Prototype interactive, multi-species model for invasive mammals
Tall tussock at Macraes Flat. Image – Andrea Byrom.
The drylands of Otago are a complex interactive ecosystem involving native animals dependent on highly variable plant and invertebrate food sources in a modified landscape consisting predominantly of tussock grassland, semi-improved pasture, and regenerating shrub. Within this system, native animals are threatened by introduced cats, ferrets, stoats and hedgehogs, supported at high levels through predation of introduced rabbits, hares, mice and rats. Left unmanaged, such threats are likely to drive the remaining, often fragmented populations of native animals to extinction.
Predator control by the Department of Conservation in predator-proof-fenced areas such as at Macraes Flat has clearly demonstrated how the removal of predators can prevent extinction of critically endangered native skink populations and, in some cases, lead to their restoration (see Hutcheon et al.). However, whether predator control is a sustainable management strategy for skink populations outside predator-proof-fenced areas is unclear. With highly mobile species such as cats and ferrets, achieving a sufficient spatial scale of their suppression in open landscapes is likely to be extremely costly to maintain for long periods. Also, with multiple interacting predator species, control of one, or just a subset of predators, could lead to the release of others, with potentially greater impacts on native species.
In systems such as the drylands of Otago, field surveys and trials are usually insufficient on their own to explore such issues. If management options are to be assessed in a meaningful and robust manner, a formal framework is required to gauge understanding and guide research. One such framework is mathematical modelling. The process of constructing models rapidly identifies knowledge gaps, and model simulations generate predictions for field testing. For both these reasons, Dan Tompkins and colleagues have been developing a computer simulation model for vertebrate pest communities in New Zealand’s drylands (Fig.).
The two questions initially addressed by the team are: (1) what are the important knowledge gaps in their understanding of this system, and (2) will control of single predator species lead to increased numbers of other predators? Model construction was based on data from experimental trials recently conducted at field sites near Macraes Flat and Alexandra (see Norbury et al.), complemented by other sources where necessary. Only mouse, rabbit, cat and ferret populations, and total pasture and pasture seed biomass, were modelled at this stage to make the exercise manageable. Possums, hares and sheep were included as fixed populations, with constant levels of pasture consumption assumed for these species.
Confidence in different model components and parameters was scored on a qualitative scale: ‘High confidence’ indicated data or processe s (e.g. food consumption rates) that had been robustly quantified with potential confounding factors accounted for; ‘Medium confidence’ indicated the same for data or processes derived from other systems; ‘Low confidence’ indicated data or processes (a) for which confounding factors had not been adequately accounted for, (b) which have been obtained by model fitting to observed population or community patterns, or (c) had been estimated based on expert opinion. This exercise demonstrated that even for just the core subset of species (Fig.), there is only low to medium understanding (Table). In particular, focused studies are needed to obtain (1) unconfounded life-history details for rabbits and mice, (2) accurate functional responses of the predators to their prey species, and (3) determinants of pasture seed production.
Table. Confidence in different model components
| Model component | Confidence | Notes |
| Pasture production | Medium | Herbivore-free measurements |
| Pasture consumption by rabbits | High | Functional response to pasture |
| Pasture seed production | Low | Pattern fitting to data at a different site |
| Rabbit numerical response | Low | Potential confounding factors (e.g. predators) |
| Rabbit predation by cats | Medium | Quantified allometric relationship |
| Rabbit predation by ferrets | Low | Rescaled allometric relationship |
| Mouse numerical response | Low | Potential confounding factors (e.g. predators) |
| Mouse predation by cats | Low | Rescaled allometric relationship |
| Mouse predation by ferrets | Low | Rescaled allometric relationship |
| Cat numerical response | Medium | Based on data from several other sites |
| Ferret numerical response | Medium | Based on data from several other sites |
Examples of the team’s model predictions at this preliminary stage are that the complete removal of cats (but no other predators) should lead to a slight increase (~10%) in mice and moderate increases in both rabbits (~20%) and ferrets (~20%), while complete removal of ferrets only should lead to a similar increase (~20%) in rabbits and a slight increase (~10%)in cats. However, a lot more research is needed before the model can be considered sufficiently well founded to inform conservation management in the Otago drylands.
This work was funded by the Ministry of Science and Innovation (Programmes C09X0505 and C09X0909)
