Biodiversity publications archive

Biodiversity and Fire: The effects and effectiveness of fire management

Proceedings of the conference held 8-9 October 1994, Footscray, Melbourne
Biodiversity Series, Paper No. 8

Biodiversity Unit
Department of the Environment, Sport and Territories, 1996

12. Assessment of fire regime options for the southern brown bandicoot Isoodon Obesulus in South Australia using population viability analysis

Hugh Possingham
Department of Applied Mathematics, The University of Adelaide, and
Brian Gepp
Department of Primary Industry (S.A. Forestry)

12.1 Abstract

One of the important uses of predictive modelling is to assess options for the management of the environment that are not amenable to experimentation. The extinction of a species area is one environmental property that cannot be experimented with, consequently we need a population modelling tool for predicting the consequences of different management options. In this paper we illustrate the use of population viability analysis (PVA) for assessing fire management options for the southern brown bandicoot Isoodon obesulus.

The southern brown bandicoot is a threatened species in South Australia (SA). In the south east of SA it occurs in about thirty remnants of native vegetation, the largest of which is Honans Native Forest Reserve. Honans is managed by the Forestry group of the SA Department of Primary Industry and one of the objectives of their management is to retain the population of bandicoots.

The principle management tool used in Honans is prescribed burning. These prescribed burns are carried out in about twenty different sections of Honans, called blocks. We use a population modelling package, ALEX, to explore the following issues. What is the relationship between control burn interval (per block) and extinction probability? What effect does changing the number of blocks have on extinction probability?

Key words: fire regime, risk assessment, brown bandicoot, Isoodon obesulus, South Australia, population viability, extinction, threatened species, mathematical modelling

12.2 Introduction

Population Viability Analysis (PVA) is a process in which the likelihood that a population will become extinct is assessed, within a specified period (say 100 years) and under particular circumstances (Shaffer 1981). This process usually involves the use of mathematical models that are explored using a computer simulation. Numerous authors have recently reviewed PVA (Burgman et al. 1988; Shaffer 1990; Possingham 1991; Akçakaya 1992; Boyce 1992; Burgman et al. 1993; Lindenmayer et al. 1993) and now there are many papers in the readily available literature that describe applications of PVA. In this paper we illustrate the application of PVA for assessing fire management options for a threatened population. Before examining the specific problem we briefly discuss issues concerning the utility of PVA.

Determining the probability that a species will become extinct is an interesting academic exercise that is useful for ranking threatened species in lists and thereby directing conservation effort, but has little other value for resource management and applied conservation biology. In this paper we emphasise that PVA is a tool that allows us to choose between different management options (two recent examples are Armbruster and Lande 1993 and Haig et al. 1993). In particular, given several management options, our primary objective should be to determine which option minimises extinction probability. The relative merits of the management options are important, not the absolute level of risk (Lindenmayer & Possingham 1995).

By carrying out a PVA we are forced to make explicit statements about the parameters and processes that influence the dynamics, and hence the viability, of a population. Although information about parameters and processes may be scanty, by modelling the dynamics of a population we expose those processes that we know little about, and test our intuition.

Information about threatened species will almost always be inadequate, so it is essential to assess the sensitivity of our results to variation in uncertain parameters and processes. We often find that our conclusions, in terms of ranking management options, are insensitive to some parameters but very sensitive to others. Where our management decisions are sensitive to a parameter or a process this means that more field work needs to be done to reduce that uncertainty (Boyce 1992). There will always be a need for further research, however we must also be prepared to make decisions on the basis of existing information and modelling.

No model is exactly correct, it is only a tool that educates us, adds rigour to our thinking, helps us to choose between management options, and allows us to extrapolate our knowledge over large spatial and temporal scales. The process of population modelling, management and data collection must proceed hand-in-hand. Collaboration between the mathematical modeller and wildlife experts is essential in all cases (Possingham et al. 1993).

Although this paper is concerned with a specific species, in a particular place, and particular management questions, the approach taken can be applied to any threatened species management program that involves fire. To make fire management decisions for a species some of the fire-related information we will need to know will include:

  1. the direct impact of different sorts of fire on animal mortality,
  2. the impact of fire on habitat quality, in time and space, and
  3. the relationship between fire regime, habitat dynamics and the demographic parameters of the species concerned.

12.2.1 Background

Although the southern brown bandicoot is not nationally endangered, the species has disappeared or declined over much of its range. In the south-east of SA it occurs in about 30 patches of remnant vegetation ranging from 5 ha to 1500 ha (Paull 1992). One of the most critical patches of habitat for the species is Honans Native Forest Reserve. Honans is managed by the Forestry group of the SA Department of Primary Industry and one of the objectives of this management is to retain the population of bandicoots. The total area of 1000 hectares is divided into about 25 blocks each of about 40 ha. In the management plan for the area, three of the central blocks are reserved from to prescribed burning, while the remainder are burnt on a staggered ten to 14 year rotation. Hence each block is burnt every twelve years but adjacent blocks are burnt in different years, preferably three to five years apart to facilitate recolonisation.

The southern brown bandicoot appears to favour mid-successional habitat in south-eastern SA. Although fires kill some individuals, and habitat immediately after fire is not suitable, habitat that is not burnt for a long time has low quality. Only at the mid-successional stage, when the heath understorey is dense, does habitat quality reach its peak (Heinsohn 1966; Paull 1992). Paull's (1992) work suggests that patch quality after a fire is low, from where it increases to a peak between five and seven years, after which it slowly declines to a low level (figure 12.1).

Figure 12.1: The relationship between habitat quality and the time since fire for the southern bandicoot in south-east South Australia used in the PVA model ALEX for simulations of this species.

Figure 12.1: The relationship between habitat quality and the time since fire for the southern bandicoot in south-east South Australia used in the PVA model ALEX for simulations of this species.

In this study we focus on two aspects of the fire regime – first the interval between burns and second the number of different blocks into which Honans is divided, and therefore managed.

12.3 Methods

The parameters we use in the model are described in Possingham and Gepp (1983), while the mechanics of the PVA model, ALEX, are described in detail in Possingham et al. (1992) and Possingham and Davies (1995). Here we briefly review the most important model assumptions and basic parameters.

ALEX models the viability of metapopulations. The model only includes information on females and does not model the genetics of the population (compare with VORTEX, Lacy 1993). Although this limits its applicability to very small populations, ALEX can model a fixed interval prescribed burn and allow for complex habitat dynamics. These two features are essential for this example.

Using an aerial photograph, the position and size of the 26 blocks in Honans was entered into ALEX. Connections facilitating dispersal between abutting patches were included. In each patch the time since last fire is tracked, so a fixed-interval prescribed burn can be imposed on the system. This time since last fire is allowed to influence the fecundity of the bandicoots according to the relationship in Figure 1. The prescribed burns were set so neighbouring blocks were not burnt in the same year. Life-history data was estimated from the extant literature and existing trapping data. Most of the animals are assumed to die in a prescribed burn; this may be a pessimistic assumption depending on the intensity of the fire and the extent of the understorey burnt.

Experiment 1: In this experiment we vary the interval between prescribed burns.

Experiment 2: By relaxing the restriction that neighbouring blocks are not burnt in the same year we test the impact of the number of blocks on which the scrub is managed given a burn interval of 11years. In one extreme all blocks are burnt simultaneously, at the other extreme they are spatially staggered. Intermediate options were trialed where the number of separate groupings of blocks was two, three, six and 12.

All experiments were run over 300 times so the standard deviation associated with the extinction probabilities presented below is at most 3 per cent.

12.4 Results

Experiment 1: The burn interval that minimises the likelihood of extinction is 15 years (Figure 12.2). Burn intervals of 11 to 17 years are only slightly worse than the optimum.

Based on the habitat quality dynamics (Figure 12.1) we may have expected the optimal interval to be around seven years, so this result needs some explanation. First note that, ignoring the death rate caused by the burn, we want to have a high average habitat quality, not maximise the amount of time at the highest habitat quality, which means the burn needs to be after the peak in habitat quality. Second, a control burn causes a dramatic local decline in bandicoot density because we assume that most of the bandicoots are killed during the prescribed burn and habitat is very poor for one or two years after a fire. By burning well after the peak in habitat quality fluctuations in population size are reduced while the average habitat quality is still reasonable.

Figure 12.2: Extinction probability within 100 years of the southern brown bandicoot in a 1000 ha patch of scrub )Honans) in the south east of South Australia as a function of the interval between control burns.

Figure 12.2: Extinction probability within 100 years of the southern brown bandicoot in a 1000 ha patch of scrub )Honans) in the south east of South Australia as a function of the interval between control burns.

Experiment 2: In Figure 12.3 we see that dividing the scrub into more patches, where each patch is burnt independently, increases population viability. However as long as there are at least three different patches the extinction probability is close to its minimum. This is an important conclusion. In mosaic burning, where we are trying to provide a range of habitats, more complex patterns may be more expensive to manage, especially in small patches of scrub like this. The result in Figure 12.3 tells us that Honans really only needs to be divided in to three blocks as far as bandicoot viability is concerned. This result may depend in part on the presence of a central unburnt core which provides a node for recolonisation.

Figure 12.3: The relationship between the number of fire management blocks in Honans and the extinction probability of the population of southern brown bandicoots with prescribed burn interval of 11 years.

Figure 12.3: The relationship between the number of fire management blocks in Honans and the extinction probability of the population of southern brown bandicoots with prescribed burn interval of 11 years.

By carrying out some limited sensitivity analysis we were able to draw some additional conclusions.

  1. If the number of bandicoots killed in a prescribed burn is lower, or the prescribed burns only affect about half a block (that is they are patchy) the optimal burn interval is lower.
  2. The presence of the central unburnt core had no effect on the results except where the number of blocks was small.
  3. By changing other parameters, like dispersal rates and birth and death rates, the absolute probability of extinction varies significantly but the minimum extinction probability generally occurs with a control burn interval around 15years

12.5 Discussion

The scenarios presented in this paper raises the question: what are we managing for in an isolated habitat island? Existing philosophies of wildlife management invariably ignore this question. We know that a single fire regime will not suit all species of flora and fauna. The existing subdivision of the area into 26 blocks allows different fire regimes to be conducted and monitored facilitating an experimental approach to fire management. Without such an experimental and monitoring approach we struggle to obtain the information necessary for optimal management. We believe that the experimental approach is essential, especially with respect to the following premises that apply to habitat islands:

In this paper we use PVA to explore a specific management question and arrive at a strategy that maximises bandicoot population viability. With our baseline parameter set the optimal strategy is to burn at 15 year intervals and divide the scrub into at least three management units. This is represents a slight increase in the prescribed burn interval and suggests the possibility of significantly reducing the number of existing management units.

A sensitivity analysis of these results is essential. This needs to be carried out by repeating the simulations with parameters varied around their expected value. For some parameters the estimate of the probability of extinction may vary markedly. This often disturbs people, especially given uncertainty about many parameters and processes in the model. However the primary purpose of PVA is not to quantify extinction risk, but to choose between management options and explore impacts. In this context it is the ranking of management scenarios that is important, in this case, what is the best prescribed burn interval, not what is the extinction risk. The ranking of options is usually much less sensitive to changes in parameter values.

The very limited sensitivity analysis we conducted shows that there needs to be more research into the response of various life-history parameters to the successional stage of the vegetation, and the direct impact of fire, because these parameters have a significant impact on our decision about when to burn. However the ranking of options was found to be insensitive to several other parameters, even though parameters like juvenile mortality have a major impact on the absolute levels of extinction risk.

Finally we note that PVA process has a number of benefits aside from helping us to choose between management strategies. It helps to bring experts together and focuses them on the parameters and processes that influence population viability. The results from the model, and the dynamics of the population can be viewed and checked. By simulating the dynamics of a population over long times, wildlife biologists gain some appreciation of the fluctuating nature of the dynamics of populations. The educational spin-offs of building a model of population dynamics, and having to state explicitly the assumptions that underlie that model cannot be over-estimated. Placing research in wildlife biology in the context of understanding population dynamics often helps managers and biologists to focus on key issues, like estimating basic demographic rates.

12.6 Acknowledgments

The ideas and results presented in this paper would not be possible without the continuing support of, and collaboration with, Tony Norton, David Lindenmayer, Ian Davies, Tony Herbert and Ian Noble. Work on the southern brown bandicoot was partially supported by a National Estates Grant to Brian Gepp.

12.7 References

Akçakaya, H.R. 1992, 'Population viability analysis and risk assessment', in Proceedings of Wildlife 2001: Populations, Ed. D.R. McCullough, Elsevier, Amsterdam.

Armbruster, P. & Lande, R. 1993, 'A population viability analysis for African elephant (Loxodonta africana): how big should reserves be?', Conservation Biology, vol. 7, pp. 602-610.

Boyce, M. S. 1992, 'Population viability analysis', Annual Review of Ecology and Systematics, vol. 23, pp. 481-506.

Burgman, M.A., Akçakaya, H.R. & Loew, S.S. 1988, 'The use of extinction models for species conservation', Biological Conservation, vol. 43, pp. 9-25.

Burgman, M.A., Ferson, S. & Akçakaya, H.R. 1993, Risk Assessment in Conservation Biology, Chapman & Hall, London.

Haig, S.M., Belthoff, J.R. & Allen, D.H. 1993, 'Population viability analysis for a small population of red-cockaded woodpeckers and an evaluation of enhancement strategies', Conservation Biology, vol. 7, pp. 289-301.

Heinhson, G.E. 1966, 'Ecology and reproduction of the Tasmanian bandicoots (Perameles gunnii and Isoodon obesulus)', University of California Publications Zoology, vol. 80, pp. 1-107.

Lacy, R. C. 1993, 'VORTEX: a computer simulation model for population viability analysis', Wildlife Research, vol. 20, pp. 45-66.

Lindenmayer. D.B. & Possingham, H.P. 1995, 'The conservation of arboreal marsupials in the montane ash forests of the central highlands of Victoria, south-eastern Australia: VII. Modelling the contribution of modified logging strategies to the persistence of the endangered species Leadbeater's Possum', Biological Conservation, vol. 73, pp. 239-257.

Lindenmayer, D. B., Clark, T. W., Lacy, R. C. & Thomas, V. C. 1993, 'Population viability analysis as a tool in wildlife management: a review with reference to Australia', Environmental Management, vol. 17, pp. 745-758.

Paull, D.J. 1992, 'The distribution, ecology and conservation of the southern brown bandicoot (Isoodon obesulus obesulus) in South Australia. MA thesis, Geography Department, University of Adelaide.

Possingham, H. P. 1991, 'The role of population viability analysis in forest management', in Conservation of Australia's Forest Fauna, ed. D. Lunney, Surrey Beatty & Sons, Sydney. pp. 35-39.

Possingham, H.P. & Gepp, B. 1993, 'The application of population viability analysis to assess management options for the southern brown bandicoot (Isoodon obesulus) in the SE of South Australia', Proceedings of the International Simulation Society Conference, Perth. pp. 633-638.

Possingham, H.P., Davies, I., Noble, I.R. & Norton, T. W. 1992, 'A metapopulation simulation model for assessing the likelihood of plant and animal extinctions', Mathematics and Computers in Simulation, vol. 33, pp. 367-372.

Possingham, H.P., Lindenmayer, D.B. & Norton, T.W. 1993, 'A framework for improved threatened species management', Pacific Conservation Biology, vol. 1, pp. 39-45.

Possingham, H.P. & Davies, I. 1995, 'ALEX: metapopulation simulation model for assessing the likelihood of plant and animal extinctions', Biological Conservation vol. 73, pp. 143-150.

Shaffer, M. L. 1981, 'Minimum population size for species conservation', Bioscience, vol. 31, pp. 131-134.

Shaffer, M. L. 1990, 'Population viability analysis', Conservation Biology, vol. 4, pp. 39-40