamorphosis of surviving frogs. I then examine whether these effects on metamorph size alter interactions with a very common post-metamorphic predator of H. spinigularis, pisaurid fishing spiders. In Chapter 4, I focus on the consequences of predator-induced early hatching (as described in Chapter 1) for larval growth and survival and size at metamorphosis given the simultaneous density effects of egg-stage predators. I found a surprising result,
6 smaller larvae that hatch early because of egg-predators actually survived better than larger, later hatched larvae. In the second part of Chapter 4, I develop an explanation for this observation using a mathematical simulation. The parameters used in this simulation derived from additional, independent experiments designed to quantify larval growth rates, and size and density specific rates of predation. In Chapter 5, I evaluate whether the combined effects of sequential eggand larval stage predators of H. spinigularis are independent (i.e., additive) or if we observe an emergent multiple predator effect (Sih et al. 1998). I evaluate whether any MPE is in the direction of risk reduction or risk enhancement, and then I quantify the relative contribution of the densityand size/age-mediated effects of egg predators on the efficacy of larval stage predators. Chapter 6 summarizes the key results of this research.
7 Figure 1-1.A typical amphibian life-cycle illustrating sequential predator effects on different life-history stages. In this example predators attack aquatic larvae and terrestrial juveniles.
CHAPTER 2 EGG PREDATION AND PREDATOR-INDUCED HATCHING PLASTICITY IN THE AFRICAN TREEFROG, Hyperolius spinigularis Introduction Indirect effects of predators (or consumers) are common and important in ecological communities (reviewed in Wootton 1994, Werner and Peacor 2003). The indirect effects of predators are mediated through one of two general mechanisms: 1) through a chain of direct interactions in which a predator affects a second species through changes in the density of a third species (density-mediated indirect effects), and/or 2) through interaction modifications, in which a species alters the strength of the interaction (i.e., changes the interaction coefficient) between an existing species pair (e.g., Schmitt et al. 1983, Wootton 1994). The latter has also been called a higher order interaction (Vandermeer 1969) and a trait-mediated indirect effect (e.g., Werner 1992, Werner and Peacor 2003). There are many examples of density-mediated effects and their effects on ecological communities [e.g., exploitative competition (Schoener 1983, Goldberg and Barton 1992); trophic cascades (Estes and Palminsano 1974, Carpenter et al. 1985, Spiller and Schoener 1996); apparent competition (Holt 1977, Schmitt 1987)]. Recent work also has demonstrated the role of interaction modifications in ecological communities (reviewed in Bolker et al. 2003, Werner and Peacor 2003). For example, predator induced trait-mediated effects on prey can alter prey interactions with other predators (Sih and Moore 1993, Van Buskirk and Schmidt 2000), competitors (Relyea 2000), and resources (Peacor 2002). Despite the evidence that both types of indirect 8
9 effects can be important in ecological communities, relatively few studies examine the effects of predators on both the density and phenotype of their prey (Vonesh 2000, Warkentin 2000, Peacor and Werner 2001). In many species, early life stages are more vulnerable to predators than later life stages (e.g., Werner et al. 1983, Alford 1999, Fuiman and Magurran 1994). Indeed, the eggs of many species are subject to high levels of predation (e.g., insects: Tanhaunp et al. 2003; lizards: Chalcraft and Andrews 1999; fishes: Dorn 2003; birds: Martin and Jornon 2003). Among anurans with aquatic eggs, Petranka and Kennedy (1999) observed catastrophic mortality (~100%) of Rana sylvatica egg masses due to predation by Rana clamitans tadpoles. High levels of predation on arboreal anuran clutches have also been observed (Vonesh 2000, Warkentin 2000, Lips 2001, Villa 1984). Given the strong selective pressure exerted by egg-stage predators, we might expect eggs to be sensitive to predation risk. Indeed, hatching may be an adaptive life history switch point (Sih and Moore 1993, Warkentin 1995, Li 2002). Life history theory predicts that organisms should time early life-history switch points to minimize the ratio of mortality/growth (Werner and Gilliam 1984, Werner 1986). In addition, if predation risk varies through time and/or among sites we might expect selection to favor plasticity in hatching strategies. Thus, theory predicts that organisms should delay hatching when the perceived risk from post-hatching predators is high relative to the threat from egg predators. The postponement of hatching may allow hatchlings to reach a greater size before encountering predators, potentially increasing their survival of such encounters (Sih and Moore 1993). Conversely, elevated risk to eggs should favor early hatching. Predator-induced hatching plasticity has been demonstrated in salamanders (Sih and
10 Moore 1993), anurans (Warkentin 1995, Chivers et al. 2001, Laurila et al. 2002, Schalk et al. 2002), arachnids (Li 2002), and fishes (Wedekind 2002, Jones et al. 2003), and may occur in crustaceans (Blaustein 1997). In this study I examine effects of egg-stage predation on both hatchling density and on the timing of hatching in the African treefrog, Hyperolius spinigularis (Fig. 2-1a-d). This species breeds during the two annual rainy seasons, depositing its eggs on vegetation overhanging water (Fig. 2-1bc). Hyperolius eggs are vulnerable to predation by the confamilial treefrog, Afrixalus fornasini (Fig. 2-2ab) and larvae of an ephydrid fly in the genus Typopsilopa (Fig. 2-3a-d), as well as to abiotic sources of mortality, such as desiccation and drowning. Upon hatching (Fig.2-1cd), larvae drop into the pond where they face a new suite of predators. In this study I describe seasonal patterns of reproduction and quantify sources of arboreal egg-stage mortality in H. spinigularis to evaluate effects of egg predators on input of hatchlings into the aquatic habitat. I also experimentally examine effects of Afrixalus fornasini and Typopsilopa fly predators on timing of hatching and hatchling characteristics. Through this combined approach, I hope to gain a better understanding of the densityand trait effects of egg-stage predators and their implications for post-hatching survival, growth, and development. Methods Study Site This research was conducted at the Amani Nature Reserve (ANR) Conservation Headquarters in the East Usambara Mountains of northeastern Tanzania (5.06 S and 38.37 E; Elevation; 900 m) from May June, 2000 and October 2001 August 2002. ANR includes 8380 ha of transitional lowland-montane rainforest. The site receives
11 approximately 2000 mm rainfall each year (Hamilton and Bensted-Smith 1986), which falls primarily in two distinct rainy seasons: October November and March June. ANR and the forests of the East Usambara Mountains are among the threatened Eastern Arc Mountain biodiversity hotspots (Myers et al. 2000) noted for their high degree of diversity and endemism. Field observations were made at Amani Pond (~200 m North ANR HQs). Amani Pond is a man-made (> 50 yr old) permanent shallow pond (average depth ~ 45 cm) bordered by submontane rainforest (Fig. 2-4a). Pond vegetation is dominated by floating mats of milfoil (Myriophyllum spicatum) and marginal patches of emergent cattails (Typha sp.). Hyperolius Abundance, Reproduction, and Egg Survival Between October 2001 and August 2002, I censused H. spinigularis adults and clutches along two randomly placed 3 x 30 m transects in Amani pond. The range of microhabitats along each transect included those available in the pond (e.g., pond edge, open water, milfoil, and cattail emergent vegetation). Adult abundance at the pond was monitored opportunistically 1 5 times each month (25 total over study) beginning at 1900 h and continuing ~ 2 h until I had searched both transects using visual encounter survey techniques (Donnelly and Guyer 1994, Heyer et al. 1999). During each census, I also measured air and water temperature, and humidity. I alternated starting locations each census to reduce biases due to search order and observer fatigue. These nocturnal surveys also enabled me to make opportunistic observations of predation on H. spinigularis clutches and adults. Clutch censuses (67 total) were conducted along the same transects at 2 3 day intervals except during the dry season (Dec 15 Feb 15) when transects were checked once per week. Each census began at ~0830 h and continued 2-8 h until both transects
12 were traversed. For each new clutch observed, I counted and staged the eggs [according to Gosner (1960)], noted the substrate type, the location of the clutch on the substrate (e.g., top or bottom of leaf or on branch), and the height of the clutch above water. Stevens (1971) noted that H. spinigularis females can remain with clutches after they are deposited, potentially providing some form of parental care. Thus, I was careful to note the presence of females on or near each clutch. Females did not appear to be disturbed by census activities. However, it is possible that females may have abandoned clutches before we observed them. I marked the location of each clutch by placing numbered and dated flagging on adjacent vegetation. On subsequent censuses, each clutch was rechecked to determine survival and developmental progress. As eggs neared hatching (> Gosner stage 16 18), I suspended a water-filled plastic cup beneath the clutch to capture hatchlings (Hayes 1983, Warkentin 1995, Lips 2001; Fig. 2-4b). Each cup had small drainage holes to prevent overflow during heavy rains and was covered with plastic screen (grid width = 10 mm), which allowed tadpoles to drop into the cup but prevented access by predators (e.g., large spiders and other frogs). I assessed egg mortality rates based on the difference between the initial number of eggs and the number of hatchlings. Because predators and other mortality agents leave specific evidence of their activity (Hayes 1983, Warkentin 2000, Lips 2001), I was almost always able to identify the primary source of egg mortality for a clutch. Multiple sources of mortality for a clutch were seldom observed, and for these clutches I considered only the largest source of mortality. To evaluate the factors most important in determining clutch success, I defined clutch success as a binomial variable (i.e., clutches either produced or failed to produce
13 surviving tadpoles) and used multiple logistic regression to test effects of a number of environmental and clutch parameters on this binary response variable (Agresti 1996). Predictors initially included in the model were date, air temperature (C), water temperature at surface (C), monthly rain fall (mm), % humidity on the day the clutch was first observed, clutch height above water (cm), water depth below clutch (cm), clutch size (initial embryos clutch-1), and three categorical variables; presence or absence of a female H. spinigularis, location on substrate (top, bottom, centerline), and substrate type (floating vegetation, primarily milfoil; emergent vegetation, primarily sedges and cattails; and overhanging vegetation, primarily tree ferns and trees). Statistical analyses were performed in R version 1.7.0, an open source language and environment for statistical computing and graphics (Ihaka and Gentleman 1996; http://www.r-project.org/). I used the STEP procedure in R to reduce these predictor variables to an optimal subset (Venables and Ripley 2002). STEP selects an optimal subset of predictors by both adding and subtracting terms and selecting the model with the lowest AIC (Akiake information criterion) value. Arboreal egg-stage predators reduce the input of tadpoles dropping into the aquatic habitat by consuming eggs. To evaluate this effect of egg predators on hatchling recruitment (i.e., tadpole input) into the aquatic habitat, I compared average daily input of tadpoles across 10-day windows through the study, given the presence and absence of particular predator effects. First, I estimated the input of tadpoles given only abiotic sources of mortality (i.e., in the absence of egg predator effects) by using the field estimated egg survival for healthy clutches and clutches experiencing abiotic mortality and by assuming that clutches that were attacked by predators experienced egg survival
14 similar to healthy clutches. I then summed the number of surviving hatchlings over each 10-d window and divided by the number of days to estimate the average daily input. This upper estimate of tadpole input from arboreal eggs was then decremented by including the mortality caused by A. fornasini alone, Typopsilopa sp. flies, and all egg predators combined. This was accomplished using the method above, but substituting the field estimated survival for clutches attacked by the particular predator type, rather than assuming healthy egg survivorship. Hatching Plasticity Experiments: Afrixalus fornasini To determine the sublethal effects of A. fornasini predation on H. spinigularis hatchlings, I compared hatching time and hatchling characteristics for control embryos with embryos that had been exposed to this treefrog. Both treatments (A. fornasini predator versus control) were replicated six times. On April 25, 2002, I collected 12 new H. spinigularis clutches from the field. These clutches were removed from their original substrate and transferred onto the upper surface of a standard sized (~14 x 8 cm) leaf of Tabor montana, a common oviposition substrate for this species in Amani pond. Each leaf-clutch was secured to a support in a vertical orientation and placed in individual 11-L plastic containers with tight fitting, air-permeable lids. Containers were randomly arranged in a 4 x 3 array and maintained indoors at ambient temperatures. Each container was filled to a depth of 1 cm with rainwater. On April 29 I began the experiment by randomly introducing one female A. fornasini to each of the six containers assigned to the predator treatment. These females had been collected from the field April 28 and were of similar size (Snoutto-vent length (SVL): 34 2.5 mm ( SD)). Prior to introducing the predators, I assessed the developmental stage (~ Gosner 16) and noted any eggs that
15 had failed to develop in each clutch. Twice per day clutches were misted using a spray bottle to prevent desiccation. Every morning I examined containers until all larvae had hatched, been consumed, or died. Hatched larvae were euthanized with an overdose of MS-222 anesthetic, preserved in 90% ethanol, and digitally photographed. The total lengths (TL) of all larvae were measured from the digital images using the open source digital image analysis software ImageJ (http://rsb.info.nih.gov/ij/). I used general linear models (Venables and Ripley 2002) to examine the effects of A. fornasini on four response variables: 1) the proportion of larvae surviving to the end of the experiment; 2) number of days to hatching; 3) developmental stage at hatching (according Gosner (1960)); and 4) larval total length (mm) at hatching. Proportional survival data were arc-sin square root transformed prior to analysis. Hatching Plasticity Experiment: Typopsilopa Fly Larvae I compared hatching time and hatchling characteristics of embryos that had been exposed to flies with those not exposed. Both treatments (fly predator versus control) were replicated four times. On May 15, 2000, I collected four new and uninfested H. spinigularis clutches from Amani pond. These clutches were carefully removed from their original leaf substrate, divided into a control and a treatment portions of 40 eggs each (any remaining eggs were not used), and transferred onto a Tabor montana leaf (as in the treefrog predation experiment). Treatment sections were inoculated with 10 Typopsilopa eggs consistent with the ratios of fly eggs to frog eggs (~1 fly: 4 frog eggs) observed between May 1 and May 15 at Amani Pond (Vonesh unpubl. data). Inoculations were done by carefully transferring fly eggs from recently infested H. spinigularis clutches collected from Amani pond (Vonesh 2000). Clutch portions were placed in a vertical orientation in 11-L plastic containers containing 1 cm rainwater and
16 were misted using a spray bottle twice a day to prevent desiccation. Every morning I examined containers until all larvae had hatched, been consumed, or died. Hatched larvae were euthanized and preserved in ethanol. Survival of fly larvae used in the experiment to pupation was high (92%). Measurements were made to the nearest 0.1 mm using calipers. A paired t-test was used for statistical analyses. Results Hyperolius spinigularis Phenology The mean monthly abundance of adult Hyperolius spinigularis adults in Amani Pond was positively correlated with monthly rainfall (r = 0.92, P <0.001). Adults were not observed in Amani Pond during the dry season (late Dec. 2001 late Feb. 2002), and adult activity was greatest during the March June rainy season. In 2002, adults appeared in late February and abundance peaked in early April (67 adults observed at the peak: ~ 0.37 adults m-1), the wettest month of the study (Fig. 2-5). Activity was lower during the October November rainy season, which may reflect the natural breeding phenology of H. spinigularis, but could also reflect that 2001 rainfall levels were well below average (October November rainfall in Amani, Tanzania: 27 mm in 2001 versus a mean of 340 mm from 1902 to 1970). Male H. spinigularis averaged ( SD) 18.8 0.8 mm SVL and weighed 0.35 0.04 g (n = 81); females were larger, averaging 24.4 1.2 mm SVL and weighing nearly twice as much as the males (0.76 0.09 g; n = 50). Hyperolius spinigularis clutch abundance was also positively correlated with rainfall (r = 0.95, P <0.001) and with adult abundance (r = 0.82, P = 0.007). Fresh clutches were observed from March through mid-July (~4.5 months), but over 75% of clutches were oviposited between 15 March and 15 April (Fig. 2-5).
17 Oviposition Characteristics and Clutch Success Over the entire study period, I followed the fates of 614 H. spinigularis clutches, counted 49,708 developing embryos, and collected 21,493 hatching tadpoles. Clutch size was ( SD) 89 18 eggs clutch-1. Most clutches (62%) were found on vegetation overhanging the pond, 32% were found in mats of floating vegetation (milfoil), and 6% were found on emergent vegetation (e.g., sedges, cattails). Clutches on overhanging vegetation were placed with similar frequency on the top (58%) versus the bottom (42%) surface of the substrate (e.g., leaf, fern frond), whereas clutches on floating and emergent vegetation were always near the plant stem (due to the structural characteristics of these plants). Nearly all clutches were placed over water, however a few clutches were found over dry ground (6%). The height of clutches above water was highly variable and averaged 123 120 cm ( SD). Clutches on the floating vegetation were 11 6 cm above the water, whereas clutches on overhanging vegetation were 166 77 cm above the water with a few clutches as high as 4 m (2%). Female H. spinigularis were found in association with 21% of the clutches. Based on the stepwise selection criteria, clutch size, clutch height, humidity, presence of a female, and substrate were included in the model describing clutch success as a function of environmental factors. Removal of any of these terms significantly reduced the fit of the model (Table 2-1). The odds ratio is an expression of relative chances and provides an intuitive metric for comparing the relative effects of these factors on clutch survival (Agresti 1996). The odds ratio ( ) is given by )1/()1/(22112odds1odds 1.)
18 where 1 is the probability of success (e.g., producing 1 hatchling) given condition 1 and 2 is the probability of success given condition 2 (Agresti 1996, Fox 1997). Conditions 1 and 2 could represent different categorical conditions (e.g., female present versus absent) or unit increases in a continuous explanatory variable. The odds ratio can equal any nonnegative number. The value = 1 indicates that the odds of success are equal (i.e., odds1 = odds2) and serves as a baseline for comparison (Agresti 1996). Odds ratios on each side of 1 reflect particular types of associations. For example, a value = 1.75 indicates that the odds of success given condition 1 were 1.75 times the estimated odds for condition 2, or the odds of success were 75% higher for condition 1 (Agresti 1996). The results show that an increase in clutch size by one egg increases the odds that a clutch will be successful (i.e., produce 1 surviving tadpole) by ~ 4% (Table 2-2). An increase in oviposition height by 1 cm reduced the odds of success by 0.6% and a one percent increase in humidity increased the odds of success by 7%. Substrate type also affected clutch success. Relative to the odds of success for clutches placed on overhanging vegetation, the odds for clutches placed on emergent vegetation were 16% greater and the odds for clutches placed on floating vegetation were 76% lower. Finally, the presence of females at a clutch negatively affected the odds that a clutch would be successful. The odds that clutches without females would be successful was approximately double that of clutches with females (Table 2-2). This did not appear to be due to egg cannibalism, as many healthy clutches with high egg survival also had females in attendance and females were never observed feeding from clutches.
19 Sources of Egg Mortality Based on direct observation of predation by the treefrog A. fornasini, Typopsilopa sp. larvae, ants, and beetle larvae, patterns of egg loss caused by these predators are distinct from each other and from abiotic sources of mortality. Afrixalus fornasini typically consume only the eggs, leaving a partially or completely empty jelly-mass. They consume ~ 5 15 entire eggs per feeding, and may return to a particular clutch multiple times within a night or (potentially) over a period of several days. Predation by Typopsilopa sp. fly larvae was easily distinguished by the presence of fly eggs and larvae within the egg mass, and individual eggs are often only partially consumed (Vonesh 2000). General insect predation (primarily ants and larval coleopterans) was characterized by partial consumption of the most exterior eggs in the clutch and (for beetle larvae) the presence of frass. In total, I identified six main types of clutch mortality: predation by A. fornasini, Typopsilopa sp. fly larvae, and other insects, as well as mortality due to the abiotic factors, desiccation, drowning, and failure to develop (H. spinigularis clutches submerged prior to Gosner stage ~ 13 15 die). Failure to develop refers to clutches in which embryonic development was abnormal or halted for an unusually large proportion of the clutch. The eggs in such clutches were often irregular in size and shape. This type of mortality was distinguished from clutches with a small number (< 5) of unfertilized/undeveloping, normally sized and shaped eggs, which was common in healthy clutches. In addition, I could not relocate 6% of clutches, so their fates could not be determined. Over the entire study period, the most common source of clutch mortality was predation by A. fornasini (41% of total clutches), followed by predation by Typopsilopa sp. flies (12%), desiccation (10%), insect predation (3%), failure to develop (~1%) and
20 drowning (< 1%). Clutches that escaped these sources of mortality, which I refer to as healthy, comprised 28% of all clutches. These sources of mortality varied throughout the course of the study (Fig. 2-6). The proportion of clutches that desiccated was higher during the October November season and after the spring rains (June July 2002). Predator effects also varied temporally. For example, the proportion of clutches attacked by A. fornasini in 2002 was greatest in March (73%), early in the breeding season, and steadily decreased through April (37%) and May (13%). In contrast, the proportion of clutches attacked by dipterans was low in March (1.6%) and peaked late in the breeding season (e.g., 27% in May, Fig. 2-6). Sources of mortality were not independent of female clutch attendance strategy (2 = 17.04, df = 4, P= 0.002). Clutches with females in attendance were slightly more vulnerable to desiccation and predation by A. fornasini. Clutches that experienced different sources of mortality differed in embryonic survival (ANOVA, F5, 554 = 27.36, P < 0.001, drowning excluded because n = 1. Fig. 2-7). Survival was highest (93%) in healthy clutches, as expected. Mortality in healthy clutches was generally due to a few unfertilized or otherwise undeveloping eggs. Survival of eggs in clutches incurring other mortality sources were low and similar, with mean survival ranging from 20 30% in response to the three predator types and 22% for clutches that experienced developmental problems. Desiccation resulted in the lowest embryonic survival, nearly always killing the entire clutch (Fig. 2-7). Egg Predator Effects on Larval Input into the Pond The average new hatchling input into the pond in the absence of predators (for 10-d periods with breeding activity) was 1.5 2 larvae m-2 d-1 ( SD), with peak input occurring in early April with an estimated 6.4 new larvae m-2 d-1 (top of white bars, Fig.
21 2-8). The addition of egg predators reduced larval input, but the magnitude of this reduction differed considerably among types of egg predators. Predation by A. fornasini substantially reduced hatchlings input into the pond during both breeding seasons (Fig. 2-8a,b). During the 2001 breeding season, A. fornasini reduced average H. spinigularis input into the aquatic habitat by 44 44%. In 2002, A. fornasini had the greatest proportional effect early in the breeding season, reducing average H. spinigularis input by 62 15% in March. The effect of A. fornasini declined through the rest of the breeding season, though A. fornasini also had a large proportional effect in June and July when clutch densities were very low. In contrast, predation by Typopsilopa sp. flies had much smaller effects on the input of hatchlings into the pond (Fig. 2-8c,d). In 2001, flies reduced tadpole input (for 10-d periods with breeding activity) by 27 48% and in 2002 they reduced input by 9 11%. Their effect was most pronounced late in the breeding seasons. The combined effects of all predators on new hatchling densities were considerable. In 2001, egg predators reduced hatchling densities by 71% on average. In 2002, the effects of egg predators on hatchling input ranged from an 82% reduction to an 18% reduction and averaged nearly a 50% reduction across the entire season (Fig. 2-8e,f). In general, predator effects on hatchling input decreased throughout the main part of the 2002 breeding season. Egg Predator Effects on Larval Traits Attack by A. fornasini and Typopsilopa fly predators reduced egg survival within a clutch, but also altered the timing of hatching and the developmental stage, and size of surviving hatchlings. Embryonic survival was high (94%) in the absence of A. fornasini, but was reduced to 42% in the predator treatment (F1,12=17.57; P=0.002, Fig 2-9a).
22 Larvae that survived A. fornasini predation hatched 2.6 0.75 ( SE) days earlier than undisturbed clutches (F1,12=24.44; P=0.001; Fig. 2-9b), at 2.3 0.46 Gosner (1960) developmental stages earlier (F1,12=39.29; P<0.001, Fig. 2-9c), and were 27% (2.2 0.48 mm) smaller in total length (F1,12=36.69; P<0.001; Fig. 2-9d). Predation by Typopsilopa larvae had similar sublethal effects. Egg survival was high (91%) in the absence of Typopsilopa fly larvae and reduced to only 4% in the predator treatment (t =7.8; P = 0.016, Fig. 2-10a). There was no survival in one predator replicate, so remaining comparisons had only three paired replicates. Larvae that survived Typopsilopa predation hatched 4 0.62 ( SE) days earlier than undisturbed clutches (t = 5.74; P = 0.029; Fig. 2-10b), at 3.5 0.64 Gosner (1960) developmental stages earlier (t = 6.47, P = 0.023, Fig. 2-10c), and 19% (1.5 0.67 mm) smaller by total length (t = 7.2; P = 0.019; Fig. 2-10d). Discussion Clutch Success There was considerable variation in H. spinigularis oviposition patterns, and this variation had important consequences for the probability of clutch success. Increased humidity early in development had a strong positive effect on the probability of clutch success. Increased humidity likely decreases several sources of clutch mortality. Most obviously, higher humidity reduces the likelihood of desiccation, which was an important factor of clutch mortality late in both breeding seasons. Humidity may also facilitate clutch survival via more indirect mechanisms. For example, well-hydrated clutches typically had a more copious jelly matrix (pers. obs.), which may provide a more substantial barrier to predation by A. fornasini.
23 Substrate type and height also affected the odds of clutch success. Clutches on floating vegetation or on very high substrates had greater odds of producing hatchlings. Clutches on emergent vegetation had the greatest likelihood of success. Emergent vegetation was the most limited substrate type in and around Amani Pond, but it never appeared that this substrate became saturated with clutches. Clutch size was also an important predictor of clutch success. Given the considerable variation in clutch size in H. spinigularis, this is likely to be an important factor in explaining the variation in clutch success at this site. Selective pressure for females to produce larger clutches may also exert selection for larger female body size as female size and clutch size can be positively correlated in anurans (e.g., Collins 1975). Perhaps most interesting was the surprising result that clutches that had females in attendance had lower odds of producing hatchlings than clutches without females. There have been several hypotheses regarding egg attendance in non-aquatic egg-laying anurans. Egg attendance may benefit clutches by reducing predation (McDiarmid 1978, Townsend et al. 1984), pathogens (Vial and Prieb 1967), or desiccation (Taigen 1981). Stevens (1971) suggested that female H. spinigularis help maintain clutch hydration. However, I found no evidence that female attendance reduced the probability of desiccation (desiccation rate was 2% vs. 8% in the absence and presence of the female). It is possible that clutch attendance may make eggs more conspicuous to predators. For example, Skutch (1949) argued that avian nest predators use parental activity to find nests. Clutches with females in attendance were attacked by A. fornasini more frequently (60%) than clutches without females (50%). While clutch attendance may provide a cue to potential predators, the presence of H. spinigularis on a clutch did not appear to be an
24 effective deterrent against A. fornasini predation. On two occasions, I observed an A. fornasini displace a H. spinigularis female before feeding on a clutch. However, clutch attendance may be more effective against fly predation. Flies attacked 6% of clutches without females and 2% with females. Thus, it is possible female H. spinigularis are caught between conflicting predator-specific egg defensive strategies attending eggs is effective against flies but facilitates predation by other frogs. Egg-stage Predator Effects on Larval-stage Density Egg-stage predation of arboreal H. spinigularis clutches reduced tadpole recruitment considerably (Fig. 2-8). This reduction in larval densities likely increases larval-stage growth and development (competition among anurans was reviewed in Skelly and Kiesecker 2001). Faster growing/developing larvae may realize higher survival to metamorphosis as they are exposed to sources of larval mortality (biotic or abiotic) for shorter periods (Wilbur 1977, Dash and Hota 1980). Faster growing larvae may also more rapidly attain a refuge from size-selective aquatic predators again increasing larval survival (e.g., Table 10.3; Werner and Gilliam 1984, Alford 1999). The magnitude of the effect of egg-stage predation on tadpole input into the pond varied with the type of egg predator and through the breeding season. Egg predation by the treefrog A. fornasini had the largest overall effect on H. spinigularis egg-stage survival. Drewes and Altig (1996) observed A. fornasini predation on the foam nests of Chromantis xerampelina in Arabuko-Sokoke Forest, Kenya (~ 200 km N Amani) and found that anuran eggs and/or larvae were the most common food item in the stomachs of specimens from that site (n = 121) and were also found in the stomach of a specimen collected from South Africa. The eggs and/or larvae of Afrixalus (likely A. fornasini), Chiromantis xerampelina, and Hyperolius (likely H. tuberilinguis) were all found in A.
25 fornasini stomachs (Drewes and Altig 1996). Hyperolius tuberilinguis eggs are suspended in a jelly matrix similar to that of H. spinigularis, A. fornasini eggs are enclosed within folded leaves, and Chiromantis eggs are suspended within a foam nest that quickly hardens after the eggs are deposited. Thus, A. fornasini is capable of preying upon a variety of clutch types. The large density effect of A. fornasini was primarily due to the large number of clutches this predator attacked (Fig. 2-6), as there was no difference in embryonic survival among clutches attacked by different types of predators (Fig. 2-7). Over both breeding seasons, Afrixalus attacked 41% of clutches, considerably more than were attacked by other predator types. Within a breeding season, the density effect of A. fornasini varied (Fig. 2-8ab). During the 2002 breeding season the proportion of clutches attacked by A. fornasini gradually decreased through the season until late June and July when clutch densities were very low and proportional effect of low levels of egg predation was large. Since A. fornasini also breeds during the rainy season, it is possible that seasonal variation in egg predation is driven in part by seasonal shifts in activity at the pond due to A. fornasini breeding activity. Dipteran larvae are common egg predators of non-aquatic frog eggs in both the Old and New World (Villa 1980,1984; Yorke 1983, Lips 2001; Davis and Disney 2003), and can be the most important egg-stage predator in some systems (Vonesh 2000, Lips 2001). However, in this system Typopsilopa sp. fly larvae had a small effect on the recruitment of H. spinigularis larvae, compared with the effects of A. fornasini (Fig. 2-8cd). While Typopsilopa sp. flies killed a comparable fraction of eggs (relative to A. fornasini) in attacked clutches (Fig. 2-7), flies attacked fewer clutches (Fig. 2-6). Across both breeding seasons, Typopsilopa sp. infestation rates averaged 12% of the available
26 clutches. This is on the low end of the range of Typopsilopa infestation rates observed for the clutches of four Hyperolius species at two ponds in western Uganda (Vonesh 2000). Typopsilopa sp. infestation rates in Uganda averaged 33% across species and sites, and ranged from 8 to 47% of the available clutches over an eight-month period (Vonesh 2000). In Amani, Typopsilopa infestation rates varied within the breeding season. In 2002, infestation rates were low in March and April and peaked in May in June, when up to 30% of the available clutches were infested. I observed similar levels of infestation in May June during my preliminary field season at Amani in 2000 (Vonesh unpublished data). Interestingly, the number of Typopsilopa infested clutches was relatively consistent through the breeding seasons in Uganda (Vonesh 2000) where A. fornasini is absent and flies were the primary egg predators suggesting the possibility that flies could have been temporally displaced by A. fornasini. Egg-stage Predator Effects on Hatching and Larval Traits In addition to their effects on larval density, egg predators can have important effects on larval traits by influencing the timing of hatching. Theory developed for organisms with complex life histories predicts that the timing of transitions between two life stages should evolve in response to variation in growth and mortality rates in the two stages (Werner and Gilliam 1984, Werner 1986). Recently, this theoretical framework has been applied to examine the timing of hatching in response to predator cues (Sih and Moore 1993). Examples of such predator-induced hatching responses are becoming common in the literature. The eggs of anurans (Warkentin 1995, 1999ab, 2000; Vonesh 2000; Chivers et al. 2001; Warkentin et al. 2001), fishes (Wedekind 2002), and arachnids (Li 2002) have been shown to hatch earlier in response to cues from egg-stage predators, potentially increasing egg-stage survival. We found a similar result, embryos that survive
27 clutch predation by both A. fornasini (Fig. 2-9b) and Typopsilopa sp. fly larvae (Fig. 2-10b) hatched earlier than undisturbed clutches. Larvae that hatched earlier because of a predator induced hatching response entered the aquatic habitat at an earlier developmental stage (Figs 2-9c and 2-10c) and at a smaller size (Figs 2-9d and 2-10d and 2-11). Predator-induced early hatching in H. spinigularis is less rapid than the behaviorally-mediated hatching response observed in some other anuran species (e.g., Agalychnis callidryas, Warkentin 1995), but is still likely to be effective at increasing egg-stage survival given the foraging modes of the egg predators in this system. Female Typopsilopa flies oviposit on Hyperolius clutches very soon after they are laid and fly larvae hatch within 24-hr and begin to feed on the developing frog embryos. Thus, embryos may detect cues of fly predation from early in development and respond by initiating the mechanisms that lead to hatching earlier in ontogeny. Similarly, A. fornasini feed on a few embryos from a clutch at time and possibly revisit clutches over multiple nights (Vonesh pers. obs.). Thus even a relatively slow response could enhance embryonic survival. While rapid behaviorally-mediated hatching responses to predators have been observed for egg predation by snakes, which can rapidly consume a clutch (Warkentin 1995), adaptive hatching plasticity in H. spinigularis can occur via slower physiological or developmental mechanisms due to the slower (relative to snakes) predation rate of A. fornasini and flies. For example, hatching plasticity in H. spinigularis may be mediated through plasticity in enzymatic process. Embryos manufacture and secrete proteolytic enzymes that digest the egg capsule and jelly enabling the embryo to escape (Caroll and Hedrick 1974).
28 Warkentin (1995) suggested that the timing of hatching is maintained by a trade-off in embryonic versus larval predation risk, because smaller, early-hatched larvae are more vulnerable to aquatic predators. Indeed, theory suggests that predator specific defenses (e.g., hatching early) should increase vulnerability to other predators (Matsuda et al. 1994). If this is the case, then the effects of egg predators on larval density (above) and larval traits may act in opposite directions. Reductions in larval density may facilitate survival, and predator effects on hatchling traits (i.e., reduced size and development) may reduce larval survival. Indeed, most evidence suggests that smaller tadpoles are more vulnerable to predators (Alford 1999). Consequences in Subsequent Life-stages In experimental studies designed to evaluate the indirect effects of egg-stage predators on predator-prey interactions during the aquatic larval-stage, I found that both the reduced density and (in contrast to expectations above) hatching trait effects of egg predators enhanced larval-stage survival in the presence of aquatic predators. Tadpoles at lower densities survived better and tadpoles that hatched earlier at smaller initial sizes survived better in the presence of larval dragonfly predators. Predator-induced early hatched tadpoles were initially smaller, but exhibited higher growth rates than tadpoles from undisturbed clutches and grew more rapidly through vulnerable size classes (Vonesh and Osenberg 2003). In addition, egg-stage predator effects on density and size both affected size at metamorphosis and the interaction between metamorphs and fishing spiders, which are size-selective predators of H. spinigularis (Chapter 3). Thus, the effects of predators of early stages of prey with complex life histories can have important consequences for predator-prey interactions in subsequent life stages.
29 Table 2-1. Results from likelihood ratio test for each term in the logistic regression model. Term 2 Df P Clutch size 38.543 1 < 0.001 Height 14.712 1 < 0.001 Humidity 20.173 1 < 0.001 Female 4.433 1 0.035 Substrate 30.539 2 < 0.001
30 Table 2-2. Summary of effects from logistic regression on clutch success for environmental and clutch parameters. Variable Coef. SE t-ratio P Odds95% CI Odds-ratio Upper Lower Intercept -7.004 1.281 -5.469 <0.001 Clutch size 0.037 0.007 5.273 <0.001 1.037 1.052 1.023 Height -0.006 0.002 -3.070 0.002 0.994 0.990 0.998 Humidity 0.065 0.014 4.639 <0.001 1.067 1.097 1.038 Female Present Female Absent 0.688 0.266 2.587 0.010 1.989 3.348 1.181 Substrate Overhanging Substrate Emergent -0.222 0.507 -0.437 0.662 0.801 2.166 0.296 Substrate Floating -1.414 0.398 -3.552 <0.001 0.243 0.531 0.112
31 Figure 2-1. Life history of H. spinigularis. A. An adult male. B. An adult female on a recently oviposited clutch. C. A clutch in the process of hatching, larvae dropping into the pond. D. Newly hatched larvae. A B. C. D.
32 Figure 2-2.A. Afrixalus fornasini adult female. B. A. fornasini preying upon a H. spinigularis clutch. A B.
33 Figure 2-3. Typopsilopa sp. dipteran predator. A. Adult fly. B. Adult female fly ovipositing Hyperolius puncticulatus clutch. C. Close-up of Typopsilopa eggs next to H. spinigularis embryo (~ Gosner stage 10). D. Close-up of Typopsilopa larvae in H. spinigularis clutch. A 1 mm scale is provided for approximate reference. A B. C. D.
34 Figure 2-4.A. Amani Pond, Amani Nature Reserve, East Usambara Mountians, NE Tanzania. B. Cup placed under H. spinigularis clutch to catch hatchlings. A B.
35 Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug No. adults or clutches 0255075100125150175 Rainfall (mm mo-1) 050100150200250300350 H. spinigularis clutches H. spinigularis adults monthly rainfall Figure 2-5.Rainfall and seasonal breeding activity of H. spinigularis at Amani Pond, Amani Nature Reserve, in the East Usambara Mountains of N.E. Tanzania between October 2001 and August 2002.
36 No. clutches 020406080100120140160 Healthy Afrixalus Dipteran Insect Dried Undeveloped Drowned Lost Date Oct-01Dec-01Feb-02Apr-02Jun-02Aug-02Proportion of clutches 0.00.20.40.60.81.0 AB Figure 2-6.Sources of H. spinigularis clutch mortality between 24 October 2001 and 19 July 2002, showing A) the mean number, or B) the proportion of clutches in each of six mortality categories. I monitored 614 H. spinigularis clutches along two transects in Amani Pond and could determine the primary source of mortality (if any) for 564 of these clutches. Clutches experienced one of six possible fates: healthy, Afrixalus fornasini predation, Typopsilopa sp. fly predation, predation by other invertebrates, desiccation, failure to develop, drowning, or unknown/lost. Lost clutches in panel B. are indicated by the difference between the top of each bar and 1.0.
37 HealthyAfrixalusDipteranInsectDesiccationUndevelopedDrownProportion embryonic survival 0.00.20.40.60.81.0 PredationAbiotic Figure 2-7. Embryonic survival for clutches in different fate categories: healthy, predation by A. fornasini, Typopsilopa sp. flies, or other invertebrates, and abiotic mortality caused by desiccation, failure to develop, or drowning. Bars indicate means 1 SD. Bars with different letter labels were statistically different from each other based on univariate ANOVA followed by post-hoc Fishers LSD. Sample sizes for each category were: Healthy/undisturbed (n=173); predation by Afrixalus fornasini (n = 297), Typopsilopa sp. flies (n = 28), other insects (n = 27); and abiotic sources of mortality, desiccation (n = 19), drowning (n = 1), and developmental failure (n = 19). A B B B B NA C
38 01234567 Afrixalus No Predator 0.00.20.40.60.81.0 Larval input (larvae m-2 d-1) 01234567 Dipteran No Predator Proportional reduction in larval input 0.00.20.40.60.81.0 Date Oct Dec Feb Apr Jun Aug 01234567 Egg predation No Predator Date Oct Dec Feb Apr Jun Aug 0.00.20.40.60.81.0 2001200220012002 A.B.C.D.E.F. Figure 2-8.Effect of egg stage predators on the input of tadpoles over 10 d intervals through the study period. The first column of panels (A, C, E) provides the estimated input of new hatchings dropping into the aquatic environment. The top of the white bar indicates densities in the absence of one or more egg stage predators (but includes abiotic mortality). The top of the black bar indicates density with the egg predator effect. The second column of panels (B, D, F) provides the proportional reduction in tadpole densities due to egg stage predators. Panels A and B show the effect of egg predation by A. fornasini on tadpole density, panels C and D show the effect of Typopsilopa sp. dipteran egg predation, and panels E and F show the effects of all egg predators combined.
39 Embryonic survival 0.00.20.40.60.81.0 Hatching time (d) 024681012 No PredatorAfrixalusDevelopmental stage (Gosner) 04812162024 No PredatorAfrixalusTotal length (mm) 0246810 ABCDABABABAB Figure 2-9.Results from the experiment to test for Afrixalus-induced effects on the timing of hatching and traits of hatchlings. A. Embryonic survival, B. Time to hatching, C. Hatchling developmental stage, D. Hatchling size. Bars indicate means 1 SE. Bars with different letter labels were statistically different from each other.
40 Embryonic survival 0.00.20.40.60.81.0 Hatching time (d) 024681012 No PredatorDipteran larvaeDevelopmental stage (Gosner) 04812162024 No PredatorDipteran larvaeTotal length (mm) 0246810 ACDBABABABAB Figure 2-10.Results from the experiment to test for Typopsilopa-induced effects on the timing of hatching and traits of hatchlings. A. Embryonic survival, B. Time to hatching, C. Hatchling developmental stage, D. Hatchling size. Bars indicate means 1 SE. Bars with different letter labels were statistically different from each other.
41 Figure 2-11. A comparison of a predator-induced early hatched H. spinigularis larva and a larva from an undisturbed clutch.
CHAPTER 3 CARRY OVER OF PREDATOR EFFECTS ACROSS THREE LIFE-HISTORY STAGES IN AN AFRICAN TREEFROG Introduction Studies of organisms with complex life cycles often focus on a single life stage independent of previous or subsequent stages. However, the studies that have examined multiple stages have found that effects on one stage can be carried over to affect performance in subsequent life stages. For example, embryonic-stage effects may alter larval performance (e.g., Warkentin 1995, Vonesh and Osenberg 2003), and larval-stage conditions may affect adult performance and fitness (Anholt 1991, Moeur and Istock 1980, Goater 1994, McPeek and Peckarsky 1998). Studies of carry-over are usually limited to adjacent stages, but these effects could be transmitted to even later life stages, although such effects are difficult to examine in most systems. In many cases, these carry over effects are mediated through effects of density on vital rates such as mortality or growth (Anholt 1991, Goater 1994, McPeek and Peckarsky 1998, Vonesh and De la Cruz 2002). For example, in amphibians, changes in larval-stage density can alter adult survival, time to maturity, lipid stores, mating success, and fecundity (e.g., Scott 1994, Altwegg and Reyer 2003). Changes in density can be driven by predation, and thus there are situations where predation on one life stage can influence the density and growth of a later life stage, which might affect performance in yet a later life stage. The degree to which performance in one stage is linked to performance in a later stage is critical for understanding the dynamics of stage-structured populations. 42
43 In this study, I examine the effects of sequential stage-specific predators of the African reed frog, Hyperolius spinigularis. Hyperolius spinigularis oviposits on vegetation above water, where eggs are vulnerable to predation by another hyperoliid frog, Afrixalus fornasini (Drewes and Altig 1996, Chapter 2) and several invertebrates (Vonesh 2000, Chapter 2). Upon hatching, larvae drop into the pond, where they are vulnerable to aquatic predators, such as larval dragonflies, and upon metamorphosis, frogs climb out of the pond onto emergent or floating vegetation, where they are vulnerable to a new suite of predators, including fishing spiders. Hence the early life history consists of three discrete stages egg, larval, and postmetamorphic that are vulnerable to different predators. The effects of predators on early stages may alter H. spinigularis performance and species interactions in subsequent stages. For example, by consuming eggs, egg-stage predators reduce the input of larvae into the pond. This reduction in larval density may reduce larval competition, potentially increasing larval-stage growth and size at metamorphosis. Reduced larval input into the pond may also alter the interactions of H. spinigularis larvae with their aquatic predators (e.g., if predation is sizeand/or density-dependent). Aquatic larval-stage predators also reduce larval densities and may alter larval traits (e.g., foraging activity), further altering larval-stage growth and size at metamorphosis. Finally, eggand larval-stage effects on traits at metamorphosis may alter interactions with postmetamorphic predators. Thus, there is the potential for the effects of predators to carry over across three life stages. The primary goals of this study are to examine how changes in larval density (reflecting the numerical effects of egg-stage predators) and the presence of aquatic predators affect the larval
44 stage survival, duration and size at metamorphosis, and then to examine whether changes in metamorph size affect survival in the presence of fishing spiders. Methods and Materials Site Information This research was conducted at the Amani Nature Reserve (ANR) Conservation Headquarters in the East Usambara Mountains of northeastern Tanzania (5.06 S and 38.37 E; Elevation; 900 m) from October 2001 August 2002. ANR includes 8380 ha of transitional lowland-montane rainforest. The site receives approximately 2000 mm rainfall each year (Hamilton and Benstead-Smith 1989), which falls primarily in two distinct rainy seasons: October November (short rains) and March June (long rains). ANR and the forests of the Usambara Mountains are among the threatened Eastern Arc Mountain biodiversity hotspots noted for their high degree of diversity and endemism (Lovett and Wasser 1993 Myers et al. 2000). Field measurements of H. spinigularis clutch densities, egg predation, and aquatic predator densities were made at Amani Pond (~200 m North ANR field station) unless otherwise noted below. Amani Pond is a man-made (> 50 yr old) permanent shallow pond (average depth ~ 45 cm) bordered by submontane rainforest. Pond vegetation is dominated by floating mats of milfoil (Myriophyllum spicatum) and marginal patches of emergent cattails (Typha sp). Natural History of Hyperolius spinigularis and its Predators Hyperolius spinigularis is endemic to a few submontane rainforest localities in Tanzania and Malawi (Schitz 1999). Male H. spinigularis average 18.8 mm snout-to-vent length (SVL) and females average 24.4 mm SVL (Chapter 2). It breeds during both annual rainy seasons by attaching its eggs to vegetation overhanging permanent or semi-permanent ponds or swamps. Mean clutch size is 89 eggs clutch-1 (Chapter 2). In 2002,
45 reproductive activity peaked early in the rainy season (i.e., in March) when the density of arboreal clutches reached 0.85 clutches m-2 (Chapter 2). Egg-stage predation, primarily by Afrixalus fornasini and ephydrid fly larvae, has two effects on tadpoles: 1) they induce early hatching, so that survivors are younger and smaller than in the absence of predators (Vonesh 2000, Vonesh and Osenberg 2003, Chapter 2), and 2) they consume eggs and therefore reduce the input of tadpoles into the aquatic habitat (Chapter 2). Here, I concentrate on the density effect. During March 2002, the first month of breeding activity (over which 75% of reproductive activity occurred), larval input into the pond from arboreal eggs over a 10 d period (~ 1 cohort given variation in hatching time) averaged 20 larvae m-2 (range: 5 42 larvae m-2). In the absence of egg predation, larval input for the same period would have averaged 43 larvae m-2 (range: 13 63 larvae m-2). Thus, egg predators reduced larval input by 53% on average (range: 40 82% reduction). Upon hatching the tadpoles fall into the water where they encounter a new suite of predators. In Amani Pond, potential aquatic predators include; dragonfly larvae (primarily libellulids), damselfly larvae (Zygoptera), adult diving beetles (Dytiscus sp.), water scorpions (Nepa sp.), African clawed frogs (Xenopus muelleri), aquatic snakes (Natricitares olivacea), and fish (primarily Gambusia sp.). Aquatic plot sampling and predation trials established that larvae of the libellulid dragonfly, Trapezostigma basilaris, were the most abundant aquatic predator in Amani Pond (Chapter 4) and were more effective predators of recently hatched H. spinigularis larvae than any of the five most abundant aquatic predators (Chapter 4).
46 Upon metamorphosis, frogs initially emerge from the water onto aquatic vegetation, where they encounter a third suite of predators, including wading birds, kingfishers, semi-aquatic snakes, and fishing spiders. Here, I focus on fishing spiders (Pisauridae, Thalassius sp.) because they prey on both metamorph and adult hyperoliid frogs in the field (pers obs) and are present at high densities in Amani Pond during late May as larvae begin to metamorphose (( SD) 7.8 4.7 spiders m-2; n = 12 random plots 18 20 May 2002). Thus, fishing spiders are likely to be an important metamorph-stage predator. My observations of predation in the field and laboratory suggest that spiders attack frogs by grasping them with their chelicerae and hanging on until venom immobilizes the prey a process that can take many minutes. During this period, the frogs try to escape, often pulling the spider along through the vegetation. Thus, it seems likely that larger body size (e.g., at metamorphosis) may be beneficial when encountering fishing spiders larger frogs may be more difficult to catch and if caught may take longer before becoming immobilized by the venom and thus may be more likely to break the spiders hold. Experiment 1: Effects of Density on Larval Performance and Postmetamorphic Traits To estimate the effect of larval density and aquatic predators on larval-stage survival, development, and size at metamorphosis, I conducted an experiment in which I varied the density of larval H. spinigularis in the presence and absence of dragonfly larvae. This experiment was performed in 32 white plastic water storage tanks (1.2 x 0.8 x 0.4 m; Chemi and Cotex Industries, Ltd. TZ.) arranged in a 2 x 16 array in a partially shaded forest clearing. Pools were initially filled to a volume of 220 L with Amani Pond water filtered through a 0.3 mm mesh screen and were immediately covered with tight
47 fitting lids made of fiberglass window screening to prevent unwanted colonization by insects and frogs. Screened drain-holes prevented overflow during heavy rains. Tanks were buried to a depth of 15 cm in shallow trenches to help stabilize water temperature. To each tank I added 200 g of washed and dried pond litter, 10 g of commercial fish food (Hikari brand, Kyorin Co., Ltd; crude protein: min 32%, Crude fiber: max 5%, Moisture: max 10%, Ash: max 12%; main ingredients: fish meal, wheat flour, wheat germ meal), 300 g (~ 25, 40 cm long stems) of washed free-floating macrophytes (M. spicatum), and an inoculum of a plankton collected from Amani Pond using a 80 m plankton net and concentrated into 0.5 L. Tanks were filled with water and constituents in early November 2001 and allowed to equilibrate for approximately two weeks before the start of the experiment. Sixteen treatments were randomly assigned to tanks: 5, 10, 15, 20, 30, 50, 75, 100 H. spinigularis larvae tank-1 (~ 5.2 to 105 larvae m-2) crossed with the presence or absence of 3 larval odonate predators. Each treatment was replicated twice. Hyperolius spinigularis larvae used in the experiment came from 18 clutches collected upstream of Amani Pond that hatched naturally between 15 20 November. On 22 November, all hatchlings were pooled and randomly assigned to treatments. To reduce the effect of initial handling on the experimental outcome, larvae were held overnight and dead larvae were replaced (< 1% initial mortality). Larvae were then digitally photographed to estimate initial size (TL: 9.6 SD 0.23 mm) and added to the tanks. Three late instar T. basilaris larvae (TL: 18.9 SD 2.26 mm) were immediately added to each tank (3 dragonflies tank-1 approximates the mean field density of 3.16 SD 4.78 dragonflies m-2; Chapter 4). Tanks were checked weekly for metamorphosed dragonflies and frogs.
48 Metamorphosed dragonflies were replaced to maintain predator density. The experiment was terminated after 4 months, all tanks were emptied, and predators and any remaining tadpoles collected and counted. I measured the snout-vent length (SVL) of all Hyperolius metamorphs using calipers, and estimated their mass to the nearest 0.01 g using an electronic balance. I examined the effects of odonate predation and quantified the relationships between initial larval density and 1) survival to the end of the experiment, 2) mean larval duration, and 3) mass at metamorphosis. Curve fitting was implemented using Kyplot (3.0) Student Version (Kyence, Inc.) using both linear and non-linear least-squares optimization approaches. Comparison of model fit was conducted using the Akaike Information Criteria (AIC, Akaike 1992). Experiment 2: Size-selective Post-metamorphic Predation To examine whether size at metamorphosis affected vulnerability to a common post-metamorphic predator, I conducted an experiment in which I varied metamorph size in the presence of Thalassius sp. fishing spiders over a 24 h period. I was unable to establish a control for metamorph survival in the absence of predators due to the limited number of appropriately sized metamorphs. However, mortality in the absence of predators should be minimal over the short duration of the experiment (e.g., I have never observed handling-induced mortality of these metamorphs). The experiment was performed in 10 plastic containers (0.2 x 0.4 m) arranged in a 2 x 5 array on a shaded porch area. Containers were initially filled to a volume of 8 L with rainwater, and were covered with tight fitting lids. To each tank I added 10 g of pond litter and 20 g (~ 4, 20 cm long stems) of washed free-floating macrophytes (M. spicatum). Metamorphs collected from the density-manipulation experiment above were used in this experiment. Metamorphs were divided into small and large size classes representing the range in
49 metamorph sizes from the experiment above given the effects of egg-stage predators on larval density crossed with the presence or absence of aquatic-stage predators. Small metamorphs were collected from the treatments without dragonflies (30 and 50 larvae tank-1 densities) and large metamorphs were collected from predator tanks (20 and 30 larvae tank-1 densities). Thus, metamorph size is confounded with predator history and, to a lesser extent, larval density: small metamorphs experienced no larval predators and higher larval densities and large metamorphs experienced aquatic predators and somewhat lower larval densities this confounding represents the natural situation (although it precludes me from isolating the effects of larval density vs. predators vs. metamorph size per se). On 22 January 2002, I added 5 metamorphs of one size class or the other to each container. After 2 h I added 1 Thalassius sp. fishing spider (mean mass: 0.61 SD 0.15 g) to each container. This spider density (1 spider container-1 12.5 spiders m-2) was within the range observed in the field (field mean: 7.8 SD 4.7 spiders m-2). Survivorship was estimated after 24 h by counting the number of live metamorphs. I compared the proportional survival of each size class by univariate ANOVA. Data were arcsine-square root transformed prior to analysis. Results Experiment 1: Survival as a Function of Initial Density I compared the fit of Shepherd (Shepherd 1982) and Beverton-Holt recruitment (Beverton and Holt 1957) and linear functions to the relationship between the number of survivors and initial density. The later two models are nested within the Shepherd function. The Shepherd function is given by )1(cbXaXS (1)
50 where S is the number of survivors, X is the initial density, and a, b, and c are fitted parameters. When c = 1, equation 1 reduces to the Beverton-Holt function and when b or c = 0, it reduces further to a linear model (with intercept = 0). The parameter c, provides an indication of the shape and strength of density dependence. The parameter a is the density-independent survivorship and b is the density-independent survivorship (a) divided by the maximum (i.e., asymptotic) survivor density (Schmitt et al. 1999). In the absence of aquatic predators, the Shepherd function provided the best description of the relationship between initial density and the number of survivors ( AIC: Shepherd versus Beverton-Holt = 10; Shepherd versus Linear: 21.7; Table 3-1a). Thus, the relationship between the number of survivors and initial density was overcompensatory i.e., the number of survivors at an initial density of 100 larvae tank-1 was less than the estimated number of survivors at 75 larvae tank-1 (~ 50 versus 55 survivors; Fig. 3-1a). In the presence of aquatic predators, the Beverton-Holt function provided the best description of the relationship between initial density and the number of survivors ( AIC: Shepherd versus Beverton-Holt = 1; Beverton-Holt versus Linear: 4; Table 3-1a). Thus, the number of survivors was a decelerating function of initial density (Fig. 3-1a). I also used linear least squares regression to quantify how the proportional survival through the larval stage changed with initial larval density. The Beverton-Holt fit to survival in the presence of predators (Fig. 3-1a) implies that proportional survival in the presence of predators in a decreasing linear function of initial density. The Shepherd fit to the survival data in the absence of predators (Fig 3-1a) implies that proportional survival is a monotonically decreasing function of initial density with the rate of decrease
51 accelerating with increasing density. Thus, I used non-linear regression to fit the polynomial, y = aX2 + bX + c, to the proportional survival in the absence of predators (Fig. 3-1b). In the absence of aquatic predators, larval-stage survival was high in the absence of intraspecific competition, i.e., at very low densities, (y-intercept (c): 0.93 [0.81 1.04], (parameter [95% C.I.])) but decreased to ~ 50% at high density (a = -5.73E-5 [-1.19E-4 5.23E-6], b = 0.0014 [-0.0051 0.008], Fig. 3-1b). Dragonfly predators reduced tadpole survival from 93% to 39% (y-intercept: 0.39 [0.29 0.48]) in the absence of density dependent effects. Interestingly, survival did not decrease significantly with increasing density in the presence of predators (slope: -0.0017 [-0.003 0.0027, P = 0.08]), although power was low because of the low survival at all densities in the presence of dragonflies. Above I used the models above to provide a general description of the relationship between density, predation, and larval survival. I can now use these relationships to quantify the survival consequences of specific eggand larval-stage predator effects. To do this I first quantified the overall effect of egg and larval-predators on survival, and then partitioned out and compared the individual contributions of egg-predator carry over effects (i.e., mediated through reduced larval input) and larval-predators effects on survival. I employ the same approach later to quantify the combined and individual contributions of egg and larval predation to larval duration and size at metamorphosis. The estimated larval input into the pond during the peak of the breeding season was 41.3 larvae tank-1 and egg-stage predation reduced this input to 20 larvae tank-1. In the absence of predator effects, (i.e., 41.3 larvae tank-1, no dragonfly) estimated larval survival was 0.89 (95% CI: 0.39 1.37, Fig. 3-1b). The combined effects of both stage
52 specific predators (i.e., 20 larvae tank-1, dragonfly) reduced this by 57% to 0.36 (95% CI: 0.22 0.49). I then partitioned this overall effect in the contributions from egg-stage and larval-stage predator sources. Egg-predator reductions in larval input (i.e., 20 larvae tank-1, no dragonfly) increased survival by 5% to 0.93 (95% CI: 0.66 1.20, Fig. 3-1b). Given that egg-predation facilitated survival, yet the combined predator effects predict a 57% decrease in larval survival, larval predators (in the absence of egg predator effects) should have a large negative effect on survival. Indeed, the estimated survival given only the effects of dragonflies (i.e., 41.3 larvae tank-1, dragonfly) was 64% less (0.32; 95% CI: 0.14 0.49, Fig. 3-1b) than in the absence of predators (i.e., 89%). Experiment 1: Duration of the Larval Stage as a Function of Initial Density The duration of the larval stage increased with increasing initial density (Fig. 3-2). This effect was greatest in the absence of aquatic predators. In the absence of predators and intraspecific density dependence (i.e., at very low densities), the estimated duration of the larval period was 48 days (parameter [95% C.I.]: y-intercept: 48.4 [44.5 52.3]). In contrast, at low densities in the presence of predators, estimated length of the larval period was 13 days longer (y-intercept: 61.6 [58.8 64.4]). This pattern reversed at high densities because the rate at which larval period increased with increasing density was three times higher in the absence of predators (slope: 0.60 [0.45 0.74]) than in the presence of predators (slope: 0.20 [0.14 0.25]). Thus, at densities of 100 larvae tank-1, the estimated larval duration in the absence of predators was more than 30 days longer than in the presence of predators so long, in fact, that many larvae in the no predator, high density treatments failed to reach metamorphosis over the course of the study. Nearly all of the surviving larvae reached metamorphosis at low and medium densities. At densities below 50 larvae tank-1, 97 SD 6% (n = 24 tanks) of surviving larvae
53 reached metamorphosis by the end of the experiment. However, at very high densities, the dragonfly and no dragonfly results diverged. The proportion of metamorphs was still high (89 SD 3%, n = 4 tanks) in the high-density (75 and 100 larvae tank-1) dragonfly treatments. However, in the high density, no predator treatments the proportion of survivors to metamorphose was only 13 SD 13%. I then used this linear model for the relationship between larval duration and initial density in the presence and absence of predators to quantify the overall effect of egg and larval-predators on larval period, and to partition out and compare the individual contributions of egg-predator carry over effects and larval-predators effects. In the absence of predator effects, (i.e., 41.3 larvae tank-1, no dragonfly) estimated larval duration was 73 d (95% CI: 64 83 d, Fig. 3-2). The combined effects of both stage-specific predators (i.e., 20 larvae tank-1, dragonfly) reduced this by 10% to 66 d (95% CI: 62 69 d, Fig. 3-2). I then partitioned this overall effect in the contributions from egg-stage and larval-stage predator sources. Egg-predator effects on larval input (i.e., 20 larvae tank-1, no dragonfly) carried over to decrease the estimated larval duration by 18% to 60 d (95% CI: 54 66 d, Fig. 3-2). Larval-stage predation had little effect on larval duration, while larval period in the presence of dragonflies was only 4% less than in the absence of predators (70 d; 95% CI: 65 75 d, Fig. 3-2). Given the interaction between larval predation and density on larval duration (Fig. 3-2), the combined independent effects of egg and larval predators did not predict the actual aggregate predator effect very well (multiplicative prediction: 21% decrease, observed: 10% decrease).
54 Experiment 1: Size at Metamorphosis as a Function of Initial Density A negative exponential model provided a better description of the relationship between initial density and size at metamorphosis than a linear model for both larvae reared in the presence and absence of dragonfly predators ( AIC Predator present: 8.43; Predator absent: 2.0; Table 3-2). For the remainder of the paper I consider only the exponential model (Fig. 3-3). Comparing across predator treatments, the estimate for maximum size (i.e., the estimated mass at very low densities) at metamorphosis was similar in the presence (0.236 g; 95% C.I.: 0.216 .257) and absence of dragonflies (0.238 g; 95% C.I.: 0.212 0.264). However, the rate at which metamorph size decreased as initial density increased was significantly higher in the absence of predators (rate of decrease [95% C.I.]: Absent: -0.018 [-0.0133 -0.0229] versus Present: -0.0073 [-0.0049 -0.0097]). As a result, the difference in the estimated size at metamorphosis between the predator present and absent treatments increased with increasing density. Thus while estimated metamorph size in the predator and no predator treatments were similar at low densities, at high densities (100 larvae tank-1) metamorphs from the predator treatment were 150% larger than those from the no predator treatment (0.13 vs. 0.05 g). I used these models for the relationship between mass at metamorphosis and initial density in the presence and absence of predators to quantify the overall effect of egg and larval-predators on metamorph mass, and partition out and compare the individual contributions of egg-predator and larval-predators carry over effects. In the absence of predator effects, (i.e., 41.3 larvae tank-1, no dragonfly) the estimated mass at metamorphosis was 0.11 g (95% CI: 0.08 0.15 g, Fig. 3-3). The combined effects of
55 both stage-specific predators (i.e., 20 larvae tank-1, dragonfly) increased metamorph mass by 91% to 0.21 g (95% CI: 0.18 0.23 g). I then partitioned this overall effect in the contributions from egg-stage and larval-stage predator sources. Egg-predator effects on larval input (i.e., 20 larvae tank-1, no dragonfly) carried over to increase mass at metamorphosis by 55% to 0.17 g (95% CI: 0.14 0.20 g, Fig. 3-3). Larval-stage predation had a similar effect; mass at metamorphosis in the presence of dragonflies was 60% greater (0.18 g; 95% CI: 0.15 0.21 g, Fig. 3-3) than in the absence of predators. Experiment 2: Size-specific Metamorph Predation Based on the experimental results above, the small metamorph size class (mass: 0.08 g SD 0.01) used in the predation trials fell within the range of sizes expected in the absence of predator effects (0.08 0.015 g). The large size class (0.20 g SD 0.01) fell within the range of size classes expected given egg predator (0.14 0.20 g), aquatic predator (0.15 0.21 g), or egg and aquatic predator effects combined (0.18 0.23 g). In the presence of spiders, large metamorphs survived significantly better (0.68 vs. 0.44) than the smaller metamorphs (F1,8 = 7.059, P = 0.029; Fig. 3-4). Discussion The effects of predators on early prey stages can carry over to alter performance in later stages, particularly by altering the strength of density-dependent processes. Field, mesocosm, and laboratory experiments with larval amphibians spanning three decades have shown that reductions in larval-stage density (independent of predator effects) can have strong positive effects on amphibian larval growth (reviewed in Skelly and Kiesecker 2001) and, in some systems, survival (Berven 1995, Van Buskirk and Smith 1991, Semlitsch and Caldwell 1982). In this study, reductions in larval input similar to
56 that observed in the field (i.e., ~50% reduction) had positive effects on larval-stage survival (5% increase), duration (18% decrease), and size at metamorphosis (55% increase). Larval predators further reduced tadpole densities (64% decrease) and larval competition. Aquatic predation may also be size selective (Alford 1999, Table 10) such that larger larvae are more likely to survive and metamorphose. Both numerical and size selective effects may have contributed to the larger metamorph sizes in the presence of dragonflies (64% increase). In combination, egg-predator effects on density and larval-stage predator effects increased size at metamorphosis by 91%. This increase in size is likely to have fitness consequences. For example, there is strong evidence that increased size at metamorphosis can increase juvenile survival and growth, size at maturity, and fecundity in amphibians (e.g., Scott 1994, Altwegg and Reyer 2003). However, these studies do not explicitly address how increased size leads to greater survival. The metamorph predation trials point to one possible mechanism larger metamorphs may be less vulnerable to size-specific metamorph predation. Thus, egg-stage effects carry over to affect larval-stage performance and both eggand larval-stage effects can carry over to affect metamorph performance. Such carry over effects point out that studies that focus on a single life stage are limited in their ability to capture important aspects of the biology of species with complex life histories (see also Vonesh and De la Cruz 2002). In general, the effects of changes in larval-stage density (through egg-predation or otherwise) were greater in the absence of larval-predators. However, density and larval predation also exhibited complex interactions, such that the consequences of being in a predator or predator-free larval habitat reversed depending on larval density. For
57 example, linear fits of larval duration to larval density in the presence and absence of predator differed in both their intercepts and slopes. Different intercepts suggest that larvae in the predator and no predator treatments differ in their minimum time to metamorphosis (i.e., larval duration in the absence of intraspecifc competition) at low densities, larvae delayed metamorphosis by nearly two weeks in the presence of predators, but at high density metamorphosed ~4 weeks sooner. This delay at low density was likely the result of predator-induced behavioral plasticity, as larvae frequently reduce foraging rates (and thus growth rates) in the presence of predators (e.g., Werner 1992, Werner and Anholt 1996). Earlier metamorphosis at high density (in the presence of larval predators) likely reflected the numerical effects of predators on prey density and competition, which more than compensated for any behaviorally-mediated reductions in growth. Density also altered how larval predators affected size at metamorphosis. The estimated upper limit for mass at metamorphosis (i.e., Table 3-1, Exponential fit, P2) was the same in the presence and absence of larval predators, but since mass at metamorphosis decreased more rapidly with increasing density in the absence of predators; at high densities metamorphs from the predator treatments were much larger than those from no predator treatments. Thus, at low densities, the effects of larval predators on surviving larvae were mostly negative (i.e., increased the length of the larval stage, no effect on metamorph size) while at high densities larval predator effects were positive (i.e., decrease larval duration and increase size at metamorphosis of survivors). In combination with other work on this system (Vonesh and Osenberg 2003, Chapter 4), we are beginning to get a more complete picture of predator-induced carry
58 over effects of across different life-stages of H. spinigularis. While in this study I have focused on the density-mediated effects of egg-stage predators (i.e., their reduction in larval input into the pond), egg-predators also induce surviving embryos to hatch early and at a smaller size. Thus, they can affect larval traits and well as larval density, and these trait-mediated effects can also carry over to alter performance in subsequent stages. We found that this trait-mediated effect (i.e., reduced initial size and age), acted to facilitate larval survival to metamorphosis (i.e., to our surprise, small larvae survived better not worse) because initially early-hatched larvae grew more rapidly through vulnerable size classes (Chapter 4). While egg-predator induced, early-hatched larvae grew faster through early size-classes (where they were vulnerable to dragonfly predation), this advantage was lost later in ontogeny and by metamorphosis early-hatched larvae were significantly smaller than larvae from undisturbed clutches. Thus, while egg-predator effects on larval density increased size at metamorphosis, their effect on larval traits decreased metamorph size. This later result is important because it points to a trade-off that could explain the maintenance of later hatching times in the field. Egg-predator induced hatching plasticity (i.e., hatching early to escape a risky egg-stage environment, Sih and Moore 1993, Warkentin 1995) comes at the cost of reduced size at metamorphosis. In addition to the potential longer term fitness consequences of small metamorph size (reverse arguments above for advantages of large size), I have also shown here that being smaller at metamorphosis can increase vulnerability to size specific metamorph predators. One potential concern regarding our assessment of size selective metamorph predation as a potential cost for hatching early is that the experiment was conducted over
59 a very short time scale (24 h) and ignores the possibility of compensatory response similar to those I observed in the larval-stage (Chapter 4). Small metamorphs might show similar patterns of compensatory growth, possibly reversing the effects of predators measured in short-term trials. However, other studies suggest this is unlikely. While growth rates in larval amphibians tend to be highly plastic, post-metamorphic amphibians generally do not compensate for small size at metamorphosis by enhancing post-metamorphic growth rates (Goater 1994, Scott 1994, Altwegg and Reyer 2003, Relyea and Hoverman 2003). For many amphibians the larval stage is thought to be primarily a growth phase (but see Werner 1986, Relyea and Hoverman 2003); thus there may be more opportunity for compensatory responses during larval-stage compared with post-metamorphic stages, where individuals are also investing in dispersal and reproduction. In addition, in contrast to some other anuran species (Werner 1986), H. spinigularis does not grow much after metamorphosis. On average, H. spinigularis larvae are likely to increase their total length by more than 500% during the larval period (~ 7 45 mm), whereas the typical metamorph (~12 mm) is only likely to increase in snout-vent length through post-metamorphic growth by ~60% in males and by ~100% in females. In fact, the largest metamorphs (~18 mm SVL) already overlapped the size range of sexually mature males. Thus, it seems unlikely that small H. spinigularis metamorphs will be able to compensate via accelerated postmetamorphic growth. In the absence of such compensatory responses, smaller size at metamorphosis due to the effects of egg predators on larval traits is likely to lead to negative consequences when encountering size-selective predators.
Table 3-1. Summary of model fits describing larval-stage survival as a function of initial larval density. Fits for the Shepherd recruitment function (Equation 1), Beverton-Holt function (B-H), and simple linear (y intercept = 0) models are given for number surviving in the presence and absence of aquatic predators (dragonfly larvae). Degrees of freedom (df), sum of squares (SS), Akaikes information criterion (AIC), parameter estimates (a, b, c), and the lower (low) and upper (high) bounds for the 95% confidence intervals for each parameter are given. Model fit Parameter estimates (95% CI) Dragonfly Model df SS AIC a low high b low high c low high Present Shepherd 13 181 92 0.39 0.17 0.60 3.2E-5 -5.8E-4 6.5E-4 2.12 -1.78 6.26 B-H 14 191 91 0.51 0.20 0.81 0.02 -0.003 0.032 Linear 15 250 95 0.19 0.12 0.26 Absent Shepherd 13 171 91 0.95 0.81 1.09 1.4E-7 -1.16E-6 1.44E-6 3.48 1.42 5.38 B-H 14 349 101 1.32 0.92 1.72 0.014 0.005 0.023 Linear 15 767 113 0.52 0.399 0.65 60
61 Table 3-2. Summary of model fits describing size at metamorphosis as a function of initial density. Fits for simple linear (y = P1x+P2) and exponential (y = P2e P1x) models are given for metamorph size in the presence and absence of aquatic predators. Degrees of freedom (df), the sum of squares (SS), the Akaikes information criterion (AIC), parameter estimates (P1, P2), and the lower (low) and upper (high) bounds for the 95% confidence intervals for each parameter are given. Model fit Parameter estimates (95% CI) Dragonfly Model df SS AIC P1 low high P2 low high Present Linear 14 0.0068 -72.78 -0.0012 -0.0015 -0.0008 0.228 0.210 0.247 Exp 14 0.0060 -74.78 -0.0073 -0.0049 -0.0097 0.236 0.216 0.257 Absent Linear 13 0.0091 -62.47 -0.0018 -0.002 -0.0014 0.0208 0.019 0.023 Exp 13 0.0052 -70.899 -0.018 -0.0229 -0.0133 0.238 0.212 0.264
62 0204060801000102030405060No. survivors No DragonfliesDragonflies 02040608010000.20.40.60.81Initial density (larvae tank-1)% Survival Figure 3-1. Relationship between H. spinigularis larval-stage survival and initial density in the presence and absence of larval-stage predators. Survivors were calculated as the number of metamorphs plus larvae remaining in the tank at the end of the experiment. A. Number of survivors as a function of initial density showing Beverton-Holt fit (Table 3-1). B. Proportion surviving as a function of initial density. A B.
63 02040608010405060708090100110Initial density (larvae tank-1)Larval development (d)No DragonfliesDragonflies 0 Figure 3-2. Relationship between the length of the larval-stage (days to metamorphosis) for H. spinigularis and initial density in the presence and absence of aquatic predators. Data points represent tank means.
64 02040608010.050.10.150.20.250.3Initial density (larvae tank-1)Metamorph mass (g)No DragonfliesDragonflies 00 Figure 3-3. Results from experiment 1 showing metamorph mass (g) at eight initial densities in the presence and absence of larval-stage predators. Fits of the exponential model are given for each predator treatment (Table 3-2). Mean metamorph size SD are given for each replicate.
65 Metamorph mass Small (0.08 g)Large (0.19 g)Metamorph survival 0.00.20.40.60.8 Figure 3-4. Proportional survival of small and large metamorph size classes in the presence of 1 Thalassius sp. spider over 24 h. SE
CHAPTER 4 CONSEQUENCES OF PREDATOR-INDUCED HATCHING PLASTICITY IN AN AFRICAN TREEFROG Introduction There is considerable evidence that prey can assess predation risk and respond to predators by changing their behavior, morphology, and life history (e.g., Lima and Dill 1990, Skelly 1992, Sih and Moore 1993, DeWitt 1998, Tollrain and Harvell 1999). Changes in the timing of metamorphosis and habitat shifts are particularly compelling because they typically involve dramatic shifts in ecology, including changes in habitat, resources, and predators. Theory developed for organisms with complex life histories predicts that the timing of transitions between two life stages should evolve in response to variation in growth and mortality rates in the two stages (Werner and Gilliam 1984, Werner 1986). Recently, this theoretical framework has been applied to examine the timing of hatching, an event that separates embryonic and larval stages (Sih and Moore 1993). For example, the eggs of some salamanders (Sih and Moore, 1993; Moore et al. 1996), anurans (Laurila et al. 2002; Schalk et al. 2002), crustaceans (Blaustein 1997), and fish (Jones et al. 2003) delay hatching in response to cues from post-hatching predators (review in Martin 1999). The postponement of hatching may allow hatchlings to reach a larger body size and more developed stage before encountering predators, potentially increasing their survival (Sih and Moore, 1993). Similarly, the eggs of anurans (Warkentin 1995, 1999ab, 2000; Vonesh 2000, Warkentin et al. 2001; Chivers et al. 66
67 2001), fish (Wedekind 2002), and arachnids (Li 2002) have been shown to hatch earlier in response to cues from egg-stage predators, potentially increasing egg-stage survival. In cases of egg predator-induced early hatching, it is thought that the timing of hatching is determined by trade-offs in egg-stage versus post-hatching mortality; e.g., hatching earlier and smaller in response to egg predators may increase egg-stage survival but at the cost of increased vulnerability to size-specific predators that prey on subsequent stages (Warkentin 1995, 1999b). However, most studies have focused on detecting a predator-induced hatching response, rather than quantifying the implications of such a response to later life stages. The few studies that have attempted to examine potential trade-offs focus on short-term effects (e.g., those that arise within 24 hrs of hatching; Sih and Moore 1993, Warkentin 1995, 1999b). Such short-term experiments ignore the potential for longer-term compensatory responses in the prey. Another limitation of previous studies of predator effects on hatching is that they have only considered trait-mediated (e.g., timing, size) effects. However, eggs consumed by predators never become larvae thus egg-stage predation also reduces larval density. Like predator effects on prey traits, this density effect may also indirectly alter predator-prey interactions in subsequent stages. Reduced larval density due to egg predation may decrease intraspecific competition, increase larval growth rates, and potentially change larval survival in the presence of larval predators (e.g., increase survival if larvae grow more rapidly through vulnerable size classes, or decrease it if predators exhibit a Type II functional response). Since the densityand trait-mediated effects of egg predators occur simultaneously, and could act in opposite directions, both of these effects need to be
68 considered when evaluating potential trade-offs associated with predator induced life history shifts (Vonesh and Osenberg 2003). In this study we evaluate the consequences of egg-stage predator effects on hatching traits and initial hatchling density in the African treefrog, Hyperolius spinigularis. First, we conducted a tank experiment in which we manipulated initial larval density (mimicking the numerical effects of egg predators), hatchling size (mimicking the effect of predators on hatchling traits), and the presence or absence of aquatic predation to examine how effects of egg-stage predators translate through the larval stage and affect metamorph characteristics. We found, in contrast to previous studies (e.g., Warkentin 1995), that small/early hatched larvae actually survived better than later hatched larvae. This surprising result motivated us to develop a mathematical model that incorporated expressions for growth and sizeand density-specific mortality parameterized from additional field experiments that explored the role of these mechanisms in determining larval survival. Through this combination of approaches we addressed the following questions: How do egg-stage predator effects on initial size and initial density affect larval survival and timing and size at metamorphosis? Are the density and size effects of predators independent? How does the relative importance of these indirect pathways change with different egg-stage predators? Finally, what mechanisms give rise to the observed patterns of survival? Materials and Methods Site and System Information Site information This research was conducted at the Amani Nature Reserve (ANR) Conservation Headquarters in the East Usambara Mountains of northeastern Tanzania (5.06 S and
69 38.37 E; Elevation; 900 m) from October 2001 August 2002. ANR includes 8380 ha of transitional lowland-montane rainforest. The site receives approximately 2000 mm rainfall each year (Hamilton and Benstead-Smith 1989), which falls primarily in two distinct rainy seasons: October November (short rains) and March June (long rains). ANR and the forests of the Usambara Mountains are among the threatened Eastern Arc Mountain biodiversity hotspots (Myers et al. 2000) noted for their high degree of diversity and endemism (Lovett and Wasser, 1993). Field measurements of H. spinigularis clutch densities, egg predation, and aquatic predator densities were made at Amani Pond (~200 m North ANR field station) unless otherwise noted below. Amani Pond is a man-made (> 50 yr old) permanent shallow pond (average depth ~ 45 cm) bordered by submontane rainforest. Pond vegetation is dominated by floating mats of milfoil (Myriophyllum spicatum) and marginal patches of emergent cattails (Typha sp). System information Hyperolius spinigularis is endemic to a few submontane rainforest localities in Tanzania and Malawi (Schitz 1999). It breeds during both annual rainy seasons by attaching its eggs to vegetation overhanging permanent or semi-permanent ponds or swamps. Upon hatching the tadpoles fall into the water. Thus, hatching can be described as an ontogenetic niche shift; animals change habitats, become mobile, gain access to external resources, and are vulnerable to a new suite of predators. Arboreal and aerial predators attack eggs, but have limited access to tadpoles, while aquatic predators prey on larvae but have no access to eggs. Mortality rates may differ before and after hatching, creating the potential for trade-offs in risk between egg and larval stages. Clutches attacked by egg-eating treefrogs Afrixalus fornasini and parasitoid flies of the genus
70 Typopsilopa hatch earlier than undisturbed clutches (Chapter 2). As a result, embryos that survive predator encounters become aquatic larvae at a smaller size and earlier developmental stage (Vonesh 2000). While treefrogs and parasitoid flies have similar effects on the timing of hatching and initial larval traits, these predators can differ considerably in their effect on initial larval density. For example, during the first month of the breeding season, egg-stage predation by Afrixalus fornasini can reduce the density of H. spinigularis tadpoles entering Amani Pond by more than 60%, while studies at several sites suggest that parasitoid flies attack many fewer clutches and have a much smaller effect on larval density (Vonesh 2000, Chapter 2). Aquatic larval-stage predators To guide our selection of an aquatic predator in the experiments described below, we estimated the relative abundance and efficacy of potential tadpole predators in Amani Pond. We estimated the relative abundance and field densities of potential tadpole predators in Amani Pond using a box sampler (0.8 m2 x 0.9 m high) made of a plastic cylindrical water tank (Simtank Ind., Ltd.) with its top and bottom removed. Sampling locations were chosen at random from a sampling grid. We deployed the box sampler by throwing it into the area to be sampled, pressing it through any vegetation into the soft bottom substrate, removing all vegetation by hand, carefully searching the vegetation for all macroinvertebrates, fishes, snakes, and larval and adult frogs, and then using dip nets to remove animals from the water column. Dip netting continued until 15 min passed without a new capture. We sampled 10 plots on 18 November 2001 and 12 plots on 9 January and 20 February 2002. This sampling window began at the end of the late 2001 wet season and continued into the subsequent dry season. Since H. spinigularis breed during the rainy seasons and then spend ~ 3 months as aquatic larvae this sampling
71 window coincides with the time that larvae are vulnerable to aquatic predators (a second window occurs in May July). A more detailed discussion of H. spinigularis breeding phenology can be found in Chapter 2. To evaluate the relative efficacy of the common tadpole predators, we conducted short term feeding trials with recently hatched H. spinigularis larvae and the five most numerous types determined above. These trials were conducted in small 11-L plastic containers containing 10-L of pond water and 20 dried leaves (each ~40 cm2) for cover. Each trial contained 10 anuran larvae (8.73 SD 0.26 mm TL) and 1 predator. Trials were conducted with five predators: 1) dragonfly larvae (T. basilaris; 14.8 SD 1.5 mm TL); 2) damselfly larvae (Zygoptera, unidentified; 14.6 SD 0.54 mm TL); 3) adult diving beetles (Dytiscus sp.; 24 SD 2.1 mm TL); water scorpions (cf. Nepa sp.; 34.6 SD 1.3 mm TL); and fish (Gambusia sp.; 26.1 SD 2.3 snout-tail base). All predators were collected from Amani Pond and held without food 48 hr prior to experimental trails. Tadpoles were introduced first into containers followed 12 hr later by predators to start the experiment. After 24 hrs we counted the number of surviving tadpoles. Each predator treatment was replicated six times. Based on the results from the aquatic plot sampling and the predation trials we used larval dragonflies (Trapeziostigman basilaris) as our aquatic predator in our experiments. Consequences of Egg-Stage Predator Effects for Growth and Survival to Metamorphosis Experimental design: density and size/age effects To evaluate the consequences of densityand size-mediated effects of egg predators, we conducted a 3 x 2 x 2 factorial experiment with a randomized block design in which we manipulated initial density, initial size, and presence of ambient densities of
72 aquatic predators (+PAQ and PAQ; Table 4-1). Each of these 12 treatments was replicated four times. The three levels of the initial density factor correspond to expected densities given no egg-stage predation and two different intensities of egg predation (N-EP, and N+EP1, N+EP2; Table 4-1). The two levels of the initial size factor represent larval hatching sizes in the presence and absence of egg-stage predation (S-EP, S+EP; Table 4-1). This experiment was performed in 48 white plastic water storage tanks (1.2 x 0.8 x 0.4 m; Chemi and Cotex Industries, Ltd. TZ.) arranged in a 3 x 16 array in a partially shaded forest clearing. Pools were initially filled to a volume of 220 L with Amani Pond water filtered through a 0.3 mm mesh screen and were immediately covered with tight fitting lids of fiberglass window screening to prevent unwanted colonization by insects and frogs. Screened drain-holes prevented overflow during heavy rains. Tanks were buried to a depth of 15 cm in shallow trenches to provide a more stable thermal regime. To each tank we added 200 g of washed and dried pond litter, 10 g of commercial fish food (Hikari brand, Kyorin Co., Ltd; Crude protein: min 32%, Crude fiber: max 5%, Moisture: max 10%, Ash: max 12%; Main ingredients: fish meal, wheat flour, wheat germ meal), 300 g (~ 25, 40 cm long stems) of washed free-floating macrophytes (Myriophyllum spicatum), and a 0.5 L inoculum of a stratified pond water sample collected from Amani Pond using a 80 m plankton net. All components were assigned to pools randomly. Tanks were filled with water on 19 March, and litter and other components were added on 22 March. Tanks were allowed to equilibrate for approximately two weeks before the start of the experiment. Treatments were randomly assigned to tanks, and the experiment was initiated on 5 April with the addition of H. spinigularis and T. basilaris larvae. Growth and survivorship were estimated monthly by
73 dip-netting out all larvae and taking digital photographs of the larvae from each tank. The total lengths (TL) of all larvae were measured using the open source digital image analysis software ImageJ (http://rsb.info.nih.gov/ij/). Measurements were calibrated for each image by referencing a ruler in each photograph. Damaged larvae or larvae hidden behind other larvae were not included in analyses. Metamorphosed dragonflies were replaced weekly to maintain predator density. After 16 weeks, tanks were emptied and any remaining larvae counted. Survival was estimated as the number of metamorphs (removed weekly) plus the number of surviving larvae (proportion of survivors reaching metamorphosis: 0.75 0.09 95% CI). We used general linear mixed effect models ( Pinheiro and Bates 2000, Venables and Ripley 2002) to examine effects of initial larval density, initial size, and presence of aquatic predators on three response variables: 1) the proportion of larvae surviving to the end of the experiment; 2) the proportion of survivors at the end of the experiment to reach metamorphosis; and 3) mass at metamorphosis. Initial size, density, and predator presence/absence were considered fixed factors; block was modeled as a random factor. Proportional data were arcsine square-root transformed prior to analysis. All analyses were performed in R version 1.7.0, an open source language and environment for statistical computing and graphics (Ihaka and Gentleman 1996; http://www.r-project.org/). Establishing the egg-predation mediated density effect: N-EP and N+EP1, N+EP2 Initial tadpole densities in the experimental trials were based upon field estimates of average clutch size, clutch density, the number of clutches attacked by each predator type, and the average survivorship of undisturbed versus attacked clutches. These parameters were estimated by monitoring reproductive effort and clutch survival along
74 two randomly located 3 X 30 m transects in Amani Pond between October 2001 and the start of the experiment in April 2002. Clutch production along each transect was monitored at 2 3 d intervals, new clutches were marked and followed through to hatching, and clutch survivorship determined by placing capture cups beneath clutches as they neared hatching (Hayes 1983, Warkentin 1995). A more detailed account of our clutch monitoring methodology is provided in Chapter 2. The density of tadpoles in the absence of egg predators (N-EP) was estimated from the mean clutch density (; clutches m-2), the average clutch size (; egg clutch-1), and the mean proportional survival to hatching of embryos in clutches that are not attacked [E; thus, N-EP = ()()(E)]. The density of tadpoles in the presence of egg predators (N+EP1,2) was estimated by including the proportion of the total clutches attacked by each predator (EP1,2) and the expected survival of embryos in clutches attacked by that predator (EP1,2). For example, initial density given the effects of egg predator 1 (N+EP1) was the larval contribution of clutches that are not attacked [()(1EP1)()(E)], plus the contribution of clutches that are attacked [()(EP1)()(EP1)]. Standard deviations for initial densities were calculated from the standard deviations of each of the estimated parameters, using standard rules for error propagation (Lyons 1991). Field estimates in the absence of egg predation (N-EP) were used to establish treatments 1, 2, 7, and 8 (Table 4-1). Estimates of initial densities in the presence of Afrixilus fornasini (N+EP1) were used to establish treatments 3, 4, 9, and 10 (Table 4-1). Estimates of initial densities in the presence of Typopsilopa flies (N+EP2) were used to establish treatments 5, 6, 11, and 12 (Table 4-1).
75 Hyperolius spinigularis breeding activity during the long rainy season 2002 began on 4 March. From the start of breeding to the start of the experiment, Hyperolius spinigularis clutch densities () in Amani Pond averaged 0.49 SD 0.27 clutches m-2 10-d-1. Hyperolius spinigularis clutch size () for the same period averaged 83.94 SD 21.9 embryos clutch-1 (n = 123), and the average proportional survival of embryos in clutches not attacked by egg-stage predators (E) was 0.91 SD 0.20 (n = 29). Thus, field estimates of larval density in the absence of predators were 37.43 SD 24.18 larvae m-2; in our experimental tanks N-EP densities were set at 38.8 larvae m-2 (i.e., 35 larvae tank-1). Afrixalus fornasini are known predators of arboreal frog eggs (Drewes and Altig 1996), and at this site attacked (EP1) an average of 0.73 SD 0.18 clutches 10-d-1. Survival of embryos in the attacked clutches (EP1) averaged 0.14 SD 0.19 (n = 86). Thus, the field estimate for initial larval density in the presence of A. fornasini predation was 14.1 SD 11.27 larvae m-2. In our experimental tanks N+EP1 densities were 11.11 larvae m-2 (10 larvae tank-1). Ephydrid flies of the genus Typopsilopa prey upon the arboreal eggs of several East African hyperoliid species (Vonesh 2000) including H. spinigularis in Amani (Chapter 2). However, early in the 2002 breeding season, when this experiment was being set-up, egg-predation by Typopsilopa was rare (Chapter 2) and only increased late in the breeding season (Chapter 2). Thus, experimental densities reflecting Typopsilopa predation (N+EP2) were based on predation estimates for four Hyperolius species from two ponds in Kibale National Park (Vonesh 2000) and H. spinigularis in Amani in late spring 2000 (Vonesh, unpublished data). Using these data, Typopsilopa flies attacked (EP1) an average of 0.27 SD 0.14 (n = 9 site x species combinations) of the available clutches.
76 Proportional survival of embryos in attacked H. spinigularis clutches (EP1) averaged 0.21 SD 0.3 (n = 10 clutches from spring 2002). Thus, field estimates for initial larval density in the presence of fly egg-predation were 29.3 SD 18.6 larvae m-2. In our experimental tanks N+EP2 densities were 27.8 larvae m-2 (25 larvae tank-1). Establishing predator-induced size effect: S-EP and S+EP In addition to the mortality caused by egg-stage predators, exposure to egg predators causes surviving embryos to hatch earlier and at a smaller initial size. To simulate this sublethal effect of egg-stage predation we collected fresh (< Gosner stage 10 (Gosner 1960); ~ 1 12 hrs old) H. spinigularis clutches from the field on two dates: 10 clutches each on 26 and 29 March (3 d age difference). On 4 April, clutches from both dates were submerged in water and gently shaken to induce hatching. Clutches within each date were pooled and larvae randomly assigned to density and predator treatments from respective pooled groups. To reduce the effect of initial handling on the experimental outcome, larvae were held overnight and dead larvae were replaced (< 1% initial mortality). Larvae were then digitally photographed to establish initial size differences among treatments. Larvae from eggs collected on 26 March (~10 d old; = 9.33 mm TL 0.08 95% CI) were used in S-EP treatments and larvae from eggs collected on 29 March (~7 d old; = 7.14 mm TL 0.06 95% CI) were used in S+EP treatments. These differences reflect a decrease in age and total length (TL) at hatching ( -3 d; -2.19 mm TL 0.11 95% CI, 23% decrease in TL) consistent with that induced by A. fornasini ( -2.6 d 1.1 95% CI; -2.17 mm TL 0.77 95% CI, 27% decrease in TL) and ephydrid flies ( -4.0 d 3.2 95% CI; -1.5 mm TL 1.1 95% CI, 19% decrease in TL) (Chapter 1). Within a given level of initial size there were no significant differences in
77 total length among different levels of the density (F2,36 = 1.1, P = 0.344) or predator (F1,36 = 1.3, P = 0.255) factors. Integrating Patterns of Larval Growth, and Densityand Size-Specific Risk to Predict Larval Mortality To better understand the mechanisms that determined the survival of H. spinigularis larvae in the presence of predators, we developed a simulation model that incorporates differences in initial size and growth rates between early and late hatched larvae observed in the density x size manipulation experiment above, as well as sizeand density-specific larval predation risk to dragonflies (as determined by two additional experiments discussed below). Larval growth We focused on growth rates during the first 30 days, as this is the period when most growth and mortality occurs, and was prior to any metamorphosis. Alford and Jackson (1993) argued that the exponential growth model provides the most accurate description of growth prior to metamorphosis in amphibian larvae, so we modeled growth as: tktijieSS (1) where St is larval total length (mm) at age t days, Si is total length at hatching, and kij is the exponential growth rate constant for the ith level of initial size and the jth level of the initial density factor. We estimated kij over the first 30 days via linear regression of log transformed size data on time. Effect of density on risk: the functional response To estimate the shape of the functional response we conducted an experiment in which we varied larval density in the presence and absence (a control) of dragonfly larvae. This experiment was conducted in 32 300-L plastic tanks prepared in early
78 November 2001 in the same manner as the density x size manipulation experiment above. Treatments consisted of eight initial densities: 5, 10, 15, 20, 30, 50, 75, 100 larvae tank-1. Each treatment was replicated twice. Hyperolius spinigularis larvae used in the experiment came from 18 clutches collected upstream of Amani Pond that hatched naturally between 15 20 November. On 22 November, all hatchlings were pooled and randomly assigned treatments. To reduce the effect of initial handling on the experimental outcome, larvae were held overnight and dead larvae were replaced (< 1% initial mortality). Larvae were then digitally photographed to establish initial size (9.69 mm TL SD 0.23). Three late instar T. basilaris larvae (TL: 18.9 mm SD 2.26) were added to each tank. Tanks were checked every other day and metamorphosed dragonflies were replaced to maintain predator densities (none emerged/were replaced). Survivorship of frog larvae was estimated after 14 d by dip netting out all H. spinigularis and dragonfly larvae from tanks. Mortality in the absence of predators was very low, and similar across densities (0.04 SD 0.03). We assumed that risk could be modeled with a type II functional response. However, the classic form of the type II functional response makes no allowance for prey depletion, and hence is only appropriate for studies of very short duration, experiments where prey are replaced as they are consumed, or in situations where predators search systematically so that they do not re-encounter previously exploited areas (Elliot 2003). Because our experimental design did not fit these criteria, we estimated values for the attack rate (D) and handling time (HD) using the random predator equation of Rogers (1972, see also Juliano 1993). The random predator equation is equivalent to a type II
79 functional response that incorporates the effects of prey depletion, and is solved by integrating the foraging rate over time as prey density declines: )eN(1NPT)H(NDaDa (2) where N is the number of prey eaten in T days, N is the initial prey density, D is the attack rate (i.e., the instantaneous rate of discovering prey by one predator); and HD is the handling time. We assume that the actual number attacked, N, for initial density N follows a binomial distribution, N~ binom (N, ), where is the probability of an individual prey being killed during the course of the experiment ( = Na/N). We then obtained estimates for D and HD that satisfy equation 2 and maximize the binomial likelihood. In our simulation model, where population densities are tracked daily and depletion therefore does not need to be modeled explicitly, we used these parameter estimates for D and HD in a type II functional response (Holling, 1959): a a a DDDNH1NK (3) where K is the expected predation rate (i.e., number of prey eaten pred-1 day-1). Ideally, this experiment would have been conducted for a fixed larval size. However, since larvae grew during the experiment, our estimated functional response is perhaps best viewed as an approximation of the functional response for larval size midway through the interval (~TL: 12.8 mm). Effect of larval size on predation risk To quantify how size and age influence the vulnerability of H. spinigularis larvae, we exposed 5 larval size/age classes to dragonfly larvae and quantified survival. Each treatment was replicated 3 times. The experiment was conducted in 35-L plastic tubs
80 (diameter = 0.32 cm2) arrayed near the larger tanks above. Each tank was filled with 25-L filtered pond water and received four 40 cm stems of floating M. spicatum, 200 g of dried pond litter, and 20 g of commercial fish food (= ad lib food). Tanks were covered with mosquito netting. Hyperolius spinigularis larvae were obtained from clutches collected on different dates, yielding 5 non-overlapping size/age classes. The smallest, youngest class was induced to hatch from 3 clutches collected from sites near Amani on 5 June (Size/Age Class 1: 0 days post-hatching, TL: 8.89 mm SD 0.04). Larvae used in the remaining treatments were reared from 3 4 near-to-hatching clutches collected on different dates but reared under similar conditions: 28 May (Size/Age Class 2: 7 days post-hatching, TL: 11.55 mm SD 0.28), 29 April (Size Class 3: 5 weeks post-hatching, TL: 20.42 mm SD 0.91), 21 April (Size Class 4: 6 weeks post-hatching, TL: 25.34 mm SD 0.75), and 29 March (Size Class 5: 9 weeks post-hatching, TL: 37.76 mm SD 1.18). Late instar T. basilaris nymphs (TL: 13.34 mm SD 1.23), were collected from the field on 2 June, fed one H. spinigularis larvae, and then held without food until the beginning of the experiment. Dragonflies (2 tank-1) were randomly assigned to the predator treatment tanks on 4 June and the experiment began on 5 June with the addition of H. spinigularis larvae. Emerged dragonflies were replaced daily to maintain predator densities (only 1 was replaced). After 3 days all surviving larvae were counted. We estimated the functional form of size specific mortality using a phenomenological model that allows for a peak in prey vulnerability at intermediate sizes:
81 S))((S))((s1 e e (4) where s is the size-specific predation rate, S is prey size, and , and are fitted constants. Note that this expression does not explicitly account for predator saturation with increasing prey density (i.e., as in equation 2). We assumed s for size S follows a binomial distribution, s ~ binom (N, s) and obtained estimates for , and that maximized the binomial likelihood. Combining size and density specific risk We next had to combine our estimate of the size-specific predation rate (which was measured at a single density) with our description of the functional response (which quantified the effect of prey density for a single prey size). We developed a model that predicted the attack rate for any combination of density and size by using the functional response to recover the attack rate without the effects of prey saturation from the predation rate we estimated from the size-manipulation experiment, s. We did this by setting K in equation 3 equal to s(N), setting prey density to Ns (i.e., the prey density used in the size-specific predation experiment) and then solving algebraically for the attack rate (now a function of size and density, so we now call it SD): DsssHN1SD (5) We assume that the handling time (HD) estimated from the functional response holds across all larval sizes. The density and size specific predation rate can then be estimated using equation 3 by replacing D with SD, providing us with a hypothesis regarding how the predation rate behaves across the range of relevant density and size parameter space.
82 Simulating larval growth and mortality We then simulated larval growth and survival in daily increments (i.e., we ignore effects of depletion and growth within a day) over the duration of the density x size manipulation experiment (112 d). Simulations were run for each experimental density and for two initial sizes (reflecting the mean size for either early or later hatched larvae). In each time step, larvae grew exponentially at density and initial size specific rates, and were killed by predators according to size and density specific predation rates. Larvae that were killed were decremented from the density at the start of the next time step. We repeated this process until 112 d or all larvae reached 40 mm TL, at which point we assumed they are no longer vulnerable to predation. We calculated the total number killed over the experiment as the sum of the number killed in each time step. Error estimates for model predictions were obtained via Monte Carlo uncertainty analysis (e.g., Caswell, 2001). The distribution for total number killed at the end of the experiment for each density was generated according to the following steps: 1) we specified the joint probability distributions of the estimates of the parameter controlling growth (equation 1), and size(equation 4) and density-specific mortality (equation 2) from the variance-covariance matrix of the maximum likelihood parameter estimates; 2) we drew random samples from the appropriate distribution for each parameter and calculated the total number killed; 3) we repeated this process (for a given set of parameter estimates) for 100 iterations of the simulation and estimated the mean of the Monte Carlo distribution of the total number killed; 4) we repeated this process to obtain 100 estimates of the models mean prediction (i.e., using 100 different parameter values, thus incorporating uncertainty in parameter estimation; sampling uncertainty for a set of parameters was greatly reduced via step 3).
83 Results Aquatic Predators Field sampling and experimental feeding trials suggest that libellulid dragonfly larvae are important aquatic predators of Hyperolius larvae in Amani Pond. Libellulid nymphs were found at higher densities than any other aquatic predator (Fig. 4-1a; 3.16 SD 4.75 larvae m-2). In addition, libellulid nymphs were the most effective of five abundant aquatic predators in 48-hr predation trials (F4,25 = 3.26 P = 0.028; Fig.4-1b). Based upon this evidence, we used libellulid dragonfly larvae (Trapezostigma basilaris) as the aquatic predator (+Paq) in treatments 1 6 at densities of 3 larvae tank-1 (3.13 m-2). Treatments 7 12 received no dragonfly larvae. We used late instar dragonfly larvae (Total Length (TL): 18.9 mm SD 2.54). Larval Survival Aquatic predators had a significant negative effect on larval survival (Table 4-2, Fig. 4-2a). Whereas survival in the predator free treatments was high (92%) and similar across density (F2,18=0.09, P=0.915) and size factors (F1,18=0.224; P= 0.642), predators reduced average survival by 44% (i.e., to 52%). Furthermore, the negative effect of predators on survival was greater for the initially larger/older larvae (i.e., initially smaller larvae survived better) and (marginally) for larvae at higher densities. Across density levels, dragonflies had a 50% greater effect on larger/older larvae compared with early hatched smaller larvae. Across initial size treatments, the average effect of dragonflies at initial densities of 25 and 35 was more than twice their effect at 10 larvae per tank; i.e., dragonflies reduced survival by ~53% on average at higher densities compared to only 23% at 10 larvae per tank. The effects of dragonflies at 25 and 35 larvae per tank were
84 similar. Most mortality in the predator treatments for both larger/older (82%) and smaller/younger (88%) occurred during the first 30 days of the experiment. Reductions in larval density and hatchling size that mimicked egg predation by treefrogs increased larval survival to a similar degree (Vonesh and Osenberg 2003). Reductions in initial density due to treefrog predation (we mimicked a ~70% reduction) resulted in a 110% increase in survival and reductions in hatchling size/age resulted in a 90% increase in average larval survival. In contrast, size-mediated effects of Typopsilopa sp. predation had the greatest effect on larval survival. Reductions in initial density due to fly predation (we mimicked ~30% reduction) had almost no effect on larval survival (< 3% increase), while the effect of reductions in hatchling size/age due to fly predation were the same as those for Afrixalus fornasini (i.e., a 90% increase in larval survival). Proportion of Surviving Larvae that Reach Metamorphosis Reductions in initial larval density (mimicking egg-stage predation) had a positive effect on larval development (Table 4-2, Fig. 4-2b). Across size and predator treatments, nearly all survivors metamorphosed from the lowest density tanks (96%), while only a little more than half (55%) metamorphosed from the highest density treatments on average. However, this difference was reduced in the presence of aquatic predators. In the absence of predators, decreasing initial density from 35 to 10 larvae tank-1 increased the proportion of survivors to metamorphosis by 156%. In comparison, in the presence of predators the same decrease in density resulted in only a 31% increase in the proportion of metamorphs (Fig. 4-2b). There was no effect of initial hatchling size on larval development the same proportion of small/early-hatched and large/later-hatched larvae had reached metamorphosis by the end of the experiment.
85 The densityand size-mediated effects of A. fornasini egg-stage predation on development were similar in magnitude but opposite in direction. Reductions in initial density due to treefrog predation resulted a 23% increase in the proportion of survivors to reach metamorphosis in the predator treatments, while smaller size/age at hatching tended to reduced the proportion of metamorphs by 20% (though this difference was not statistically significant, P = 0.23). Reductions in initial density due to fly predation had no effect on the proportion of metamorphs in the predator treatments (< 3% increase), while the effect of reductions in hatchling size/age due to fly predation were the same as those for treefrog predators. Mass at Metamorphosis Larval density, aquatic predators, and initial size all affected mass at metamorphosis (Table 4-1, Fig. 4-2c). Metamorphs from the lowest density treatments had 50% greater mass on average than metamorphs from the highest density treatments. Metamorphs from predator treatments averaged 48% greater mass than metamorphs from no predator treatments. In addition, metamorphs that were initially larger/older at hatching had 13% greater mass compared to initially smaller (and younger) hatchlings. As a result there was a considerable range from 0.08 to 0.22 g in mean mass at metamorphosis among treatments. The densityand size-mediated effects of A. fornasini egg-stage predation on metamorph mass were asymmetrical in size and direction. Reductions in density mimicking treefrog predation increased metamorph mass by 30%, while reductions in size decreased mass by 13%. Thus, we expect a net gain in mass of ~17% (0.028 g) for density and size effects combined (assuming they are independent). Indeed, metamorph size from the treatment with both the density and size effects of treefrog predation were
86 13% larger on average than metamorphs from the predator treatments with no effect of egg-stage predators (Fig. 4-2c). In comparison, densityand size-mediated effects of fly egg-stage predation on metamorph mass also acted in opposite directions but were similar in magnitude. Reductions in initial density due to fly predation increased average mass 11%, while the effect of reductions in hatchling size/age reduced average mass by 13%. Growth Rates Over the first 30 days of larval development, early hatched smaller larvae (mimicking predator-induced effects on hatching traits) exhibited higher growth rates ( = 0.0393 SD 0.005 mm d-1) than later hatched larvae ( = 0.0296 SD 0.003 mm d-1; Table 4-3, Fig. 4-3a). Furthermore, the difference between the growth rates of these two initial size classes increased as initial density increased. At low densities (10 larvae tank-1) growth rates for early hatched larvae were 18% higher than later hatched larvae, while at highest densities (35 larvae tank-1) early hatched growth rates were 48% higher on average (Table 4-3, Fig. 4-2a). Density and Size-Specific Risk Due to the combined effects of encounter rates and handling time, predation risk from the dragonfly larvae, T. basilaris, decreased with density and was a hump-shaped function of size (Table 4-3,. 4-3b,c,d). Vulnerability to dragonflies was initially low, increased until ~13 mm TL, and then declined to zero as larvae approached 25 mm TL (Table 4-3, Fig. 4-3c). We combined the expressions for sizeand density-specific risk to generate a predation risk surface that described predation rates as a function of both larval size and density (Fig. 4-3d).
87 Comparison of Observed versus Simulated Larval Survival We examined the contribution of the compensatory growth we observed in early hatched larvae (see above) to survival by implementing a simulation with two growth responses; 1) we averaged small and large growth rates for a given initial density (i.e., removed the compensatory response), and 2) we used size and density specific estimates for growth (i.e., included compensatory response). If compensatory growth was responsible for the reduced risk of small larvae, then only the second simulation should produce patterns similar to those observed. Under both of these conditions the simulation predicted similar survival across densities as we observed in our experiment (Fig. 4-4a-c, Proportional survival ( SD) Experiment: 0.52 0.18. Model: No compensatory response: 0.49 0.022; Compensatory response: 0.49 0.05). In the absence of the compensatory growth response, the simulation predicted similar survival for early and late hatched tadpoles (late hatched larvae survived ~3% better, Fig. 4-4b). When the compensatory growth response of early-hatched larvae was included in the simulation this pattern reversed itself early-hatched larvae survived better than late hatchers. This survival difference increased with increasing density: early hatched larvae survived 7%, 15%, and 30% better than later hatched larvae as initial densities increased from our low to medium to high, respectively (Table 4-3, Fig. 4-4c). While this qualitatively matches the pattern observed in the experimental results (i.e., small larvae survived better, particularly at high densities), in the experimental results the difference in survival between these initial size classes was considerably larger. In addition, the model also appears to underestimate the mean survival of both initial size classes at low densities.
88 Discussion Previous empirical work on predator-induced hatching found that early hatching increased mortality during the larval stage (Warkentin 1995). Theoretical studies support this result: predator-specific defenses (e.g., predator-induced early hatching to avoid egg-predators) should result in increased vulnerability to other predators (e.g., aquatic stage predators) (Matsuda et al. 1993, 1994, 1996). This trade-off maintains the complex life-history in the absence of the trade-off, larvae should hatch earlier and earlier, eventually eliminating the egg-stage from the life history. Interestingly, we found no evidence of a predation risk trade-off across egg and larval stages in our system. While early-hatched/smaller and later-hatched/larger larvae survived equally well in the absence of predators, early hatchlings survived better in the presence of aquatic predators (Fig. 4-1a). If hatching early is a successful strategy for evading egg-predators and there is no trade-off in embryonic versus larval predation risk or even a survival benefit why don't larvae hatch early all the time? Other costs of early hatching may arise later in the life history. For example, hatching early could delay larval development and lead to longer exposure to larval predators. However, there was no significant effect of early hatching on larval development rate, as measured by the proportion of survivors that reached metamorphosis (Fig. 4-1b). There was, however, a significant effect of size at hatching on size at metamorphosis (Fig. 4-2c). Metamorphs from early-hatched larvae were 13% smaller than those from late-hatched larvae. Smaller size at metamorphosis can reduce post-metamorphic survival and growth and lead to smaller size at sexual maturity, and hence lower reproductive success (e.g., Smith 1987, Berven 1990, Altwegg and Reyer 2003). Thus, a trade-off between egg survival and the costs and benefits
89 associated with metamorph size may help maintain a balance between early and late hatching. The fact that small/early hatched larvae survived better than large/late larvae was a surprise to us. We expected later-hatched larvae, by virtue of their greater size, would escape relatively more larval mortality. Indeed, a number of previous studies indicate that larval anuran risk to odonate predators decreases with increased tadpole size (summarized in Alford 1999, Table 10). However, in our system, risk to libellulid predation was greatest at intermediate size classes with size refugia for small and large tadpoles. This pattern may reflect the combination of predator-induced effects on hatching size and developmental traits in our study. The initial increase in vulnerability with size/age is likely driven by changes in activity associated with the shift from reliance upon yolk stores to active foraging (i.e., a developmental effect), which increases predator encounter rates, while the subsequent decline in risk is likely driven by the size effects observed in previous studies. Indeed, greatest risk at intermediate size classes is expected when encounter rates increase and capture probabilities decrease with increasing size (Osenberg and Mittelbach 1989). However, a hump-shaped risk function alone is insufficient to explain the better survival of early-hatched tadpoles. Both early and late hatched larvae drop into the aquatic environment at small sizes before risk to dragonfly predation peaks. Thus, both early and late hatchers must grow through the vulnerable size classes. If they grow at the same rate, small/early hatched larvae should still suffer higher mortality because large/late hatched larvae will maintain their head start and spend less time in vulnerable size classes. This result is confirmed in simulation results when early and late hatch
90 larvae have equal growth rates, late-hatched larvae survived slightly better (Fig. 4-4b). However, early-hatched larvae in the tank experiment exhibited compensatory growth, growing more rapidly during the first thirty days of the larval-stage than later hatched larvae. This difference in growth tended to be greater at higher densities (Table 4-2, Fig. 4-3a). If increased growth rates of early-hatched larvae enabled them to grow more quickly than late-hatched larvae through the sizes vulnerable to libellulid predation, the survival benefit of compensatory growth may swamp the negative consequences of hatching early. As a result, smaller/early-hatched larvae would survive better than late-hatched larvae. Indeed, the simulation results suggest that these mechanisms could generate the pattern of survival we observed in the tank experiment. When we include the compensatory growth response of early hatched larvae in the model, we find that early hatched larvae exhibit higher survival than late hatched larvae and that the survival difference between these size classes increases with increasing density (Fig. 4-4c). Recent theoretical studies have highlighted that the strength of trait-mediated indirect effects of predators is sensitive to the timing of experimental manipulations and the length of observation (Luttbeg et al. 2003). Indeed, differences in the time scale of experimental manipulations may explain the discrepancy between our results and previous studies. For example, Warkentin (1995) found that hatching early increased vulnerability to aquatic predators, while we observed that early-hatched larvae survive better in the presence of predators. However, Warkentins (1995) predation trials focused on the first 24-hr post-hatching. In our system it appears that early-hatched larvae survive better because they grow faster than later-hatched larvae through vulnerable size classes. This type of a response to predators was not possible in Warkentins (1995)
91 studies due to their short duration. However, in a longer study without predators, Warkentin (1999a) observed faster growth rates as well as more rapid onset of feeding and development of feeding structures in predator-induced early versus unexposed, later hatched red-eyed treefrog larvae, and noted that this could yield an advantage to early hatchers. Thus, the longer-term survival consequences of early hatching may swamp the early survival costs Warkentin (1995) observed for small larvae, yielding no net difference (or even a reversal) in the larval survival of earlyand late-hatched larvae. As a result, it is not clear if there is a trade-off between early hatching and larval survival. In our study system there were no short-term costs (because of the hump-shaped survival function), and thus hatching early only had beneficial effects on larval survival. Compensatory growth in larval anurans has been reported in response to poor conditions or stress early in ontogeny associated with low resources (Alford and Harris 1988) and decreased pH (Rasanen et al 2002) and predator-induced early hatching (Warkentin 1999a, this study). Indeed, given that delayed larval development or reduced metamorph size can have potential long-term fitness consequences, it is to be expected that selection will favor compensatory strategies, provided that compensation in and of itself is not too costly. However, a growing body of studies, from diverse taxa, show that such compensatory responses are costly, and that these costs are frequently manifested much later in ontogeny (reviewed in Metcalfe and Monaghan 2001). For larval anurans potential costs of compensatory growth include elevated predation rates (due to increased foraging activity; Anholt and Werner 1998) and delayed timing of metamorphosis (Downie and Weir 1997). In our study, reduced mass at metamorphosis may represent a cost of compensatory growth for early-hatched tadpoles. While small/early-hatched
92 larvae grew more rapidly than larger/later-hatched larvae through the first month (when they were vulnerable to predation) this trend gradually reversed itself later in larval ontogeny. Thus, while early-hatched larvae gained a survival benefit from rapid compensatory growth early in larval ontogeny it may have come at the cost of later growth/mass at metamorphosis. While the simulation results confirm that the combination of size-selective predation and compensatory growth mechanisms can generate survival patterns consistent with the experimental data, there are other mechanisms that could explain or contribute to greater survival in early hatched tadpoles that we were unable to evaluate. For example, morphological or behavioral differences between earlyversus late-hatched tadpoles have been documented in other systems (e.g., Warkentin 1999, Laurila et al. 2001), and may result in differences in their relative vulnerability to aquatic predation (e.g., Van Buskirk and McCollum 1999). In addition, the simulation was relatively poor at predicting the proportion of larvae that survived at low densities. Data from the predator-present treatments show that larvae at low densities tended to survive better than larvae from higher density treatments, while the simulation results indicate that the proportion surviving is similar or slightly increasing with increasing density. The pattern observed in the experiment is consistent with a sigmoidal functional response (i.e., Type 3; Holling 1959), which could arise, for example, if predators switched to alternative prey (e.g., small invertebrates) at low densities. Our study highlights that the consequences of predator-induced hatching plasticity need to be examined within the context of the numerical effects of predators. For example, in our study both the numerical and size/age effects of egg-predators acted to
93 enhance larval survival thus, considering only the predator effects on hatchling traits would have led us to underestimate the consequences of egg predation on survival. Furthermore, our study highlights the importance of considering the potential for compensatory responses in the prey when evaluating life history trade-offs arising from predator effects on prey traits. In our system, prey compensatory responses reversed any initial negative effects of egg predators on hatchling size/age. For prey with complex life cycles, ignoring prey compensatory responses could lead to misidentifying the trade-offs that maintain the timing of key life history transitions.
94 Table 4-1. Experimental design. N-EP is the density of Hyperolius spinigularis larvae entering the aquatic habitat when no egg-predation occurs (i.e., highest density treatment); N+AF is the density of larvae entering the aquatic habitat with ambient Afrixalus fornasini egg-stage predation; N+TY is the density of larvae entering the aquatic habitat given ambient Typopsilopa fly egg-stage predation. S-EP is the average body size of tadpoles at hatching when not exposed to the sublethal effects of egg-predation (i.e., bigger, same for both egg predators); S+EP is the size at hatching of tadpoles surviving an egg predator encounter (i.e., smaller); and + PAQ is the presence of aquatic predators (libellulid dragonfly nymphs) at ambient field densities.
95 Table 4-2. Results of analysis for effects of density, predator, and size treatments on Hyperolius spinigularis larval survival, proportion of survivors to metamorphose, and metamorph mass Response/Factor df F P Survival Density 2,33 3.73 0.035 Predator 1,33 86.69 <0.001 Size 1,33 2.64 0.11 Density*Predator 2,33 2.87 0.07 Density*Size 2,33 0.52 0.60 Predator*Size 1,33 4.91 0.03 Density*Predator*Size 2,33 0.27 0.77 Proportion of survivors that metamorphosed Density 2,33 10.99 0.0002 Predator 1,33 1.97 0.17 Size 1,33 1.50 0.23 Density*Predator 2,33 2.60 0.09 Density*Size 2,33 0.13 0.88 Predator*Size 1,33 0.74 0.40 Density*Predator*Size 2,33 0.57 0.57 Mass at metamorphosis Density 2,33 23.67 <0.001 Predator 1,33 60.93 <0.001 Size 1,33 4.32 0.046 Density*Predator 2,33 1.90 0.17 Density*Size 2,33 0.21 0.81 Predator*Size 1,33 0.53 0.47 Density*Predator*Size 2,33 0.31 0.74
96 Table 4-3. Model parameter estimates (and 95% confidence limits) decribing larval growth, the functional response of T. basilaris larvae preying upon recently hatched H. spinigularis larvae, and size-specific predation risk of H. spinigularis larvae to late instar T. basilaris. 95% Confidence Limits Parameters Estimate Low High Growth Initial size Small (Sis) 7.14 mm 7.08 7.2 Growth rate Small 10 (KS10) 0.039 0.026 0.051 Growth rate Small 25 (KS25) 0.039 0.036 0.041 Growth rate Small 35 (KS35) 0.040 0.035 0.045 Initial size Large (SiL) 9.33 mm 9.25 9.41 Growth rate Large 10 (KL10) 0.033 0.023 0.043 Growth rate Large 25 (KL25) 0.029 0.025 0.034 Growth rate Large 35 (KL35) 0.027 0.022 0.032 Functional response No. predators (density) 3 (3.1 m-2) Duration 14 d Larval size over interval (TL) 12.79 mm 11.96 13.83 Attack rate (D) 0.0055 pred-1 0.004 0.006 Handling time (HD) 0.84 d 0.389 1.302 Size-specific attack rate Larval density 125 m-2 No. predators (density) 2 (25 m-2) Duration 3 d Parameter 1 () -0.594 -1.079 -0.096 Parameter 2 () 1.679 0.912 2.485 Parameter 3 () 12.91 10.65 15.06
97 Density (m-2) 01234 SEA. Mortality 0.00.10.20.30.4 LibellulidZygopteranDytiscidNepidGambusia Figure 4-1. Aquatic predators A. Relative abundance and estimated field densities of the five most common tadpole predators in Amani Pond based on aquatic plots sampled prior to the start of the main experiment. B. Results from 24-hr predation trails with the five most common predators. Based on this evidence, we selected libellulid dragonfly larvae the the most abundant and most effective predators of Hyperolius larvae as the aquatic-stage predator to be used in the size and density manipulation experiment.
98 Figure 4-2. Results from the tank experiment. A. Proportion of tadpoles surviving to metamorphosis or the end of the experiment (4 months) for each treatment; B. Proportion of surviving frogs (metamorphs + larvae) that reached metamorphosis; C. Mass at metamorphosis. Means SE. Our experimental design crossed 3 levels of initial larval density with 2 levels of initial larval size and the presence or absence of aquatic predators. The three levels of initial density represent larval densities in the absence of egg stage predator effects and at two different levels of egg stage predation predation by the treefrog Afrixalus fornasini and predation by Typopsilopa sp. parasitoid flies. The two level of the density factor represent hatchling size in the absence of egg predator trait effects and predator induced size and age at hatching. Aquatic predators, larvae of the libellulid dragonfly Trapeziostigma basilaris, were either absent or present at field ambient densities.
Proportion surviving 0.20.40.60.81.0 Proportion metamorphosing 0.20.40.60.81.0 Initial larval density (larvae/tank) 5101520253035Metamorph mass (g) 0.050.100.150.200.25 No Pred Large No Pred Small Pred Large Pred Small A.B.C.
Figure 4-3.Parameterization of the model. A. Larval growth over the first 30-d for each density and initial size combination (means SE). The slopes from these regression were used to parameterize larval growth in the model (Table 4-3). B. The proportion of larval H. spinigularis (~12 mm TL) killed by T. basilaris (Eq. 2). C. The functional form of H. spinigularis size-specific predation risk by T. basilaris (Eq. 4, Table 4-3). D. The hypothesized relationship among predation rate and tadpole size and density used in the simulation model (Eq. 3 using SD). 100
Date 3/31/034/7/034/14/034/21/034/28/035/5/03Total body length (mm) 78915202510 10 Small 25 Small 35 Small 10 Big 25 Big 35 Big A .B. C. D.
102 0.00.20.40.60.81.0 Small/Early Hatched Large/Late Hatched 0.400.450.500.550.60 102535Proportion of larvae surviving 0.400.450.500.550.60 Initial density (larvae tan k -1) A B. C. SE SD Figure 4-4.Larval survival: comparison of experimental and simulation results. A. Experimental results: Mean ( SE) proportional surivival to the end of the experiment in the presence of predators at each initial density. Simulations results (note change in scale): B. Equal growth rates: Mean ( SD) proportional survival through 112 d as predicted by the simulation with equal growth rates for small/early hatched and large/late hatched larvae. C. Compensatory response: Mean ( SD) proportional survival through 112 d as predicted by the simulation with initial size and density specific growth rates (Table 4-3, Fig. 4-3a).
CHAPTER 5 MULTI-PREDATOR EFFECTS ACROSS LIFE-HISTORY STAGES: NON-ADDITIVITY OF EGGAND LARVAL-STAGE PREDATION IN AN AFRICAN TREEFROG Introduction A growing number of studies demonstrate that pair-wise predator-prey interactions can be altered by the presence of additional predators, such that aggregate predator effects cannot be predicted from their independent effects (reviewed in Sih et al. 1998, Bolker et al. 2003, Werner and Peacor 2003). Given that most prey face more than one predator species (Schoener 1989, Polis 1991), understanding when and where such non-additive multi-predator effects (MPEs) arise is a prerequisite for understanding community dynamics. In a review of 23 studies, Sih et al. (1998) showed that multiple predator effects often combined non-additively (27/43 comparisons), and that most often the combined effects were less than expected from their independent effects (20/27 comparisons), a result referred to as risk reduction. Risk reduction primarily arose due to predator-predator interactions, such as intraguild predation (Polis et al. 1989) or interference, but has also been shown to arise via predator effects on prey traits that reduce the efficacy of a second predator (Peacor and Werner 1997). Interestingly, nearly all of the studies reviewed by Sih et al. (1998) focused on prey that undergo ontogenetic habitat shifts; however, all of these studies focused on interactions of predators of a single life stage (e.g., Van Buskirk 1988, Fauth 1990, Peckarsky 1991, but see Briggs and Latto 2001). In addition to being exposed to multiple predators within life stages, species 103
104 that undergo ontogenetic habitat shifts are exposed to multiple predators across life stages. Because these predators are unlikely to interact directly, the mechanisms that most commonly give rise to risk reduction are absent. As a result, the apparently predominant occurrence of risk reduction revealed by Sih et al. (1998) may simply reflect this bias in the types of multi-predator systems studied. There have been no studies of multiple predator effects across life history stages and habitats, despite the common occurrence of spatially stage-structured prey populations (Werner and Gilliam 1984, Werner 1986, McPeek and Peckarsky 1998). Although predators of different prey stages may not interact directly, prey encounters with predators early in their life history may indirectly influence predator-prey interactions in later life-stages via changes in prey density and traits (e.g., behavior, morphology, life history). The magnitude and nature of interactions between predators across prey life-stages are not necessarily similar to those observed for predators of the same life stage. To explore this issue, we examined the effects of sequential stage-specific predators of the African reed frog, Hyperolius spinigularis. This is a model system for examining the implications of MPEs that arise from sequential stage-specific predation because the life history of the prey allows us to isolate the densityand trait-mediated indirect effects of an early life-stage predator on subsequent predator efficacy. Hyperolius spinigularis oviposits on vegetation above water, where eggs are vulnerable to predation from other treefrogs. Upon hatching, larvae drop into the pond, where they are vulnerable to predators such as larval dragonflies. Hence the early life history consists of two stages that are vulnerable to different suites of predators. Two classes of mechanisms may lead to non-additive MPEs in this system: (1) Density
105 mediated effects Egg-stage predation will decrease aquatic larval densities, potentially reducing larval competition and increasing growth. If mortality risk decreases with prey size, as is often the case with aquatic predators of anuran larvae (reviewed in Alford 1999, Table 10.3), then faster growth can reduce risk to aquatic predators. As a result, larval survival in the presence of larval predators should be greater given previous exposure to egg predators (i.e., risk reduction). In addition, reduced larval density can increase or decrease larval survival depending on the functional response of the aquatic predator (Soluk 1993). (2) Size/Age-mediated effects Egg predation induces surviving eggs in the clutch to hatch earlier at smaller sizes (Warkentin 1995, Vonesh 2000). If younger or smaller hatchlings are more vulnerable to larval predators, then previous exposure to egg predators should lead to lower larval survival (i.e., risk enhancement). Our goal was to test for a non-additive MPE between eggand larval-stage predators of H. spinigularis and to quantify the direction and relative contribution of densityand size/age-mediated effects to the overall MPE. Materials and Methods Site Information Research was conducted at the Amani Nature Reserve Conservation Headquarters in the East Usambara Mountains of northeastern Tanzania (5.06 S and 38.37 E, Elevation: 950 m) from October 2001 August 2002. Field observations were made at Amani Pond, an artificial permanent shallow pond bordered by submontane rainforest. Pond vegetation is dominated by floating mats of milfoil, Myriophyllum spicatum, and marginal patches of emergent cattails, Typha sp. Hyperolius breeding activity at Amani Pond in 2002 began on March 4, peaked in April (90% clutches laid in March April) and continued intermittently through July 17. Our surveys and lab studies revealed that
106 most egg-stage mortality was due to predation by another hyperoliid treefrog, Afrixalus fornasini, and that larvae of the libellulid dragonfly, Trapezostigma basilaris, were the most abundant aquatic predator and were substantially more effective predators than any of the other five most abundant aquatic predators in shortterm predation trials (Chapter 4). Experimental design To quantify MPEs and isolate the mechanism leading to emergent effects, we conducted a 2 x 2 x 2 factorial randomized block design (Table 5-1) in which we mimicked the two types of effects of the egg predator: effects on 1) initial larval density (N-: a control mimicking density without egg predators; N+: density with egg predators) and 2) initial larval size/age (S-: a control; S+: size and age that results from exposure to egg predators). These treatments were crossed with the presence (+PA) or absence ( PA) of aquatic predators. Each treatment was replicated four times. Nwas 38.8 larvae m-2, based on estimates of larval density that would arise without egg predators (37.43 24.2 SD larvae m-2), derived as the product of mean clutch density along two 3 x 30 m random transects between March 4 and the start of the experiment on April 4 (0.49 0.27 SD clutches m-2 per10-d), the average clutch size (83.94 21.9 SD embryos clutch-1, n=123), and the average expected survival of clutches that were not attacked (0.91 0.20 SD, n=29). N+ was 11.1 larvae m-2, approximating the field-derived estimate of 14.1 11.3 SD obtained by decrementing Nby the proportion of clutches attacked by A. fornasini (0.73 0.18 SD clutches per 10-d) and the survival of eggs in attacked clutches (0.14 0.19 SD, n = 86). +PA was 3.13 dragonflies m-2, approximating the mean field density of
107 3.16 4.78 SD dragonflies m-2. We used late instar T. basilaris larvae (Total Length (TL): 18.9 mm 2.54 SD). To simulate the sub-lethal effect of egg-stage predation on hatchling size and age we collected fresh (< Gosner stage 10, ~1-12 hrs old) H. spinigularis clutches from the field 10 each on March 26 and 29. On April 4, clutches were placed in water and gently shaken to induce hatching, yielding large, 10-day-old larvae (TL: 9.33 0.08 mm 95% CI) for Treatments 1, 2, 5, and 6 and smaller, seven-day-old larvae (TL: 7.14 0.06 mm 95% CI) for Treatments 3, 4, 7, and 8 (Table 5-1). This reflects a decrease in hatching age and size (-3 days and -2.19 0.11 mm 95% CI) consistent with that induced by A. fornasini ( x 95% CI: -2.6 1.1 d and -2.17 0.77 mm). The experiment was conducted in plastic tanks (1.2 x 0.8 x 0.4 m) arrayed in a partially shaded forest clearing near the pond, filled with 220 L of filtered pond water and covered with fine-mesh screens. To each tank we added 200 g of washed and dried pond litter, 10 g of commercial fish food (Hikari brand), 300 g of washed M. spicatum, and a 0.5 L inoculum of pond phyto/zooplankton collected using a 80 m plankton net. The experiment was initiated on April 5, with the addition of newly hatched larvae. Metamorphosed dragonflies were replaced weekly to maintain predator density. After 16 weeks, tanks were emptied and remaining larvae counted. Survival was estimated as the number of metamorphs (removed weekly) plus the number of surviving larvae (proportion of survivors reaching metamorphosis: 0.75 0.09 95% CI). Detecting Multiple Predator Effects. Testing for a non-additive MPE requires comparison of the proportion of the initial cohort that survived when no predators were present (Treatment 5), only egg predators
108 were present (Treatment 8), only aquatic predators were present (Treatment 1), and both predators were present (Treatment 4). The expected survival from oviposition to metamorphosis assuming independent risks (Billick and Case 1994) in the presence of both egg-stage and aquatic-stage predators,, can be derived from their separate effects: AE EAEANP()()() (1) which can be log-transformed into the linear (i.e., additive) form, EAEANPln()ln()ln()ln() (2) where i is the proportion of eggs that survive from oviposition to metamorphosis in the presence of predator assemblage i (i = E: egg predator only; A: aquatic predator only; E+A: both the egg and aquatic predator; NP: no predator). Because we simulated effects of egg predators through their effects on density and size/age, the fraction of eggs surviving the egg stage was either 1.0 (in the absence of egg predators) or N+/N(in their presence); the final survival depended on how these effects (of density and size/age) translated through the aquatic stage. We tested for the presence of an emergent MPE with a two-factor ANOVA using log-transformed number (or proportion) surviving to the end of the experiment as the response variable (Billick and Case 1994; Wooton 1994) a significant interaction indicates non-additivity (i.e., deviation from the model in Equations 1 and 2). Greater survivorship than expected indicates risk reduction; lower survivorship indicates risk enhancement. We then decomposed this overall response into effects attributable to densityand size/age-mediated mechanisms using the remaining treatments.
109 Results The overall multi-predator effect of Afrixalus egg-stage and dragonfly larval-stage predation was non-additive (Fig. 5-1). The expected survival in the presence of both egg and larval predators was 3.04 frogs (8.6%); however, the observed survival was more than twice that: 7.5 frogs 3.05 95% CI (21%; Fig. 5-1). Despite large effects of dragonflies in the absence of Afrixalus, dragonflies had little effect when they acted after Afrixalus. Both the densityand size/age-mediated effects of Afrixalus egg predation reduced the efficacy of dragonflies to a similar extent (Density: 54% reduction; Size/Age: 44% reduction, Fig. 5-2). Under combined Afrixalus density and size/age effects the efficacy of dragonfly predation was reduced by 70% (Fig. 5-2). Discussion This study represents the first empirical test of whether the aggregate effects of predators that attack different life-stages in different habitats can be predicted from their independent effects or whether densityand/or trait-mediated indirect effects acting across stages lead to a non-additive MPE. We found that the combined effects of arboreal egg-stage and aquatic larval-stage predation were less than expected from their independent effects, leading to risk reduction (Fig. 5-1). Previous studies focusing on MPEs within a particular life-stage have frequently observed risk reduction (Sih et al. 1998), but in the majority of these cases risk reduction arose as a result of predator-predator interactions (e.g., intraguild predation leading to reductions in predator density or foraging behavior). In our study system, where predators are separated in time and space, there is no opportunity for direct predator-predator interactions. Instead, risk reduction resulted from changes in prey (rather than predator) density (Schoener 1993) and prey traits (Peacor and Werner 1997). Only a few studies have directly compared the
110 relative effects of densityand trait-mediated indirect interactions within a system (discussed in Werner and Peacor 2003). Our study provides evidence that trait-mediated effects acting across stages (i.e., shifts in the timing of hatching and associated size/developmental consequences) can be of similar magnitude to density-mediated effects across stages (Fig. 5-2). Counter to our initial expectations, density and size/age effects acted in the same direction. As expected, reduced density led to risk reduction, suggesting that larvae released from intraspecific competition grew more rapidly through vulnerable size classes and/or that libellulid predators were less effective at lower tadpole densities. The egg predator induced size/age effect also led to risk reduction early hatching/smaller initial size increased larval survival. This is perhaps our most surprising result and is inconsistent with the results from previous short-term studies in similar systems (Warkentin 1995) and the expectations from theoretical studies (Matsuda et al. 1993), which indicate that predator-specific defenses (e.g., predator-induced early hatching) should result in risk enhancement to other predators. The greater survivorship of younger (and smaller) larvae we observed appears to be driven by the functional form of larval size-specific risk to libellulid predators and compensatory growth in early-hatched larvae (Chapter 4). Vulnerability to dragonflies initially increases until ~ 13 mm TBL and then declines as larvae attain a size refuge after 25 mm TBL (metamorphosis typically occurs between 40 50 mm TBL). The initial increase in vulnerability with size is likely driven by changes in activity associated with the shift from reliance upon yolk stores to foraging. Early-hatched larvae exhibit greater growth rates early in ontogeny compared to later-hatched larvae (a trend that is lost after reaching 25 mm) and as a result pass
111 more rapidly through the window of vulnerability. Additional mechanisms may also increase the survival of early-hatched larvae. For example, morphological differences between earlyversus later-hatched tadpoles (of the same age) have been observed in other systems (Warkentin 1999) and may result in differences in their relative vulnerability to aquatic predators (e.g., Van Buskirk and McCollum 1999). This study represents a first attempt to test for non-additive MPEs across stages and habitats. To increase our power to examine this phenomenon, we used a simplified model system. Because much of our observed response depends upon prey density, our ability to generalize beyond the particulars of this study would be limited if we found that intraspecific competition in our artificial ponds differed substantially from that in natural ponds. Two arguments suggest our results can provide insight into mechanisms that operate in natural systems. First, we established our artificial ponds using conditions very similar to the adjacent natural pond. Second, a recent meta-analysis of intraspecific competition among larval anurans demonstrated that experimental density manipulations conducted in natural ponds yield very similar estimates of intraspecific competition as studies conducted in artificial ponds (Skelly and Kiesecker 2001, mean ln (response ratios) 95% CI: artificial ponds 0.134 0.014; field enclosures 0.117 0.023). Thus, it is unlikely that our results are simply an artifact of our experimental venue. Instead, they highlight the potential for densityand trait-mediated indirect interactions to act across life-stages and habitats, resulting in non-additive multi-predator effects. Additional work in other systems with stage-structured prey populations will facilitate comparative analyses of non-additive MPEs that will help determine if risk reduction is a general feature of predator-prey systems.
112 Table 5-1. Experimental design. Nis the density of Hyperolius spinigularis larvae entering the aquatic habitat when no egg-predation occurs; N+ is the density of larvae entering the aquatic habitat with ambient Afrixalus fornasini egg-stage predation (i.e., reduced). Sis the age and average body size of tadpoles at hatching when not exposed to egg-predation; S+ is the age and size at hatching of tadpoles surviving an egg predator encounter (i.e., younger and smaller); +PA is the presence of aquatic predators (libellulid dragonfly larvae) at ambient field densities; and PA indicates their absence.
113 Larval-stage predators 2468102040 No predators (5)Larval predator only (1)Egg predator only (8)Both Observed (4)Both Expected (Eq. 2)AbsentPresentNo. surviving Figure 5-1.Testing for a non-additive MPE Effects of egg-stage predation (via Afrixalus fornasini) and larval-stage predation (via dragonfly larvae). No Predators (Treatment 5) had no larval-stage predators and an initial density and size/age that occurs in the absence of egg-stage predators. Egg predators only (Treatment 8) lacked larval-stage predators and had an initial density and size/age representative of Afrixalus egg predation. Larval predators only (Treatment 1) had dragonflies and initial density and size representative of no egg stage predation. Both (Treatment 4) indicates treatments with larval predators and initial density and size/age representative of the egg stage predators. The dashed line indicates the expected effect of both predators, assuming independent effects (Equations 1, 2). Values are means SE. The effects of the two predators are not independent: interaction term, F1,12 = 8.96, P = 0.011.
114 % Larval-stage survival 2030405060708090100 NoneDensitySize/AgeDensity +Size/AgeNone ABBBCC12345 AbsentLarval Predator:PresentEgg Predator: Figure 5-2.Examining the mechanisms Effects of the density and size/age effects of Afrixalus egg predation on aquatic larval survival in the presence of dragonflies. Numbers within bars correspond to treatment codes in Table 5-1. The first four bars give the % survival in the presence of larval-stage predators; the last bar gives the survival in the absence of both egg and larval stage predators. Treatments are further distinguished by the type of egg-stage predator effect that was included: none, reduced density, reduced size (and age), or reduced density and size/age. Larval survival was similar in all treatments without larval predators (i.e., we present treatment 5, but larval-stage survival in treatments 5 8 was similar; ANOVA, F3,12=1.84, P = 0.19). Values are means SE. Bars with different upper-case letters identify treatments that were statistically different. Responses are not equal across treatments (ANOVA, F4,15 = 7.22, P = 0.002), the independent and combined effects of density and size/age resulted in greater survival relative to the effect of dragonflies alone in the No Egg Pred treatment (Post hoc Fishers LSD; Density: P = 0.019; Size/Age: P = 0.046; Density + Size/Age: P = 0.004). Survival in the presence of both density and size/age effects and in the presence of dragonflies was not significantly different from survival without dragonflies (Fishers LSD, P = 0.10).
CHAPTER 6 SUMMARY Predators can affect prey density (i.e., by consuming prey) and induce changes in the phenotypes (e.g., behavior, morphology, life history) of surviving prey. Recent reviews have highlighted the importance of both types of effects for understanding species interactions in simple food webs (e.g., Schoener 1993, Werner and Peacor 2003). However, most studies focus on only density or trait effects few studies have examined both within a single predator-prey system. In Chapter 2, I examined both the density and trait effects of egg-stage predators of the East African treefrog, Hyperolius spinigularis, a species with arboreal clutches and aquatic larvae. To quantify the density effects of egg predators, I monitored seasonal clutch production and survivorship over two breeding seasons. This enabled me to estimate the effect of different predators on the density of larvae entering the aquatic habitat through time. I also quantified egg predator effects on hatchling traits. There is increasing evidence that embryos respond to predation risk by altering the timing of hatching (e.g., hatching early when egg predation risk is high). To test for predator-induced hatching plasticity in H. spinigularis, I conducted experiments to evaluate the effects of the two most common egg predators on the timing of hatching and hatchling phenotype. I found that the treefrog, Afrixalus fornasini, and the ephydrid fly, Typopsilopa sp., were the most important predators of H. spinigularis eggs. Egg predation resulted in a substantial reduction in aquatic larval densities, though the magnitude of reduction varied with egg predator and within and across breeding seasons. Both Afrixalus and Typopsilopa predation altered the timing of hatching and phenotype 115
116 of surviving larvae. Surviving larvae hatched earlier and were smaller and less developed than larvae from undisturbed clutches. Both the density and trait effects of egg-stage predators may affect survival of H. spinigularis in subsequent life-stages, but they may act in opposite directions. Reductions in larval-stage densities are likely to increase larval growth and survival (e.g., via reduced intraspecific competition), while reduced initial size/age is likely to reduce larval survival (e.g., via size-selective aquatic predation). In Chapter 3, I examine how the density effects of predators early in ontogeny affect size at metamorphosis and how size affects encounters with post-metamorphic predators. I estimated the functional form of density-dependent size at metamorphosis, through an experiment in which I varied larval density at eight levels in the presence and absence of aquatic predators (libellulid dragonfly larvae). I then used these functional forms to estimated the effect of egg and larval predators on size at metamorphosis. Predators of arboreal eggs reduce larval input into the aquatic habitat by consuming eggs, potentially reducing intraspecific competition. Similarly, aquatic-stage predators consume larvae and may be size-selective. Both egg and larval predators increased size at metamorphosis of survivors. I then conducted a second experiment to evaluate where the differences in size at metamorphosis due to predators affected the interaction between metamorphs and Thalassius sp. fishing spiders a common metamorph predator. The results showed that larger metamorphs had significantly higher survival in the presence of spiders. Thus, the effects of predators early in ontogeny can alter predator-prey interactions in later stages. Here the effects of early predators (even 2 life-stages earlier)
117 on size at metamorphosis facilitated larval survival in the presence of post-metamorphic predators. In Chapter 4, I focused on the consequences of predator-induced early hatching. Chapter 2 showed that developing H. spinigularis embryos can respond to predators by altering the age at which they hatch from eggs. Past studies in similar systems suggest that this kind of hatching plasticity involves a trade-off between embryonic and hatchling predation risk. However, these studies have primarily focused on detecting a predator-induced hatching response and have only considered the short-term consequences of potential trade-offs. Evidence of trade-offs based on short-term consequences (e.g., those arising within 24 hrs of hatching) may be misleading, because it does not allow for longer-term compensatory responses. Long-term responses, which may even extend to subsequent life stages, can stem either from changes in prey physiology and behavior or from predator-driven changes in prey population density. Thus, any trade-offs associated with trait-mediated predator effects on hatching must be examined within the context of the simultaneous effects of embryonic predators on larval density. In order to explore the consequences of the density and size/age-mediated effects of egg-stage predators for larval-stage growth and survival, I conducted an experiment in which I manipulated initial larval size and density (mimicking the effects of egg predators) and the presence of aquatic predators. Based on evidence from the literature, I expected that small, predator-induced, early hatchlings would exhibit lower survival in the presence of aquatic predators than larger, later-hatched larvae. Surprisingly, I found that both the density and size/age effects of predators enhance larval survival. These results motivated me to develop a model parameterized from two additional experiments to explore whether a
118 combination of mechanismscompensatory growth, and density and size specific predation could give rise to this result. Patterns of larval survival in the simulation were consistent with those in the experiment, suggesting that compensatory growth in early-hatched larvae enables them to grow more rapidly through vulnerable size classes than later hatched larvae, leading to higher overall survival. Thus, in this system there does not appear to be a trade-off in vulnerability between egg and larval predators. Instead, the results suggest that the cost that balances the survival benefit of hatching early to evade egg predators arises later in the life history, as a result of smaller size at metamorphosis. Other studies have documented the negative fitness consequences of being small at metamorphosis for amphibians and, in Chapter 3, I showed that small size at metamorphosis can increase H. spinigularis vulnerability to post-metamorphic predators. Finally, in Chapter 5, I examined whether the effects of sequential predators across eggand larval-stages of the H. spinigularis were non-independent. The effects of multiple predators on their prey are frequently non-additive because of interactions among predators. When prey shift habitats through ontogeny, many of their predators cannot interact directly. However, predators that occur in different habitats or feed on different prey stages may still interact through indirect effects mediated by prey traits and density. I conducted an experiment to evaluate the combined effects of arboreal egg-stage and aquatic larval-stage predators of H. spinigularis. Egg and larval predator effects were non-additive more Hyperolius survived both predators than predicted from their independent effects. Egg-stage predator effects on aquatic larval density and size and age at hatching reduced the effectiveness of larval-stage predators by 70%. These
119 results indicate that densityand trait-mediated indirect interactions can act across life-stages and habitats, resulting in non-additive multi-predator effects. Given that many prey have complex life histories and are probably vulnerable to predators in each stage, such indirect effects acting across stages may very well be common in nature and potentially important aspects of many food webs.
LIST OF REFERENCES Agresti, A. 1996. An Introduction to Categorical Data Analysis. Wiley, New York. Akaike, H. 1992. Information theory and an extension of the maximum likelihood principle. In: Breakthroughs in Statistics (eds Kotz S. and Johnson N.) Springer-Verlag, New York, pp. 610-624. Alford, R. A. 1999. Ecology: Resource use, competition, and predation. In: Tadpoles: The Biology of Anuran Larvae (eds. McDiarmid, R.W. and Altig, R.). The University of Chicago Press, Chicago, pp. 240-278. Alford, R. A., and Harris, R.N. 1988. Effects of larval growth history on an anuran metamorphosis. Am. Nat. 133: 91-106. Alford, R. A., and Jackson, G. D. 1993. Do cephalopods and larvae of other taxa grow asymptotically? Am. Nat. 141: 717-728. Altwegg, R., and Reyer, H. 2003. Patterns of natural selection on size at metamorphosis in water frogs. Evolution 57: 872-882. Anholt, B.R. 1991. Measuring selection on a population of damselflies with a manipulated phenotype. Evolution 45: 1091-1106. Anholt, B. R., and Werner, E. E. 1998. Predictable changes in predation mortality as a consequence of changes in food availability and predation risk. Evol. Ecol. 12: 729-738. Azevedo-Ramos, C., Magnusson, W. E., and Bayliss, P. 1999. Predation as a key factor structuring tadpole assemblages in an area in central Amazonia. Copeia 1999: 22-33. Berven, K. A. 1990. Factors affecting population fluctuations in larval and adult stages of the wood frog (Rana sylvatica). Ecology 71: 1599-1608. Berven, K.A. 1995. Population regulation in the wood frog, Rana sylvatica, from 3 diverse geographic localities. Austral. J. Ecol. 20: 385-392. Beverton, R.J.H., and Holt, S.J. 1957. On the Dynamics of Exploited Fish Populations. Chapman and Hall, London. 120
121 Billick, I., and Case, T. J. 1994. Higher order interactions in ecological communities: What are they and how can they be detected? Ecology 75: 1529-1543. Blaustein, L. 1997. Non-consumptive effects of larval Salamandra on crustacean prey; can eggs detect predators? Oecologia 110: 212-217. Bolker, B., Holyoak, M., Kivan, V., Rowe, L., and Schmitz, O. 2003. Connecting theoretical and empirical studies of trait-mediated interactions. Ecology 84: 1101-1114. Briggs, C. J., and Latto J. 2001. Interactions between the egg and larval parasitoids of a gall-forming midge and their impact on the host. Ecol. Entomol. 26: 109-116. Carpenter, S. R., Kitchell, J. E., Hodgson, J. R., Cochran, P. A., Elser, J. J., Elser, M. M., Lodge, D. M., Kretchner, He, X., and von Ende, C. N. 1987. Regulation of lake primary productivity by foodweb structure. Ecology 68: 1863-1876. Carroll, E. J., and Hedrick, J. L. 1974. Hatching in the toad Xenopus laevis: morphological events and evidence for a hatching enzyme. Devel. Biol. 38: 1-13. Caswell, H. 2001. Matrix Population Models. 2nd edition. Sinauer Assoc. Inc., Sunderland, MS. Chalcraft, D. R., and Andrews, R. M. 1999. Predation on lizard eggs by ants: species interactions in a variable physical environment. Oecologia 119: 285-292. Chivers D. P., Kiesecker J. M., Marco, A., DeVito, J., Anderson, M. T., and Blaustien, A. R. 2001. Predator-induced life history changes in amphibians: egg predation induces hatching. Oikos 92: 135-142. Collins, J. P. 1975. A comparative study of the life history strategies in a community of frog. PhD Thesis. University of Michigan, Ann Arbor, MI. Connell, J. H. 1978. Diversity in tropical rain forests and coral reefs. Science 199: 1302-1310. Dash, M. C., and Hota, A. K. 1980. Density effects on the survival, growth rate, and metamorphosis of Rana tigerina tadpoles. Ecology 61: 1025-1028. Davis, R. A., and Disney, R. H. L. 2003. Natural history and description of Aphiura breviceps Schmitz a scuttle fly (Diptera: Phoridae) whose larvae prey on the eggs of frogs (Anura: Myobatrachidae) in western Austraila. Austral. J. Entomology 42: 18-21. Dewitt, T. J. 1998. Costs and limits of phenotypic plasticity: tests with predator induced morphology and life history in a freshwater snail. J. Evol. Biol. 11: 465-480.
122 Donnelly, M. A., and Guyer, C. 1994. Patterns of reproduction and habitat use in an assemblage of Neotropical hylid frogs. Oecologia 98: 291-302. Dorn, N. 2003. Effects of omnivorous crayfish on fish populations and the structure of littoral communities. PhD Thesis. Michigan State University, East Lansing, MI. Drewes, R. C. and Altig, R. 1996. Anuran egg predation and heterocannibalism in a breeding community of East African frogs. Trop. Zool. 9: 333-342. Elliot, J. M. 2003. A comparative study of the functional response of four species of carnivorous stoneflies. Freshwater Biol. 48: 191-202. Estes, J. A., and Palminsano, J. F. 1974. Sea otters: their role in structuring nearshore communities. Science 185: 1058-1060. Fauth, J.E. 1990. Interactive effects of predators and early larval dynamics of the treefrog Hyla crysoscelis. Ecology, 71: 1609-1616. Fox, J. 1997. Applied Regression Analysis, Linear Models, and Related Methods. Sage Publications, Thousand Oaks, California. Fuiman, L. A., and Magurran, A. E. 1994. Development of predator defenses in fishes. Rev. Fish. Biol. Fish. 4: 145-183. Goater, C.P. 1994. Growth and survival of postmetamorphic toads: Interactions among larval history, density, and parasitism. Ecology 75: 2264-2274. Gascon, C. 1991. Population and community level analysis of species occurrences of central Amazonian rainforest tadpoles. Ecology 72: 1731-1746. Gascon, C. 1992. Aquatic predators and tadpole prey at a central Amazonian site: Field data and experimental manipulations. Ecology 73: 971-980. Goldberg, D. E., and Bartron, A. M. 1992. Patterns and consequences of interspecific competition in natural communities: a review of experiments with plants. Am. Nat. 139: 771-801. Gosner, K. L. 1960. A simplified table for staging anuran embryos and larvae with notes on identification. Herpetol. 16: 203-210. Hamilton, A. C., and Bensted-Smith, R. 1989. Forest Conservation in The East Usambara Mountains Tanzania. IUCN, Gland, Switzerland and Cambridge, U.K. Hayes, M. P. 1983. A technique for partitioning hatching and mortality estimates in lead-breeding frogs. Herpetol. Rev. 14: 115-116.
123 Hero, J-M., Gascon, C., and Magnusson, W. E. 1998. Direct and indirect effects of predation on tadpole community structure in the Amazon rainforest. Austral. J. Ecol. 23: 474-482. Heyer, W. R., Donnelly, M. A., McDiarmid, R. W., Hayek, L. C., and Foster, M. S. 1994. Measuring and Monitoring Biological Diversity. Standard Methods for Amphibians. Smithsonian Institution, Washington, D.C. Holling, C. S. 1959. Some characteristics of simple types of predation and parasitism. Canad. Entomol. 91: 385-398. Holt, R. D. 1977. Predation, apparent competition, and the structure of prey communities. Theor. Pop. Biol. 12: 197-229. Hurd, L. E., and Eisenberg, R. M. 1990. Arthropod community responses to manipulation of a bitrophic predator guild. Ecology 71:2107-2114. Ihaka, R. and Gentleman, R. 1996. R: A language for data analysis and graphics. J. Comp. Graph. Stat. 5: 299-314. Janzen, D. H. 1970. Herbivores and the number of tree species in a tropical forests. Am. Nat. 104: 501-28. Jones, M., Laurila, A., Peuhkuri, N., Piironen, J., and Sepp, T. 2003. Timing an ontogenetic niche shift: response of emerging salmon alevins to chemical cues from predators and competitors. Oikos. 122: 155-163. Juliano, S. A. 1993. Nonlinear curve fitting: Predation and functional response curves. In: Design and Analysis of Ecological Experiments (eds. Scheiner, S.M. and Gurevitch, J.). Chapman and Hall, New York, pp. 159-182. Laurila A., Crochet, P. A., and Merila, J. 2001. Predation-induced effects on hatching morphology in the common frog (Rana temporaria). Canad. J. Zool. 79: 926-930. Laurila, A., Pakkasmaa, S., Crochet, P., Merila, J. 2002. Predator-induced plasticity in early life history and morphology in two anuran amphibians. Oecologia. 133: 524-530. Li, D. Q. 2002. Hatching responses of subsocial spitting spiders to predation risk. Proc. Roy. Soc. London, Series B. 269: 2155-2161. Lima, S. L., and Dill, L. M. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canad. J. Zool. 68: 619-640. Lips, K. R. 2001. Reproductive trade-offs and bet-hedging in Hyla calypsa, a Neotropical treefrog. Oecologia 128: 509-518.
124 Lovett, J. C., and Wasser, S. K. 1993. Biogeography and Ecology of the rain forests of eastern Africa. Cambridge University Press, Cambridge, U.K. Luttbeg, B., Rowe, L., and Mangel, M. 2003. Prey state and experimental design affect the relative size of traitand density-mediated indirect effects. Ecology 84: 1140-1150. Lyons, L. 1991. A Practical Guide to Data Analysis for Physical Science Students. Cambridge University Press, Cambridge, U.K. Martin, J-L, and Joron, M. 2003. Nest predation in forest birds: influence of predator type and predators habitat quality. Oikos 102: 641-653. Martin, K. L. M. 1999. Ready and waiting: Delayed hatching and extended incubation of anamniotic vertebrate terrestrial eggs. Am. Zool. 39: 279-288. Martin, T. H., Wright, R. A., and Crowder, L. B. 1989. Non-additive impact of blue crabs and spot on their prey assemblages. Ecology 70: 1935-1942. Matsuda, H., Abrams, P. A., and Hori, M. 1993. The effect of adaptive antipredator behavior on exploitative competition and mutualism between predators. Oikos 68: 549-559. Matsuda, H., Hori, M., Abrams. P. A. 1994. Effects of predator-specific defense on predator persistence and community complexity. Evol. Ecol. 8:628-638. Matsuda, H., Hori, M., and Abrams, P. A. 1996. Effects of predator-specific defense on biodiversity and community complexity in two-trophic-level communities. Evol. Ecol. 10: 13-28. McDiarmid, R. W. 1978. Evolution of parental care in frogs. In: The Development of Behavior: Comparative and Evolutionary Aspects. (eds. Burkhardt, G.M, and Bekoff, M.) STPM Press, New York, pg. 127-147. McPeek, M. A., and Peckarsky B. L. 1998. Life histories and the strengths of species interactions: Combining mortality, growth, and fecundity effects. Ecology 79: 867-879. Metcalfe, N. B., and Monaghan, P. 2001. Compensation for a bad start: grow now, pay later? Trends Ecol. Evol. 16: 254-260. Mittelbach, G. G., and Osenberg, C. W. 1993. Stage-structured interactions in bluegill: Consequences of adult resource variation. Ecology 74: 2381-2394. Moeur, J.E., and Istock, C.A. 1980. Ecology and evolution of the pitcher-plant mosquito. IV. Larval influence over adult reproductive performance and longevity. J. Anim. Ecol. 49: 775-792.
125 Moore, R. D., Newton, B., and Sih, A. 1996. Delayed hatching as a response of streamside salamander eggs to chemical cues from predatory sunfish. Oikos 77: 33-335. Morin, P. J. 1995. Functional redundancy, non-additive interactions, and supply-side dynamics in experimental pond communities. Ecology 76: 133-149. Myers, N., Mittermeir, R. A., Mittermeir, C. G., da Fonseca, G. A. B., and Kent, J. 2000. Biodiversity hotspots for conservation priorities. Nature 403: 853-858. Osenberg, C. W., and Mittelbach, G. G. 1989. Effects of body size on the predator-prey interaction between pumpkinseed sunfish and gastropods. Ecol. Monogr. 59: 405-432. Osenberg, C.W., Mittelbach G.G., and Wainwright, P.C. 1992. Two-stage life histories in fish: the interaction between juvenile competition and adult performance. Ecology 73:255-267. Osenberg, C. W., Olson, M. H., and Mittelbach, G. G. 1994. Stage structure in fishes: resource productivity and competition gradients. In: Theory and Application in Fish Feeding Ecology. (eds. Stouder, D. J., Fresh, K. L., and Feller, R. J.).University of South Carolina Press, Columbia, pp. 151-170 Paine, R. T. 1966. Food web complexity and species diversity. Am. Nat. 100: 65-75. Peacor, S. D. 2002. Positive effects of predators on prey growth rate through induced modifications of prey behavior. Ecol. Let. 5: 77-85. Peacor, S. D., and Werner E. E. 1997. Trait-mediated indirect interactions in a simple aquatic food web. Ecology 78: 1146-1156. Peacor, S. D., and Werner, E. E. 2001. The contribution of trait-mediated indirect effects to the net effects of a predator. Proc. Nat. Acad. Sci. USA 98: 3904-3908. Peckarsky, B. L. 1991. Mechanisms of intraand interspecific interference between larval stoneflies. Oecologia 85: 521-529. Petranka, J. W., and Kennedy, C. A. 1999. Pond tadpoles with generalized morphology: is it time to reconsider their functional roles in aquatic communities? Oecologia 120: 621-631. Pinheiro, J. C., and Bates, D. 2000. Mixed-effects Models in S and S-plus. Springer-Verlag. New York. Polis, G. A. 1991. Complex trophic interactions in deserts: an empirical critique to food web theory. Am. Nat. 138: 123-155.
126 Polis, G. A., Myers, C. A., and Holt, R. D. 1989. The ecology and evolution of intraguild predation: potential competitors that eat each other. Ann. Rev. Ecol. Syst. 20: 297-330. Rsnen, K., Laurila, A., and Merila, J. 2002. Carry-over effects of embryonic acid conditions on development and growth of Rana temporaria tadpoles. Freshwater Biol. 47: 19-30. Relyea, R. A. 2000. Trait-mediated indirect effects in larval anurans: reversing competitive outcomes with the threat of predation. Ecology 81: 2278-2289. Relyea, R. A. 2001. Morphological and behavioral plasticity of larval anurans in response to different predators. Ecology 82: 523-540. Relyea, R. A., and Hoverman, J. T. 2003. The impact of larval predators and competitors on the morphology and fitness of juvenile treefrogs. Oecologia 134: 596-604. Rogers, D. J. 1972. Random search and insect population models. J. Animal Ecol. 41: 369-383. Rosenheim, J. A., Wilhoit, L. R., and Armer, C. A. 1993. Influence of intraguild predation among generalist insect predators on the suppression of a herbivore population. Oecologia 96: 439-449. Schalk G., Forbes, M. R., and Weatherhead, P. J. 2002. Developmental plasticity and growth rates of green frog embryos and tadpoles in relation to a leech (Macrobdella decora) predator. Copeia 2002: 445-449. Schitz, A. 1999. Treefrogs of Africa. Chimara, Meckenheim, Germany. Schmitt, R. J. 1987. Indirect interactions between prey: apparent competition, predator aggregation, and habitat segregation. Ecology 68: 1887-1897. Schmitt, R.J., Osenberg, C.W., and Bercovitch, M. G. 1984. Mechanisms and consequences of shell fouling in the kelp snail, Norrisia norrisi (Sowerby) (Trochidae): Indirect effects of Octopus drilling. J. Exp. Mar. Biol. Ecol. 69: 267-281. Schmitt, R.J., Holbrook, S.J., and Osenberg, C.W. 1999. Quantifying the effects of multiple processes on local abundance: a cohort approach for open populations. Ecol. Let. 2: 294-303. Schoener, T. W. 1989. Food webs from the small to the large. Ecology 70: 1559-1589. Schoener, T. W. 1993. On the relative importance of direct versus indirect effects in ecological communities. In: Mutualisms and Community Organization (eds. Kawanabe, H., Cohen, J. E. and Iwasaki, K.). Oxford University Press, Oxford, pp. 365-411.
127 Scott, D.E. 1994. The effect of larval density on adult demographic traits in Ambystoma opacum. Ecology 75: 1383-1396. Semlitsch, R.D., and Caldwell, J. P. 1982. Effects of density on growth, metamorphosis, and survivorship in tadpoles of Scaphiopus holbrooki. Ecology 63: 905-911. Shepherd, J.G. 1982. A versatile new stock recruitment relationship for fisheries, and the construction of sustainable yield curves. J. Cons. Inter. Explor. Mer. 40: 67-75. Sih, A., and Moore, R. D. 1993. Delayed hatching in salamander eggs in response to enhanced larval predation. Am. Nat. 142: 947-960. Sih, A., Englund, G., and Wooster, D. 1998. Emergent impacts of multiple predators on prey. Trends Ecol. Evol. 13: 350-355. Skelly, D. K. 1992. Field evidence for a cost of behavioral antipredator responses in a larval amphibian. Ecology 73: 704-708. Skelly, D. K., and Werner, E. E. 1990. Behavioral and life-historical responses of larval American toads to an odonate predator. Ecology 71: 2313-2322. Skelly, D. K., and Kiesecker, J. M. 2001. Venue and outcome in ecological experiments: manipulations of larval anurans. Oikos 94: 198-208. Skutch, A.F. 1949. Do tropical birds rear as many young as they can nourish? Ibis 91: 430-458. Smith, D. C. 1987. Adult recruitment in chorus frogs: effects of size and date at metamorphosis. Ecology 68: 344-350. Soluk, D.A. 1993. Multiple predator effects: Predicting combined functional response of stream fish and invertebrate predators. Ecology 74: 219-225. Spiller, D. A., and Schoener, T. W. 1996. Food-web dynamics on some small subtropical islands: effects of top and intermediate predators. In: Food webs: Integration of Pattern and Dynamics. (eds. Polis, G. A., and Winemiller, K. O.).Chapman and Hall, New York, pp. 160-169. Stevens, R. A. 1971. A new treefrog from Malawi. Zool. Afr. 6: 313-320. Taigan, T. L. 1981. Water balance of the eggs of Eleuthrodactylus coqui: importance of parental brooding. In: Annual Meeting S.S.A.R./H.L., Memphis, pp. 72. Tanhuanp, M., Ruohomaki, K., Kaitaniemi, P. 2003. Influence of adult and egg predation on the reproductive success of Epirrita autumnata (Lepidoptera: Geometridae). Oikos 102: 263-272.
128 Tollrain, R. and Harvell, C. D. 1999. The Ecology and Evolution of Inducible Defenses. Princeton University Press, Princeton. Townsend, D.S., Stewart, M.M., Pough F.H., and Brussard, P.F. 1984. Male parental care and its adaptive significance in a neotropical frog. Anim. Behav. 32: 421-431. Van Buskirk, J. 1988. Interactive effects of dragonfly predation in experimental pond communities. Ecology 69: 857-867. Van Buskirk, J., and Smith, D.C. 1991. Density-dependent population regulation in a salamander. Ecology 72: 1747-1756. Van Buskirk, J., and McCollum, S. A. 1999. Plasticity and selection explain variation in tadpole phenotype between ponds with different predator composition. Oikos 85: 31-39. Van Buskirk, J., and Schmidt, B. R. 2000. Predator-induced phenotypic plasticity in larval newts: trade-offs, selection, and variation in nature. Ecology 81: 3009-3028. Vandermeer, J.H. 1969. The comparative structure of communities: an experimental approach with protozoa. Ecology 50: 362-371. Venables, W. N., and Ripley, B. D. 2002. Modern Applied Statistics with S. Fourth edition. Springer-Verlag, New York. Vial, J. L., and Prieb, F. B. 1967. An investigation of antibiosis as a function of brooding behavior in the salamander, Plethodon cinereus. Trans. Missouri Acad. Sci. 1: 37-40. Villa, J. 1980. Frogflies from Central and South America with notes on other organisms of the amphibian egg microhabitat. Brenesia 17: 49-68. Villa, J. 1984. Biology of a Neotropical Glass Frog, Centrolenella fleischmanni (Boettger), with special reference to frogfly associates. Milwaukee Pub. Mus.: Contrib. Biol. Geol. 55: 1-59. Vonesh, J. R. 2000. Dipteran predation on the eggs of four Hyperolius frog species in western Uganda. Copeia 2000: 560-566. Vonesh, J.R. and De la Cruz, O. 2002. Complex life cycles and density dependence: assessing the contribution of egg mortality to amphibian declines. Oecologia 133: 325-333. Vonesh, J. R., and Osenberg, C. W. 2003. Multi-predator effects across life-history stages: non-additivity of egg-and larval-stage predation in an African treefrog. Ecol. Let. 6: 503-508.
129 Warkentin, K. M. 1995. Adaptive plasticity in hatching age: A response to predation risk trade-offs. Proc. Nat. Acad. Sci. USA 92: 3507-3510. Warkentin, K. M. 1999a. Effects of hatching age on development and hatchling morphology in the red-eyed treefrog, Agalychnis callidryas. Biol. J. Linn. Soc. 68: 443-470. Warkentin, K. M. 1999b. The development of behavioral defenses: a mechanistic analysis of vulnerability in red-eyed tree frog hatchlings. Behav. Ecol. 10: 251-262. Warkentin, K. M. 2000. Wasp predation and wasp-induced hatching of red-eyed treefrog eggs. Anim. Behav. 60: 503-510. Warkentin, K. M., Currie, C. R., and Rehner, S. A. 2001. Egg-killing fungus induces early hatching of red-eyed treefrog eggs. Ecology 82: 2860-2869. Wedekind, C. 2002. Induced hatching to avoid infectious egg disease in whitefish. Current Biology 12: 69-71. Werner, E. E. 1986. Amphibian metamorphosis: growth rate, predation risk, and the optimal size at transformation. Am. Nat. 128: 319-341. Werner, E. E. 1992. Individual behavior and higher-order species interactions. Am. Nat. 140: S5-S32. Werner, E. E., Gilliam, J. F., Hall, D. J., and Mittelbach, G. G. 1983. An experimental test of the effects of predation risk on habitat use in fish. Ecology 64: 1540-1548. Werner E. E., and Gilliam, J. F. 1984. The ontogenetic niche and species interactions in size-structured populations. Annu. Rev. Ecol. Syst. 15: 393-425. Werner, E.E., and Anholt, B.R. 1993. Ecological consequences of the trade-off between growth and mortality rates mediated by foraging activity. Am. Nat. 242-272. Werner, E. E., and Anholt, B. R. 1996. Predator-induced behavioral indirect effects: consequences to competitive interactions in anuran larvae. Ecology 77: 157-169. Werner, E. E. and Peacor, S. D. 2003. A review of trait-mediated indirect interactions in ecological communities. Ecology 84: 1083-1100. Wilbur, H. M. 1977. Density dependent aspects of growth and metamorphosis in Bufo americanus. Ecology 58: 196-200. Wilbur, H. M., and Fauth, J. E. 1990. Experimental aquatic food webs; interactions between two predators and two prey. Am. Nat. 135: 176-204. Wooton, J. T. 1993. Indirect effects and habitat use in an intertidal community: interaction chains between two predators and two prey. Am. Nat. 135: 71-89.
130 Wooton, J. T. 1994. Predicting direct and indirect effects: an integrated approach using experiments and path analysis. Ecology 75: 151-165. Yorke, C. 1983. Survival of embryos and larvae of the frog Ploypedates leucomystax in Malaysia. J. Herpetol. 18: 235-241.
BIOGRAPHICAL SKETCH James Vonesh received his Bachelor of Science in biology from Eckerd College, St. Petersburg, Florida, in 1991. After finishing his undergraduate studies he spent three years teaching English in Japan. In 1998 he received a Master of Science from the zoology program at the University of Florida, Gainesville, Florida. He is married to Sophia R. Balcomb and has one daughter, Savannah. 131