Carcasses of invasive species are predominantly utilized by invasive scavengers in an island ecosystem

Authors


  • Corresponding Editor: Robert R. Parmenter.

Abstract

Invasive species have significantly affected ecosystems, particularly islands, and species invasions continue with increasing globalization. Largely unstudied, the influence of invasive species on island ecosystem functions, especially scavenging and decomposition, could be substantive. Quantifying carcass utilization by different scavengers and shifts in community dynamics in the presence of invasive animals is of particular interest for understanding impacts on nutrient recycling. Invasive species could benefit greatly from carcass resources within highly invaded island ecosystems, through increased invasion success and population growth, subsequently exacerbating their impacts on native species. We quantified how experimentally placed invasive amphibian, reptile, small mammal, and bird carcasses were utilized by vertebrate and invertebrate scavengers on the Big Island of Hawai'i in three island habitats: a barren lava field, a vegetated lava field, and a rainforest. We used camera traps to record vertebrate scavengers removing carcasses and elapsed time until removal. We evaluated differences in scavenging between vertebrates and invertebrates and within the vertebrate community across different habitats and carcass types. Despite the small carcass sizes (<1 kg) used in this study, 55% of carcasses were removed by vertebrate scavengers, all invasive: mongoose, rodents, cats, pigs, and a common myna. Our data indicate that invasive vertebrate scavengers in this island ecosystem are highly efficient at assimilating a range of carrion resources across a variety of habitats. Carcasses of invasive animals could contribute substantially to energy budgets of other invasive vertebrate species. This may be a critical component contributing to successful invasions especially on islands and subsequent impacts on ecosystem function.

Introduction

Invasive species have become prevalent worldwide and are considered one of the greatest threats to biodiversity, especially in island ecosystems (Reaser et al. 2007). Moreover, invasive species drive significant ecological change on islands, causing greater harm compared to mainland ecosystems, particularly regarding ecosystem function (D'Antonio and Dudley 1995). The impact invasive species have on energy flow can be particularly destructive and have widespread ecosystem effects (Rodda and Savidge 2007, Mortensen and Dupont 2008, McNatty et al. 2009). Invasive species that utilize flexible feeding strategies, such as facultative scavengers, could have severe impacts on energy flow and an increased capability to invade (Chapple et al. 2012).

There is increasing recognition for the amount of necromass processed by scavengers and the prevalence of facultative scavenging across vertebrate species, particularly invasive species (Putnam 1983, DeVault et al. 2003, Wilmers et al. 2003, Brown et al. 2015, Mateo-Tomás et al. 2015, Huijbers et al. 2016). The redistribution of energy by vertebrate scavengers, as opposed to invertebrates and decomposers, could be far more influential in ecosystem-wide trophic interactions than is currently understood (DeVault et al. 2003, Wilson and Wolkovich 2011). Carcass utilization by vertebrates varies widely and depends on various environmental factors (such as temperature, moisture, and season), habitat fragmentation, carcass size, predation risk, scavenger densities, and scavenger community composition, particularly presence of invasive species (McKillup and McKillup 1994, DeVault et al. 2004, 2011, Selva et al. 2005, Beasley et al. 2012, Olson et al. 2012, Brown et al. 2015, Moleón et al. 2015). Determining how carcass resources are partitioned between native and invasive vertebrate and invertebrate scavengers throughout the landscape could reveal ecosystem-wide impacts of invasive species on energy flow.

Additionally, many invasive animals are able to achieve high densities and lack predators in island ecosystems, potentially leading to a large number of carcasses produced via mechanisms other than predation, for example, starvation, exposure, and disease (Schoener and Spiller 1995, Crooks and Soule 2001, Reaser et al. 2007, Beard et al. 2009, Borroto-Paez 2009). Given the functional importance and taxonomic breadth of vertebrate scavenging and the potential necromass of invasive species, research examining nutrient recycling by invasive animals in invaded island ecosystems could reveal possible synergistic impacts of invasive scavengers and carcasses. Studies evaluating either scavenging by invasive animals or of invasive animal carcasses suggest that invasive species do alter nutrient recycling (Richards and Goff 1997, Howald et al. 1999, Read and Wilson 2004, Gangoso et al. 2006, Beckmann and Shine 2011, Wilson and Wolkovich 2011, Huijbers et al. 2013, 2016, Schlacher et al. 2013, Brown et al. 2015, Mateo-Tomás et al. 2015). However, previous studies usually only evaluate scavenging on one carcass type, and few studies have simultaneously examined both scavenging on and by vertebrate invasive species (Sebastián-González et al. 2013).

To elucidate the involvement of invasive animals in nutrient recycling in an invaded island ecosystem, we investigated the utilization of invasive amphibian, reptile, small mammal, and bird carcasses by scavengers across three habitat types on the Big Island of Hawai'i. Hawai'i is currently in an invasion crisis and leads the United States in number of federally endangered species and extinctions, with invasive species considered the top threat to native species (Dodson et al. 1997, Cox 1999, Holt 1999). Therefore, understanding the impact of invasive species on nutrient recycling in Hawai'i is pertinent for conservation and management efforts. By placing camera traps baited with these carcasses at three sites on Hawai'i, we sought to achieve four objectives. Objective (1) was to quantify the proportion of carcasses utilized by vertebrates vs. the proportion scavenged either wholly or primarily by invertebrates and the species composition of the vertebrate scavenger community. We predicted that vertebrates would utilize more experimental carcasses than invertebrates given the efficient vertebrate scavengers known to be on the island, mongoose (Herpestes javanicus Geoffroy) and pigs (Sus scrofa Linnaeus). We predicted that the vertebrate scavenger community would be composed predominantly of invasive species given the low population densities of native scavengers. Also, we sought to determine whether habitat and/or carcass type influenced the proportion of carcasses scavenged by (2) vertebrates and invertebrates and (3) species within the vertebrate scavenger community. In addition to predicting that habitat and carcass type would influence carcass removal, we predicted that relative abundances of vertebrate scavengers would influence scavenging. Objective (4) was to determine whether the elapsed time to find carcasses and remove carcasses by vertebrate scavengers differed across habitats and carcass types. We predicted that in more open habitats carcasses would be found and removed more quickly, and we predicted that more massive carcass types would be found quicker, relative to smaller carcass types (DeVault et al. 2004, Ruzicka and Conover 2012, Moleón et al. 2015). This research provides novel information about the scavenging ecology of a highly invaded island ecosystem and allows us to predict how invasive species both as scavengers and as carcasses influence island ecosystem function and the potential for nutrient recycling to contribute to invasions.

Methods

Study sites

We chose three sites within Hawai'i Volcanoes National Park and Pu'u Maka'ala Natural Area Reserve on the Big Island of Hawai'i with similar rainfall and air temperature, but that, due to differences in lava flow history and subsequent substrate and vegetation development, presented different habitats. One site, referred to throughout as “lava site,” was a coastal lava field, with relatively low elevation, intact, highly reflective lava, and sparse grasses with very few trees and shrubs (Fig. 1). Carcasses were placed approximately 500 m to 2 km from coastal cliffs at this site. Our second site, “scrub site,” was further inland, approximately 10 km, and had soil varying from intact lava to small pebblelike substrate. The vegetation consisted of trees, shrubs, and grasses that were thick in some areas but short in height. Wide, deep cracks were interspersed throughout this site. Our third site, “rainforest site,” was also inland (16–25 km) and had moist soil that gave rise to thick vegetation, mostly shrubs and tall trees.

Figure 1.

Sites used for invasive species scavenging research on the Big Island of Hawai'i. Pictures show the lava site (a), an area characterized by lava fields with little vegetation; the scrub site (b), characterized by weathered lava fields and vegetation of short trees, shrubs, and grasses; and the rainforest site (c), a rainforest with lush, thick vegetation.

Study species

We used the following invasive species as experimentally placed carcasses: herpetofauna—coqui frogs (Eleutherodactylus coqui Thomas), geckos (various spp.), and cane toads (Rhinella marina Linnaeus); mammals—mice (Mus musculus Linnaeus), rats (Rattus rattus Linnaeus and Rattus exulans Peale), and mongoose; and birds (various spp., see Appendix S1 for full species list, carcass mass, and introduction dates to Hawai'i). Across Hawai'i and at our study sites, rodents and mongoose have large, established populations (Baldwin et al. 1952, Atkinson 1977). Geckos and non-native birds are considered common islandwide, especially in urban areas, and can be found at all three sites but likely do not have large populations at these sites (Kraus 2005, Pyle and Pyle 2009). Cane toads and coqui frogs are common in some areas of Hawai'i (Lever 2003, Kraus 2005, Beard et al. 2009); however, while cane toads and coqui frogs are found occasionally near our rainforest site and threat of establishment by coqui frogs in Hawai'i Volcanoes National Park is considered high, they are not currently established at our sites (Kraus 2005).

Experimental design

Camera trap trials, which consisted of one carcass placed in front of a camera (Reconyx HyperFire PC900, Holmen, Wisconsin, USA) recording photographs for 6 d, were run from June–August 2013 to July–September 2014. In 2013, over 90 carcasses (at least 30 of each herpetofauna taxa, Appendix S1) were placed at each site (n = 287). Animals were collected from the Big Island of Hawai'i and euthanized following an Institutional Animal Care and Use Committee protocol approved through the University of Georgia (A2013 04-007-Y1-A0). In 2014, 120 carcasses (30 of each mammal/bird taxa, Appendix S1) were placed at each site (n = 360). Mammal and bird carcasses were salvaged from USDA, Wildlife Services in Hilo, HI. While the euthanasia methods for species differed, all carcasses were in a predominantly intact state, and no chemicals (apart from carbon dioxide) were used.

During a trial, a single carcass was placed on a pressure-sensitive external triggering device approximately 0.3–1 m from a camera that was attached to a tree or rock a few centimeters to 0.5 m above the ground (DeVault et al. 2004). Cameras were programmed to record five (2013) and three (2014) photographs when triggered by motion or by the pressure-sensitive external trigger. In 2014, the number of photographs taken at one time was reduced to allow for additional time-lapse photographs to be taken every 15 min, more accurately documenting invertebrate scavenging and decay of larger carcasses. At each site, cameras were placed throughout the day, so as not to introduce systematic bias based on time of carcass placement. Cameras were placed at least 100 m from other baited camera traps along predetermined transects and at least 50 m from major roads (Michaud et al. 2012, Pitt et al. 2015). To confirm the independence of our carcasses, we conducted post hoc spatial clustering analyses using Moran's I statistic in R version 3.2.4 (Moran 1950, R Development Core Team 2015). No clustering of scavengers was detected within study sites, and individual carcasses were considered independent throughout analyses. After 6 d, remaining carcass material was collected and described, and evidence of scavenging was noted. Camera traps were then relocated 50–100 m away and reset with fresh carcasses. At least 30 successful trials were run for each carcass type (refers to taxa shown in Appendix S1) at each site. A trial was considered successful if the fate of the carcass could be determined through photographs or the carcass was present at the end of the trial.

To qualitatively confirm that carcasses removed from the view of the camera by vertebrate scavengers were consumed, we outfitted a separate subset of carcasses with small (2 g; 1 × 1 × 1.5 cm transmitter with a 15 cm long antenna) internally secured radiotransmitters (Advanced Telemetry Systems, Isanti, Minnesota, USA). Transmitters were placed in the esophagus with the antenna visible outside the carcass, and the mouth was safety pinned shut. Transmitters allowed us to locate carcasses that had been carried away from camera traps and record evidence of vertebrate scavenging. In August–September 2014, three cane toads were placed at each site (n = 9); carcasses of a single rat, bird, and mongoose were placed at the lava site, as well as at the scrub site (n = 6); and carcasses of a mouse and a bird were placed at the rainforest site (n = 2).

Data analysis

All photographs were examined for vertebrates, and for each new visit to a carcass, the species, time of observation, and whether or not the carcass was scavenged (i.e., fully consumed or removed from the field of view) was recorded. A visit was considered new if it occurred ≥2 min from the previous visit by the same species. If not scavenged by a vertebrate, a carcass could be removed or consumed wholly or primarily by invertebrates. To address objective (1), we calculated the percentage of carcasses across all sites that were scavenged by vertebrates vs. those that were either wholly or primarily scavenged by invertebrates and compiled a list of all vertebrate species observed scavenging carcasses.

For objectives (2) and (3), we analyzed herpetofauna carcasses separately from mammal/bird carcasses for ease of interpreting results, considering the distinctly different carcass properties of these groups. We used log-linear models to compare the count data of carcasses scavenged (2) by vertebrates vs. invertebrates and (3) among vertebrate scavenger species for different carcass types at each site. The dependent variable is the count data, that is, number of carcasses scavenged by each scavenger, and carcass type, site, and scavenger are the independent variables. Initially, we evaluated the three-way interaction of site, carcass type, and scavenger; if significant, we further evaluated each of the pertinent two-way interactions with the data sorted by the third variable (e.g., site × scavenger sorted by carcass type and scavenger × carcass type sorted by site) to clarify the underlying cause of the three-way interaction. If the two-way interaction was nonsignificant at a particular site or for a particular carcass type, we evaluated the main effects for significance. If the initial three-way interaction of site, carcass type, and scavenger was nonsignificant, we did not sort the data and evaluated two-way interactions and main effects (if two-way interactions were also nonsignificant). We conducted log-linear models using CATMOD procedure in SAS, Version 9.3 (SAS Institute, Cary, North Carolina, USA).

To calculate relative abundance at carcasses of scavenger species at each site, we summed all new visits that a single vertebrate scavenger species made to all camera traps placed within one site and divided that by the total number of visits made by all vertebrate scavengers to camera traps that year. This provided relative abundance for each vertebrate scavenger at each site, with abundances calculated separately for different years. If a scavenger species was behaviorally driven to conduct multiple visits to a carcass before or while consuming it, this approach assumes that this behavior was similar for that species across sites.

To evaluate potential differences in the total time (dependent variable) since deployment, it took vertebrate scavengers to find and to then remove each carcass type at each site, objective (4), we used two 2-way factorial ANOVAs to analyze the combined herpetofauna, mammal, and bird data. Site and carcass type were incorporated as fixed, independent variables, and the 2-way factorial ANOVA accounted for variation between and within both independent variables. If significant, we performed a Tukey's HSD test.

Results

Carcass removal and vertebrate species composition

Of our 647 experimental carcasses, 353 were scavenged by vertebrates (55%), and 294 carcasses were consumed either wholly or primarily by invertebrates (45%; Fig. 2). See Appendices S2 and S3 for the number and percentage of carcasses of each type taken by each scavenger from each site. The following species, all invasive, composed the vertebrate scavenger community: mongoose, rodents, cats (Felis catus Linnaeus), pigs, and the common myna (Acridotheres tristis Linnaeus; Fig. 3). Mongoose and rats were the only vertebrates to consume cane toad carcasses, and mongoose and pigs were the only vertebrates to consume mongoose carcasses. Species-specific proportions of scavenging by the vertebrate scavenger community across sites loosely reflected observed relative abundance at carcasses (Table 1). Mongoose and rats had the highest relative abundances at carcasses and removed the most carcasses.

Figure 2.

Percentage of carcasses removed by individual scavengers is shown separately for seven carcass types at three sites (n = 647 carcasses). A single bubble represents 10% of the carcasses of one carcass type placed at one site, and a column of bubbles represents all the carcasses of one type placed at one site (~30 carcasses per column). Bubble fractions and colors represent the percentage of carcasses removed by a particular scavenger. Data used to calculate these percentages are shown in Appendices S2 and S3.

Figure 3.

Invasive vertebrate scavengers on the Big Island of Hawai'i. Pictures show a mongoose, the most successful vertebrate scavenger, scavenging a cane toad at the rainforest site (a) and a mongoose carcass at the scrub site (b), a pig removing a mongoose at the scrub site (c), and a cat removing a mouse at the rainforest site (d).

Table 1. Number of carcasses removed by each vertebrate scavenger and relative abundance of each vertebrate scavenger at each site
ScavengerLava siteScrub siteRainforest site
Carcasses removedScavenger RACarcasses removedScavenger RACarcasses removedScavenger RA

Notes

  1. Relative abundance was calculated by summing all new visits (≥2 min apart) of a scavenger to camera traps in each site, regardless of whether carcasses had already been scavenged, divided by the total number of visits by all vertebrate scavengers to all sites that year. The last row of each data set shows the total number of carcasses removed and the visits by all scavengers.

  2. a

     The totals are the number of carcasses removed and the total number of visits by all scavengers.

Herpetofauna data—2013
Mongoose130.047330.131280.098
Rodent 0.03660.052200.602
Cat30.00410.005 0.007
Pig  20.012  
Myna  10.005  
Totala16854320048688
Mammal/bird data—2014
Mongoose590.108730.207400.132
Rodent20.012170.149350.362
Cat80.00720.00330.004
Pig 0.00170.014  
Myna      
Totala6933899985781310

While invertebrates did not set off the motion sensor for our cameras, we used photographs recorded by time lapse and the external trigger device to document the following large invertebrates, many of which are non-native species, consuming large portions of carcasses at the lava and scrub sites: yellowjackets (non-native, Family: Vespidae), cockroaches (non-native, Order: Blattodea), ants (Family: Formicidae), and fly larvae (Order: Diptera). Lone centipedes (non-native, Class: Chilopoda) and swarms of cockroaches at these sites also moved a number of carcasses outside the view of cameras (i.e., coqui frogs (5), geckos (6), mice (3), and a rat (1); Fig. 4). At the rainforest site, only fly larvae appeared to consume substantial necromass. For the majority of carcasses remaining at the end of trials at the lava and scrub sites, invertebrates had consumed all flesh, whereas at the rainforest site unconsumed flesh was still available.

Figure 4.

Trigger and time-lapse photographs of a centipede (Class: Chilopoda) removing a rat carcass at the lava site (a–d).

Influence of site and carcass type on scavenging of vertebrates vs. invertebrates

For the analysis of herpetofauna and mammal/bird data, three-way interactions were significant for both data sets, justifying our examination of two-way interactions sorted by the remaining variable to clarify the nature of the three-way interactions. When the herpetofauna data were sorted by site, there was a significant interaction between scavenger and carcass type only at the scrub site, indicating that vertebrates and invertebrates removed significantly different proportions of some carcass types at that site, but not at the lava or rainforest sites (Table 2). This result was driven by the large differences in proportional removal of frog (18% vs. 44%) and toad (56% vs. 13%) carcasses by vertebrates and invertebrates, respectively, at the scrub site (Appendix S2). In addition, there was a significant effect of scavenger at the lava field site, in the absence of an interaction between scavenger and carcass type at that site (Table 2). This main effect was driven by the much larger proportion of herpetofauna carcasses consumed by invertebrates (84%) vs. vertebrates (16%) at this site (Appendix S2).

Table 2. Results of log-linear models used to analyze three-way interactions of herpetofauna or mammal/bird data and two-way interactions of these data sorted by site or by carcass type
Interactions P Main effect
ScavengerCarcassSite

Note

  1. The interactions and main effects of scavenger, site, and carcass that significantly influenced the scavenging efficiency of vertebrates compared to invertebrates or mongoose to other vertebrate scavenger species are shown.

  2. a

     Invalid to test main effects in light of significant two-way interaction.

  3. b

     Insufficient data to test three-way interaction.

Scavenger—vertebrates vs. invertebrates
Herpetofauna data
Scavenger/carcass/site<0.01   
Scavenger/carcass
Lava siteNS<0.01NSNS
Scrub site<0.01 a a a
Rainforest siteNSNSNSNS
Scavenger/site
Frog<0.01 a a a
Gecko<0.01 a a a
Toad<0.01 a a a
Mammal/bird data
Scavenger/carcass/site<0.01   
Scavenger/carcass
Lava siteNSNSNSNS
Scrub site<0.01 a a a
Rainforest site<0.01 a a a
Scavenger/site
Mouse<0.01 a a a
Rat<0.01 a a a
BirdNS<0.01NSNS
Mongoose<0.01 a a a
Scavenger—mongoose vs. other vertebrates
Herpetofauna data
Scavenger/carcass/siteNAb   
Scavenger/carcass<0.01 a a a
Scavenger/siteNSNSNSNS
Mammal/bird data
Scavenger/carcass/siteNS   
Scavenger/carcass<0.01 a a a
Scavenger/site<0.01 a a a

When the mammal/bird data were sorted by site, there were significant interactions between scavenger and carcass type at the scrub and rainforest sites, but not at the lava site (Table 2). This result was driven largely by the large differences in proportional removal of rat (29% vs. 5%) and mongoose (19% vs. 52%) carcasses by vertebrates and invertebrates, respectively, at the scrub site, as well as the ability of vertebrate scavengers to remove higher percentages of mouse (34% vertebrate vs. 5% invertebrate), rat (32% vs. 12%), and bird (33% vs. 12%) carcasses and of invertebrates to scavenge more mongoose carcasses (71% invertebrate vs. 0% vertebrate) at the rainforest site (Appendix S3). No main effects were significant for the mammal/bird data from the lava site.

When the herpetofauna data were sorted by carcass type (frog, gecko, or toad), there were significant interactions between scavenger and site for each of the three carcass types (Table 2), indicating that the proportion of frog, gecko, or toad carcasses removed by vertebrates vs. invertebrates each varied as a function of site. This result is driven by the much larger proportions of frogs, geckos, and toads that went to invertebrates vs. vertebrates at the lava site vs. the scrub or rainforest sites (Appendix S2). When the mammal/bird data were sorted by carcass type (mouse, rat, bird, mongoose), there were significant interactions between scavenger and site for mouse, rat, and mongoose, but not for bird carcasses (Table 2). These results indicated that the proportion of mouse, rat, and mongoose carcasses that were scavenged by vertebrates vs. invertebrates differed among sites, while bird carcasses were scavenged equally by vertebrates and invertebrates across the three sites (Appendix S3). In addition, there was a significant effect of scavenger for bird carcasses, in the absence of an interaction between scavenger and site for this carcass type (Table 2). This main effect was driven by the much larger overall proportion of bird carcasses removed by vertebrates (80%) than invertebrates (16%; Appendix S3).

Vertebrate scavenger community dynamics

Due to the small number of carcasses removed by most vertebrate species, the number of carcasses removed by mongoose was compared to the carcasses removed by all other vertebrate scavengers combined. For herpetofauna carcasses, there were insufficient data to analyze the three-way interaction. Almost half of the counts for the 18 observational categories fell below the 5% threshold of the total counts (107), that is, less than five observations in each category. Given this limitation, we chose to examine the model for only the pertinent two-way interactions (scavenger × site and scavenger × carcass type) and main effects (in the absence of a significant two-way interaction) and detected a significant interaction between scavenger and carcass (Table 2). This analysis indicated that when the data were pooled across sites, the number of carcasses removed by mongoose vs. other vertebrates differed among herpetofauna carcass types, a result likely driven by the larger number of cane toads scavenged by mongoose vs. other vertebrates (Appendix S2).

For the analysis of the mammal/bird data, the three-way interaction was not significant, but interactions of both scavenger × carcass and scavenger × site were significant in the full model (Table 2). The scavenger × carcass interaction was driven by the fact that mongoose removed more of all carcass types than other vertebrates, although only slightly more mouse carcasses (Appendix S3). The scavenger × site interaction was driven by the fact that mongoose removed vastly more carcasses than other vertebrates at the lava and scrub sites but not at the rainforest site (Appendix S3).

Elapsed time for vertebrates to find and remove carcasses

For comparison purposes, time data were only used from mongoose scavengers, as they were the only scavenger to take all carcass types from each site. For herpetofauna, mammal, and bird carcasses, there was no significant difference in the amount of time that elapsed before a mongoose made an initial visit among the three sites, regardless of what ultimately scavenged the carcass. However, there was a significant main effect of carcass type on elapsed time to carcass removal (P = 0.0009), which did not always occur on the initial visit. A Tukey's HSD test revealed that significantly more time elapsed before the removal of mongoose carcasses compared to gecko and mouse carcasses, regardless of site (P = 0.0033, 0.0135, respectively). Mongoose removed 74 of 91 herpetofauna carcasses they found (81%) and 172 of 225 mammal/bird carcasses they found (76%). See Appendix S4 for the average elapsed time until mongoose found and removed each carcass type, as well as the range of elapsed time until removal and average number of visits to carcasses by a mongoose.

Confirmation of carcass consumption

Of the 17 carcasses outfitted with radiotransmitters, cameras revealed that 13 were removed and one was scavenged on camera by mongoose, one was removed by a rodent, and two were not moved or scavenged by vertebrates. Mongoose and the rodent carried carcasses 0.3–23 m away, and transmitters were recovered in the open (one carcass), heavy grass (one), under shrubs (seven), in a burrow (one), or in shallow lava caves/cracks (four). The following carcass material was found with transmitters: the head of a cane toad, the head and spine of one mongoose, and the feathers of two birds. The remaining transmitters were recovered without carcass material or were inaccessible (detected in burrows, caves, or cracks) and not recovered. This evidence suggests that when vertebrate scavengers did not fully consume carcasses in view of cameras, carcasses were moved a short distance away to a refuge where vertebrates then consumed the majority of the carcass.

Discussion

Influence of invasive species as scavengers and carcasses on ecosystem function

With the worldwide distribution of invasive facultative scavengers and the propensity for invasive species to reach high densities in areas lacking predators, such as islands, carcasses of invasive species could represent important, sustaining resources for invasive scavengers, resulting in drastically altered nutrient recycling within these ecosystems. Our study reveals that invasive vertebrates extensively utilize the carcasses of invasive species widely available on the Big Island of Hawai'i. Invasive mammals (mongoose, rats, cats, and pigs) removed 55% of carcasses before invertebrates were able to completely consume carcasses. As predicted, native Hawaiian vertebrate species capable of scavenging, such as the hawk (Buteo solitaries Peale) or owl (Asio flammeus sandwichensis Bloxam), were not observed in our study, and it is unlikely that these species significantly contribute to nutrient recycling, given their current low population densities (Klavitter et al. 2003, Pyle and Pyle 2009). Additionally while identifying invertebrates was outside the scope of this study, it was noted that many invertebrates consuming carcasses were non-native. Quantifying carcass utilization by invertebrate species will be critical to fully understanding how carcass resources are partitioned between native and invasive scavengers. Our results show that invasive vertebrates and invertebrates are utilizing a majority of the carcass resources otherwise available to native invertebrates and decomposers, redistributing resources more widely on the landscape than has historically occurred on Hawai'i and possibly contributing to positive feedback loops (Gurevitch 2006). A positive feedback loop could occur if as invasive species invade, making more carcasses available (to be utilized by both vertebrate and invertebrate invasive species), more scavenging species are also able to invade. Theoretical scavenging ecology has already begun to expand the idea of positive feedback loops (Hobbs 1996, Wipfli et al. 1999, Towne 2000, Chaloner et al. 2002, Bump et al. 2009). These data also have the potential to extend other existing theories, such as the invasional meltdown theory (Simberloff and Holle 1999). Invasional meltdowns occur when the impacts of invasive species are magnified in the presence of other invasive species due to synergistic interactions. Expanding these theories would allow scavenging behaviors of invasive species to be better incorporated into invasion ecology.

Our results and the successful invasions of many scavenging species suggest that invasion success could be tied to an ability and willingness to find and take advantage of carcass resources, even if they represent novel carcass types, such as coqui frog and cane toad carcasses in our experiment (Mateo-Tomás et al. 2015). Due to the extremely small carcass size of coqui frogs and geckos (only a few grams), it is surprising that some vertebrate species were able to find and scavenge them before invertebrates. Even though the percentages removed (22 and 39% of coqui and gecko carcasses, respectively, across all sites) were not as high as those for the other carcass types, these taxa are able to reach incredibly high densities in some areas (>91,000 adult coqui frogs/ha on Hawai'i), and with few predators on invaded islands, even these small species could provide large amounts of necromass (Beard et al. 2009). For mouse, rat, and bird carcasses, we found that vertebrates scavenged a large percentage of the carcasses, between 76 and 80%, which was comparable to or higher than in previous research conducted in natural settings with similar carcasses (see DeVault et al. 2003 for a review). Considering the potential necromass and relatively high percentage of carcasses scavenged by vertebrate invasive species, the amount of nutritive resources that carrion provides to the invasive species on Hawai'i must be substantial. Scavenging could enhance invasion success and sustain populations of invasive species, while also increasing competition for carcass resources among native invertebrate scavengers, ultimately impacting native species persistence and energy flow.

Influence of site, relative scavenger abundance, and carcass type

We determined that the proportion of carcasses scavenged by vertebrates vs. invertebrates and within the vertebrate community varied by site and carcass type on Hawai'i. These differences are likely due to factors including habitat characteristics, vertebrate and invertebrate densities, and carcass skin properties and size. We discuss these mechanisms to provide a better understanding of how ecosystem function may be differentially affected by a variety of invasive animals.

The data indicate that the three-way interaction for herpetofauna is largely driven by the scavenger × carcass interaction at the scrub site but not at the lava or rainforest sites. In addition, the data indicate that the three-way interaction for mammals/birds is driven by scavenger × carcass interactions at the scrub and rainforest sites but not at the lava site. Clearly, there are site differences influencing the proportion of carcasses scavenged by vertebrates vs. the proportion scavenged either wholly or primarily by invertebrates, as well as carcass characteristics. While determining the specific site characteristics causing this difference, for example, microhabitat characteristics, available resources, relative abundance of scavengers, was outside the scope of this study, we do present the relative abundances of scavengers as recorded by our camera traps (Table 1). In both 2013 and 2014, the higher relative abundance of vertebrate scavengers documented at the scrub and rainforest sites, as well as the decreased invertebrate carcass consumption observed at the rainforest site, could have contributed to the higher proportion of carcasses removed by vertebrates at the scrub and rainforest sites, compared to the lava site (DeVault et al. 2004, 2011).

Within each site our data indicate that carcass attributes, such as skin properties, also may have influenced whether vertebrates or invertebrates removed a carcass. For example, the highly reflective lava habitat at the lava site likely heated up quickly, causing swifter physical degradation of and bacterial activity on herpetofauna carcasses than at the scrub or rainforest sites, even though rainfall and air temperature were generally consistent across sites throughout the year (Shean et al. 1993, Archer 2004, Parmenter 2005, Sharanowski et al. 2008, Parmenter and MacMahon 2009). The skin of herpetofauna, compared to mammal/bird carcasses, has more rapid water loss and is more easily penetrated by invertebrates, resulting in faster physical degradation, especially at the lava site given the habitat (Shean et al. 1993). Invertebrates at the lava site scavenged 84% of herpetofauna, compared to 43% of mammal/bird carcasses scavenged by invertebrates at this same site the following year. Thus under the conditions on the lava field, smaller carcasses would have been available for less time. Moreover, bird data only showed a main effect of scavenger, suggesting that (regardless of site) birds were removed in similar numbers by the vertebrate scavengers. Perhaps the unique skin properties and feathers of a bird allowed them to persist in an attractive form across all sites for a similar amount of time. Given that all carcass types were found by mongoose in relatively the same amount of elapsed time at all three sites, persistence of carcasses, perhaps related to skin properties, across sites likely influenced the number and type of carcasses vertebrates were able to scavenge.

Furthermore, differences in the time to remove specific carcass types suggest that other factors, such as palatability, toxicity, and size, affected whether a vertebrate removed a carcass after it was found. The bufotoxin located in the parotoid glands of the cane toad has been shown to negatively affect cats and pigs and can remain potent several days after death, but mongoose and rats readily consume cane toads (Nellis and Everard 1983, Shine 2010). We had photographic evidence of mongoose and rats eating cane toad carcasses at the camera trap, and our transmitter trials showed that when mongoose removed cane toads they likely consumed them, entirely or at least in large part, in a refuge away from the camera trap.

Although mongoose carcasses persisted the longest, making them available for vertebrate scavenging over a greater time period, only two vertebrate species, mongoose and pigs, scavenged mongoose carcasses and tended to remove fewer of this carcass type overall. Also, the average number of visits that a mongoose made to a mongoose carcass prior to removal was greater than for any other carcass type (6.1 compared to 1.0–1.3, Appendix S4). This may reflect difficulty in handling a large carcass and/or low palatability of a mongoose carcass. At the rainforest site, the most complex habitat with presumably the most available resources, no mongoose carcasses were removed, although many were found by mongoose. This suggests that in resource-limited habitats (the lava and scrub sites) the nutritional benefits of partaking in cannibalism may outweigh the costs to mongoose scavengers. In most species, the costs of cannibalism, such as pathogen transmission, have made it a rare practice (Pfennig et al. 1998). To the knowledge of the authors, cannibalism has not previously been documented in small Indian mongoose.

Conclusions

Our study is the first to describe the vertebrate scavenger community of Hawai'i and quantify carcass removal by vertebrates and invertebrates in an island ecosystem. This study is also the first to quantify vertebrate scavenging on herpetofauna carcasses in a setting away from a road and cannibalism in small Indian mongoose. We show that the vertebrate scavenging community of Hawai'i is functionally composed of invasive species and that they can scavenge a majority of carcasses before invertebrates, many of which are likely non-native. Given the ability of some invasive animals to reach high densities in invaded island environments and a lack of predators, carcasses may be widely available. Our results suggest that novel carcass resources could influence ecosystem function, invasive species community dynamics through positive feedback loops, and possibly contribute to the process of island invasion in a manner that may extend existing theoretical models (e.g., invasional meltdown). We suggest that further research seeks to determine available invasive and native species necromass, what portion of an invasive scavenger's diet is carrion, and the population dynamics of new and established invasive animals as species continue to invade. This research would advance predictability for which species will invade and how invasive scavenger populations will respond to carcass availability throughout an invasion. Ideally, this would help manage the negative impacts of invasive species on native species and ecosystems through a heightened awareness of the influence of carcass resources (human mediated or natural) on invasion success and invasive species persistence.

Acknowledgments

We appreciate the field assistance of Shem Unger. We sincerely thank Kelton Kotake, Bob Sugihara, Dean Foster, Tom McAuliffe, and Aaron Shiels at the USDA National Wildlife Research Center Hilo Field Station for making this study possible. We thank Hawai'i Volcanoes National Park and the Natural Area Reserve System of HI DLNR for giving us a location and permit to conduct our study. This work was supported through Cooperative Agreements among the University of Georgia Research Foundation, the USDA NWRC Hilo Field Station (No. 12-7415-0936-CA), and the US Department of Energy (No. DE-FC09-07SR22506). This paper was prepared as an account of work sponsored by an agency of the US Government. Neither the US Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the US Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the US Government or any agency thereof.

Ancillary