Non-native mammals are weak candidates to substitute ecological function of native avian seed-dispersers in an island ecosystem

Although prominent examples exist of non-native species causing substantial ecological harm, many have neutral or positive effects, including filling surrogate roles once performed by extinct native organisms. We tested the ecological roles of two non-native mammals as seed dispersers or seed predators in Guåhan, which, due to invasive brown treesnakes (Boiga irregularis), is devoid of native seed dispersers – birds and bats. We conducted feeding trials with captive rats (Rattus spp.), which are present but uncommon due to predation by snakes, and pigs (Sus scrofa), which are abundant. We examined if and how they interacted with common forest fruits. We then compared how any gut-passed or handled seeds germinated compared to seeds left in whole fruit or depulped seeds.

Rats and pigs interacted with most of the fruits and seeds (>80%) that they were fed. Of those, most seeds were destroyed – 78% for rats and 90% for pigs, across both native and non-native plant species. Compared to seeds germinating within whole fruits, rats improved germination of the seeds that they handled without ingesting, while pigs diminished the germination of seeds that they handled. The small percentage of seeds (approximately 1.5% for rats and 5% for pigs) that survived gut passage germinated in higher proportions than those in whole fruits. Percentages of seed survival to germination are lower than those found in similar studies with native avian frugivores. Our results indicate that pigs and rats have mixed effects on seeds, but are not suitable surrogates for native seed dispersers.

Methods
Fruit collection
We collected common fruit species present in Guåhan’s limestone forests to use in feeding and germination trials. This included both native and non-native fleshy-fruited plant species varying in fruit and seed size and in growth form (trees, shrubs, and vines) (Table 1). Most species were collected on the day of the feeding trials. In cases when these fruits were not ripe at the time of picking, Morinda citrifolia (ladda) and Premna serratifolia (åhgao) were held for up to four days to allow for ripening before feeding. Fruits of a single species were picked from different sites and at least three different individual trees per site to account for individual effects. Fruits were mixed and divided into four groups: 1) a depulped treatment (to approximate flesh-removal without digestion) where we extracted seeds from the fruit by hand, 2) a whole-fruit treatment (representing fruits falling to the ground in the wild), where fruits were left intact, and two treatments, 3) seeds that had passed through the gut or 4) had been spat out by rats or pigs, that were determined after being fed to animals. Fruits were fed to captive animals within a mix of multiple species collected that day, or less commonly, as a single species.

Rat feeding trials – immediate seed fate
Rat feeding trials occurred from July 2017 until February 2018. Rats were captured from the wild in Guåhan and kept in a captive outdoor facility. After collecting fruits, we reserved some for whole fruit and depulped treatments, while the rest of the fruits were fed to rats. We fed rats in the late afternoon or evening.  As rats normally did not begin feeding immediately, fruits were left overnight. Based on preliminary trials, any ingested seeds were passed within 12 hours. All remaining fruits and seeds, and all scats were collected the following morning and again later in the day to allow enough time for gut passage. Seeds that had flesh removed, which we considered “rat-handled,” were set aside to be planted as one of our germination treatments. Scats were picked apart under a dissecting scope to locate all gut-passed seeds, and, if they appeared mostly intact, these were also set aside to be planted as our gut-passed germination treatment.

Pig feeding trials – immediate seed fate
Pig feeding trials occurred between July 2018 and May 2019. Again, a portion of fruits and seeds were set aside for whole fruit and depulped germination treatments. The remaining fruits were placed in pens in feeding bowls or in spots where each pig’s food was normally placed. Because feeding trials were coordinated with private landowners, timing of feeding depended on when we had permission to access their pigs. However, all feeding trials were done in the afternoon or early evening. We left fruits with pigs for approximately one hour while we remained out of sight before collecting any remnant fruits. Seeds that were spat out and partially-eaten fruits were set aside for our “pig-handled” treatment. We collected scat once daily until seeds of those species ceased to appear in scats, usually four to five days, which was consistent with gut passage time in other studies (Elston, et al., 2005; Schmidt, et al., 2004). Scats were processed by spraying water through size-appropriate mesh sieves. Any intact or mostly intact seeds caught in the mesh were identified, counted, and then planted in the nursery within one day as our gut-passed treatment. Scats collected in the days after feeding the very small-seeded species of Ficus prolixa and Pipturus argenteus were placed intact in soil mix and germination trays, and the entire scat was watered and monitored for germination. 

We fed 14 species of fruits to both rats and pigs. An additional four species were fed only to rats, and four species only to pigs due to differences in fruit availability during the respective feeding trials (Table 1).

Germination Trials
We planted seeds as one of four treatments, as mentioned above and repeated here: 1) a depulped treatment (to approximate flesh-removal without digestion) where we extracted seeds from the fruit by hand, 2) a whole-fruit treatment (representing fruits falling to the ground in the wild), where fruits were left intact, and two treatments, 3) seeds that had passed through the gut or 4) had been spat out by rats or pigs, that were determined after being fed to animals. While preparing seeds for our depulped treatment, we also selected several fruits from which to tally seed counts – this allowed us to estimate average seed counts for whole fruits with multiple seeds. If whole fruit treatments did not fit in seedling trays, as in the case of larger fruits with many seeds (Carica papaya, Morinda citrifolia) which normally break when they fall to the ground in the wild, we divided the fruit into chunks and planted each chunk separately. 

We conducted germination trials in an open-air shadehouse covered by 70 percent shadecloth. We sowed seeds from all treatments in a 1:1 mix of peat moss and perlite and watered them daily, and recorded germination three times a week. When a seedling sprouted in a cell, we took one of two courses of action: 1) if only one seedling was expected to grow from the cell, then the seedling and all contents of that cell were discarded or 2) if the cell contained additional seeds that could germinate, the seedling was clipped and other contents left undisturbed. Germination was monitored from July 2017 until June 2018 for treatments planted with rat feeding trials, and from July 2018 until November 2019 for treatments planted with pig feeding trials. Based on knowledge from previous germination studies (Fricke et al., 2018), we did not expect most species to be viable after six months, and therefore, discontinued our monitoring of these species if they had not germinated within seven months. One species, Elaeocarpus joga (yoga), can germinate after more than a year, so we monitored this species for 14 months after planting.

Analyses Overview
We ran separate parallel analyses for pig trials and for rat trials. All analyses were run in R (R Core Team 2020) using the lme4 (Bates et al. 2015) and emmeans (Lenth, et al., 2020) packages. Each analysis consisted of three separate steps (Figure 1). First, we analyzed whether an animal interacted with a fruit. Second, for the fruits with which animals interacted, we analyzed whether seeds were handled (cleaned of flesh), ingested and passed, or destroyed. Finally, we analyzed germination, comparing germination success between separate animal treatments (gut passed by an animal or handled by an animal), with seeds that were depulped by us, and those planted in whole fruit. In addition, we estimated the overall cumulative effect of each animal species on each plant species by multiplying the probability of a seed reaching each subsequent step in the feeding to germination process.

Immediate seed fate analysis
We used a generalized linear mixed-effects model (GLMM) with a binomial error distribution to determine if animals ignored or interacted with fruit. The response variable combined the number of “successes” (fruits interacted with) and the number of failures (fruits ignored), using the cbind function in R. The fixed effect was the fruit species, and the random effect was the individual identity of the pig or rat to account for individual variation between animals, and the date of feeding trial, to account for the variation in viability of fruits between collection bouts. 

We ran further analyses using the subset of species with which pigs or rats interacted to determine what type of interaction an animal had with a seed. For species that were gut passed (which tended to be smaller-seeded), we used GLMMs with binomial error distributions to determine if seeds were gut passed or destroyed by rats and pigs. For species that were handled (which tended to be larger-seeded), we again used GLMMs with binomial error distributions to determine if seeds were handled or destroyed by rats or pigs. For each GLMM, the response variable was the number of seeds gut passed or the number of seeds spat out combined with the number of seeds destroyed, again using the cbind function; the fixed effect was fruit species, and the random effect was individual pig or rat. There was almost no overlap between gut-passed and handled seeds for either pigs or rats. Only one species, Carica papaya, had seeds that were both gut-passed and handled by rats, so we used a GLMM to determine how many seeds were either gut-passed or handled and used these respective numbers in the models above. We did not conduct post-hoc tests, as we were not interested in the differences between species, but instead report the predicted proportion of seeds that were gut passed or handled relative to the proportion destroyed for each species.

Germination analysis 
We used separate generalized linear models (GLMs) to compare germination proportions between animal-treated seeds (gut-passed or handled) and control seeds (planted in whole fruit or depulped). Similar to the immediate seed fate analysis, this analysis was split into plant species where seeds passed through the gut intact or species where seeds were handled and not passed. We used the emmeans package (Lenth et al., 2020) to determine if treatment (animal handled or gut-passed) significantly improved germination over the other two treatments (whole fruit or depulped).

Overall effects
We estimated overall frugivore effects on seeds by sequentially combining our model-predicted proportions for fruit fate, immediate seed fate, and germination. To do this, we multiplied predicted proportions, from GLMMs described above, across each step for each plant species to get a cumulative proportion of seeds that germinated out of total seeds fed to each animal.