Invasive species-driven trophic cascades: Are cane toads indirectly contributing to small mammal collapses across tropical Australia?

Apex predators are fundamentally important in regulating many ecosystems, and perturbations of their populations are frequently implicated in ecosystem declines or collapses. In considering small mammal declines in northern Australia, most attention has focused on interactions between a mammalian apex predator—the dingo Canis dingo—and a meso-predator, the feral cat—Felis catus. Little consideration has been given to the possible implications of changed reptilian predator assemblages resulting from invasion by a toxic anuran invader, the cane toad (Rhinella marina), on small mammals. We used reptile removal records from licenced reptile catchers in three widely spaced towns in the savannas of northern Australia to explore potential impacts of toads on apex and meso-predatory snakes and large lizards. In addition, simultaneous fauna survey data from one town with reptile removal records, coinciding with toad invasion, were used to identify cascading impacts through the savanna ecosystem. Intervention analyses revealed empirical linkages between toad invasion, apex predator declines, meso-predator increases and declines of small mammals and other prey groups. Based on the timing and strength of intervention we postulate a novel conceptual model linking recent mammal declines with trophic cascades following toad invasion, where the loss of large, anurophagous (toad-eating) reptilian apex predators allowed increases in mammal-eating meso-predatory snakes. The conceptual model is discussed in relation to prevailing hypotheses regarding northern Australia’s dramatic small mammal declines. Future studies will need to quantify these putative interactions and test their comparative importance so that appropriate management can be implemented to stem the ongoing losses of mammal fauna.


INTRODUCTION
arrival of the cane toad [39][40]. Quolls actively hunt and ingest toads and are subject to high 116 rates of mortality due to lethal poisoning [40][41]. However, declines among other small 117 mammals have not been linked to toad arrival. This is because most mammals do not eat cane 118 toads, or escape poisoning by avoiding the toxic glands [40]. Similarly, small mammals are 119 not known to be eaten by cane toads [39]. (partially anurophagous and reptile-eating) elapid snakes such as the king brown snake 129 (Pseudechis australis) and varanid lizards (from here referred to as apex predators) as seen 130 during previous studies [15,39,42]. We interpret these collapses as due to poisoning upon 131 ingestion of toxic cane toads [39,15]. Another change at study sites was an immediate 132 increase among smaller-gaped, smaller-bodied (i.e. meso-predatory), dietary specialist 133 snakes, lizards and anurans, including mammal-eating pythonid and cobubrid snakes. 134 Increases among meso-predatory reptiles was also previously reported in several studies [11-135 13, 19]. We interpret increases as due to a meso-predator release following loss of large 136 generalist apex reptilian predators [15]. Additional changes associated with cane toad 137 invasion at one of our study sites where we had continuous fauna monitoring data were 138 declines among fauna which is prey to meso-predatory reptiles. Prey included small 139 mammals, very small skinks (< 8 cm long) and many invertebrates. Declines among 140 invertebrates following toad invasion has been reported before [43], though this was 141 interpreted as due to an increase in predator biomass due to cane toad presence, not due to 142 general increases among a range of meso-predators. We interpret declines among prey groups 143 as being driven by increases among their meso-predators, including mammal-and lizard-144 eating snakes, medium-sized lizards and frogs, as these species all increased following cane 145 toad invasion. We synthesise these empirical observations in the form of a conceptual model 146 that articulates the trophic links between small mammals, cane toads and reptilian predators 147 (Fig. 1). This provides us with a complementary hypothesis to the cat -fire/disturbance 148 driven hypothesis that has dominated the literature on north Australian mammal declines over  Study areas 153 The main study location was the town of Kununurra (2016 census population 5,300) and its 154 surrounding savanna landscapes including Mirima National Park in far north-eastern Western 155 Australia (Fig. 2a, b). The region has a tropical monsoonal climate, with high temperatures 156 year-round (daily mean maximum 29.6-36.0 o C), and rainfall (913 mm annually) occurring 157 predominantly from November to April. Several tropical savanna habitats occur around 158 Kununurra. Aside from urban and agricultural (broad-acre cropping) habitats, these include 159 black soil plains, eucalypt woodlands dominated by tussock grasses, pindan (Acacia tumida) 160 savanna woodlands dominated by Triodia hummock grasses and annual Sorghum on 161 sandplain, and shrub/Triodia spp. dominated woodland on rocky sandstone. Kununurra is 162 adjacent to perennial riparian habitats and permanent water due to the damming of the Ord 163 River (Fig. 2a, b). Minor study locations at Katherine (popn. 6,300) and Darwin (popn.   (Fig. 2b). Wildlife officers are compelled to attend snake callouts for reasons of public safety, so data can be considered representative 177 of snake occurrences in the town. Records included date and time of removal, the officer's 178 name who attended, the location/address, the species and size (length) of the animal removed. 179 Data are presented as monthly counts for analysis. In Katherine, snake removal records were 180 available from 1998 to 2008 and covered the pre-invasion (1998)(1999)(2000) to post-cane toad

267
Reptile removal and fauna survey data 268 Apex predatory species, including five species of elapid snakes (n = 364) and six species of 269 varanids (n = 42); meso-predatory species/taxa, including 23 snake (n = 6584), nine lizard (n = 561) and seven frog species (n = 487); and prey species/taxa, including five mammals (n = 271 104), six lizards (n = 599) and six invertebrate taxa (n = 1221) were recorded during 272 removals and surveys (Table S2) Responses to toad invasion among predator and prey groups 282 As predicted under the conceptual model ( Fig. 1), apex predators declined significantly after 283 cane toad invasion, almost all meso-predators increased, and most prey groups -including 284 small mammals -also declined based on intervention tests ( indicating a strong impact of toad invasion relative to other explanatory variables (Table 1).

287
The strongest additional explanatory variable for apex predators was a 6 month temporal  (Table 1).

292
Four of the five identified meso-predator groups, including the mammal-eating pythons ( Fig.   294 3b), skink-eating elapids, colubrid snakes and combined frogs, agamids and large skinks ( Fig.   295 4b), showed significant increasing intervention responses (Table 1). Intervention responses 296 had higher coefficient estimates than for all other explanatory variables for meso-predators, 297 indicating that toad invasion was the strongest predictor of change among these groups 298 ( Table 1). All three meso-predatory groups from Katherine and Darwin (mammal-eating 299 pythons, skink-eating elapids and colubrid snakes) also had significant intervention responses 300 (  Table S2) which did not show a significant intervention response (Table 1). In addition to 312 intervention, the strongest explanatory variables for small mammals were rainfall in the 313 previous 2 months, and months since fire (Table 1). Herbivorous invertebrates and small 314 skinks responded most strongly to rainfall in the previous month, seasonal auto-correlation 315 and vegetation cover (Table 1). Large carnivorous invertebrates responded most strongly to seasonal auto-correlation, maximum temperature and rainfall in the previous month (Table   317 1).

319
Species-specific apex, meso-predator and prey responses generally conformed to responses 320 predicted in the conceptual model ( Fig. 1) with a few minor exceptions (Table S4). Among  (Table S4). Among small skinks, Carlia spp. showed a positive 326 intervention response which was the opposite to small skink responses overall (Table S4). 327 Carlia spp. were the largest among the small skinks (ca. 11 cm) with others in the group < 10 328 cm long (Table S2).     than for cats and also much smaller home ranges. This means that even if snake ingestion 421 rates are much lower than for cats, they may cause comparable overall predation pressure. 422 We know that snakes in some cases can have very large impacts on mammalian and avian 423 assemblages. These include one meso-predatory snake from this study (e.g. B. irregularis) 424 [16]. In addition, there is information from a cat exclosure experiment in northern Australia

425
[35] that showed similar predation by pythons on savanna rodents to that by cats. However, 426 the hypothesis that reptilian predation could be equivalent to that of feral cats, and the 427 possibility that cumulative impacts could be substantial, needs to be tested more widely 428 across Australian savanna landscapes if we are to establish its plausibility in playing part in 429 regional mammal declines.             Significance codes . 0.1, * 0.05, ** 0.01, *** 0.001 Table S2. Snake, reptile and mammal species represented in callout and survey records from Kununurra (KNX, Mirima), Katherine (KAT) and Darwin (DAR). Fauna species and taxa are grouped according to predatory hierarchy (apex-and meso-predators and prey) and functional groupings within the conceptual model (see Fig. X). Numbers in bold are those for which intervention analyses were undertaken. Dietary information for reptile and frog species/genera was taken from Cogger (2014) Table S4. Time series generalised linear models (INGARCH log-linear models with logarithmic link) from occurrence records of species/functional/trophic groups of apex, meso-predators and prey trophic levels at different sites. Model coefficient estimates, standard error (SE) and 95% confidence intervals (CI) are shown. Intervention effect at the time of cane toad invasion (and/or at a lagged time after cane toad invasion) is indicated as significantly negative (↓), significantly positive (↑), or non-significant (NS) and the time of intervention effect is given. For definitions of variables see Table S2.  Meso-predatory reptiles include mammal-eating snakes such as pythons (e.g. Aspidites melanocephalus and Liasis olivaceus), colubrid snakes (e.g. Dendrelaphus punctulatus) and small skink-eating elapids (e.g. Furina ornata). Other savanna meso-predators include frogs (e.g. Platyplectrum ornatum), large scincids (e.g. Ctenotus robustus) and agamids (Lophognathus gilberti). Arrow thickness represents the strength of the interactions between trophic levels (apex and meso-predators and prey species). Thin, dashed lines or arrows indicate putative declines or weakened interactions. Violet arrows/lines represent interactions with the invasive cane toad, red lines/arrows represent apex predators and their interactions, blue lines/arrows meso-predators and their interactions, and green lines are key savanna prey species/groups. a) Represents a conceptual model of trophic interactions in savanna ecosystems prior to cane toad invasion. Pre-invasion reptilian and amphibian assemblages were dominated by the apex predators, which were the large-gaped anurophagous/generalist reptiles, which suppressed many of the meso-predatory savanna species, including reptilian, amphibian and mammal species. In this pre-invasion ecosystem, prey groups including small mammals, small skinks and invertebrates persisted at moderate abundance. b) Shows how these interactions are predicted to alter following cane toad invasion. With the loss of ca. 80% of the large, anurophagous/generalist apex reptilian predators, meso-predatory snakes, frogs, skinks and agamids increased by ca. 250 % and cane toads were introduced as an additional meso-predator. Under this scenario, there was increased predation pressure on prey groups including small mammals, small skinks and some invertebrates (herbivorous) which resulted in declines in these groups of ca. 30-80%. Note that large predatory invertebrates including carabid beetles and centipedes neither declined nor increased following cane toad invasion.