DISCUSSION
The introduction of northern quolls to Indian Island was associated with lowered survival and an apparent drop in population size in quoll-invaded melomys populations. This numerical effect on melomys density had an impact on seed predation rates, because seed take is strongly associated with the density of melomys in this system. This is a classic trophic cascade: predation suppresses herbivore density, which reduces the pressure that herbivores place on primary producers. Our study, however, also reveals an additional, subtler, cascade effect; driven by altered prey behaviour rather than by altered prey density.
Within months of quolls appearing on the island, invaded populations of melomys were measurably shyer than nearby, predator-free populations of conspecifics. This rapid but generalised response to a novel threat appears to have had a subtle effect on seed predation rates: when we examine unscented seeds, per capita seed take is slightly lower in quoll-invaded populations. This generalised response appears to have been supplemented over time with more threat-specific antipredator behaviours. Although the boldness of predator-exposed melomys converged through time with that of predator-free melomys, predator-exposed melomys continued to be more neophobic than their predator-free conspecifics throughout the study. Meanwhile, predator-scent aversion, as evidenced by seed plots, steadily increased over time. Presumably the novel predation pressure imposed by quolls resulted in selection on behaviour and/or learning in impacted rodent populations, allowing them to fine-tune their behavioural response (decrease general shyness, but maintain neophobia, and respond to specific cues) as the nature of the threat became clearer. These changing behavioural responses imply a generalised reduction in seed take that becomes fine-tuned over time, with high risk sites (those that smell of predators) ultimately displaying substantially lower seed take than low risk sites. We see the emergence of a fine-scaled aversive response (varying on a spatial scale measured in the tens of metres) and affecting per capita rates of seed predation.
Although our study documented dramatic population declines in predator-invaded melomys populations, and we are assigning the primary cause of these declines to the introduction of quolls, we acknowledge there is potential for confounding factors to affect our results. We do not believe these confounds can explain our results, however (see Supporting Information). The primary confound is the unplanned fire that burnt through northern Indian Island after completion of our population monitoring in 2017. Such fires are commonplace in the Australian wet-dry tropics (Russell-Smith & Yates 2007); a regular disturbance that is often rapidly offset by the annual monsoon driven wet season. Grassland melomys are adapted to fire in this system and populations have been shown to be very robust to its short- and long-term impacts (Griffiths & Brook 2015; Liedloff et al. 2018). For these reasons (see Supporting Information for detailed rationale), we suspect the fire was unlikely to be directly responsible for the demographic effects we observed, and fire cannot in any way explain the behavioural response we observed to quoll-scented seeds. Our interpretation of these changes as being driven mostly by the addition of a novel predator to the system is the most parsimonious and globally coherent interpretation of the data.
Predation is a pervasive selective force in most natural systems, driving evolutionary change in prey morphology, physiology, life history and behaviour. Unlike morphology and physiology, however, the labile nature of behaviour makes it a particularly powerful trait for rapid response in a changing world (Réale et al. 2007; Sih et al. 2010b; Dall & Griffith 2014). Behavioural comparisons of wild populations exposed to differing predation regimes provides some support for the prediction that reduced boldness would be selected for under high predation scenarios (Åbjörnsson et al. 2004; Bell 2005; Brydges et al. 2008) and that the appearance of novel predators can result in bold individuals becoming shyer (Niemelä et al.2012). The opposite pattern or a non-response can also occur, however (Brown et al. 2005; Urban 2007)(Laurila 2000; Carlson & Langkilde 2014). Interestingly, a number of studies have demonstrated that individuals from high-predation areas were quicker to emerge (Harris et al. 2010) and were bolder and more aggressive (Bell & Sih 2007; Dingemanse et al. 2007) than predator-naïve conspecifics. Although we found the opposite pattern to this immediately following the arrival of a novel predator, by the second year after predator introduction we found the boldness of melomys converging with that of predator-free populations. Thus, it is clear that the behavioural composition of these populations is dynamic, and it seems likely this dynamism (and perhaps the capacity of the prey species to identify specific threats) may explain some of the variation between earlier studies.
Although boldness may change over time, neophobia, as a generalised adaptive response to predation pressure, is now well supported across a number of studies (Crane et al. 2019). Individuals living under high predation risk scenarios have been shown to typically display generalized neophobia (Brown et al. 2015; Elvidge et al.2016), and neophobia can increase the survival of predator-naïve individuals in initial encounters with predators (Ferrari et al.2015; Crane et al. 2018). Certainly, in our study, predator-exposed melomys were significantly more neophobic than their predator-free conspecifics; an effect maintained throughout the study.
Despite reduced survival, significant population declines, and clear behavioural changes in invaded populations, it is impossible to determine from our data whether changes in the behaviour of predator-invaded melomys populations are the result phenotypic plasticity (learning) or natural selection. The low between trapping session survival of melomys in quoll-invaded populations means few individuals survive between sessions, so natural selection is a possibility, and selection on these behavioural traits is potentially very strong. Although behavioural changes in predator-invaded populations have been documented in a few systems where predator introductions have been staged and experimentally controlled (Lapiedraet al. 2018; Blumstein et al. 2019; Cunningham et al. 2019; Pringle et al. 2019), elucidating whether these observed changes arise because of behavioural plasticity or natural selection can be exceptionally difficult. Rapid behavioural responses of vulnerable prey to recovered predators has been observed in a single prey generation, presumably due to behavioural plasticity (Bergeret al. 2001; Cunningham et al. 2019). Similarly, behavioural adjustments to an introduced predator have been observed as a result of natural selection on advantageous behavioural traits (Lapiedra et al. 2018). In our study we had measures of individual behaviour, but our between session recapture rates of these individuals was too low to test whether individuals were altering their behaviour or whether natural selection was resulting in population-level change. It thus remains possible (and quite likely) that both mechanisms were in play.
Although northern quolls represent a novel predator to melomys on Indian Island, the two species’ shared evolutionary history on the northern Australian mainland may provide some explanation as to why this staged introduction resulted in rapid, finely-tuned behavioural adjustment in melomys, rather than extinction. Isolation from predators can result in rapid loss of antipredator behaviours from a prey species’ behavioural repertoire (Blumstein & Daniel 2005; Jolly et al. 2018a), dramatically increasing an individual’s susceptibility to predation following the introduction of either predator or prey (Carthey & Banks 2014; Jolly et al. 2018b). But such outcomes are not inevitable: length of isolation, co-evolutionary history, degree of predator novelty, density-dependent effects, population size, and pre-existing predator-prey associations (Berger et al. 2001; Blumstein 2006; Banks & Dickman 2007; Sih et al. 2010a; Carthey & Banks 2014) are all likely hugely influential in determining whether an invaded population adjusts to the invader or proceeds towards extinction. Recently, a conservation introduction of Tasmanian devils to an island previously lacking them found that their possum prey rapidly adjusted their foraging behaviour to accommodate this newly arrived predator (Cunningham et al. 2019). Despite possums having lived on the island in isolation from devils since the 1950s, presumably, their long evolutionary history together on mainland Tasmania had them primed to respond to this predatory archetype (Sih et al. 2010a; Carthey & Banks 2014; Cunningham et al. 2019). This shared evolutionary history is likely responsible for both possums’ and melomys’ ability to rapidly mount appropriate antipredator responses. The predators are novel within an individual’s lifetime, but the individual’s ancestors have encountered them before.
Although our results suggest that invaded melomys populations are beginning to adjust to the presence of northern quolls, there has been no sign of demographic recovery on the island. Data from our seed removal experiment clearly demonstrated that the function of melomys as seed harvesters and dispersers scales with density. Trophic cascades resulting from the addition and loss of predators from ecosystems has been observed in a number of systems globally (Ripple et al.2001; Terborgh et al. 2001; Estes et al. 2011), and the results can profoundly shape entire systems. As the only rodent and the dominant granivore in this system, while melomys populations may or may not go extinct as a result of quoll invasion, their reduced abundance and weakened ability to harvest and disperse seeds may have yet to be observed, longer-term consequences for the vegetation structure and ecosystem function of Indian Island (McConkey & O’Farrill 2016). Currently, grass is a rare vegetation feature on Indian island (though it is a dominant feature of nearby savanna woodlands on the mainland), and this is quite possibly a result of the high density of melomys on this (previously) predator-free island. The presence of quolls may well change that, as both numerical and behaviour responses of melomys cascade down to the grass community.
Empirical research on the effects of novel predators on recipient communities under controlled conditions on a landscape-scale is exceptionally difficult and remains relatively rare. The introduction of threatened predators to landscapes from which they have been lost (Cunningham et al. 2019) or where they are entirely novel (Lapiedra et al. 2018), however, provides a unique opportunity to observe how naïve prey can respond to novel predators, and the mechanisms by which predators can structure communities. Our study provides empirical support that some impacted prey populations can adjust rapidly to the arrival of a novel predator via a generalised behavioural response (decreased boldness) followed by development of a species-specific antipredator response (behavioural fine-tuning). The arrival of the novel predator appears to have set off a trophic cascade that was driven, not only by changed prey density, but also by changed prey behaviour. Thus, rapid adaptive shift may allow prey populations to persist, but large-scale, system-wide changes may still follow.