SEARCH

SEARCH BY CITATION

Keywords:

  • dispersal;
  • habitat quality;
  • population management;
  • resource quality;
  • simulation model;
  • wild rabbit

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Patterns of connectivity influence pest population system dynamics, and it is essential to consider connectivity when planning effective management strategies. Traditional connectivity models often consider populations embedded in a matrix of unsuitable habitat. This approach is unlikely to be applicable to those pest species that can utilize most of the landscape in which they live. There is therefore a need for a simple and flexible tool to assess connectivity in such systems.
  • 2
    In this study, we developed a new model in which contiguous resource patches that differ in quality, and landscape elements that impede dispersal, impact on connectivity within a population system. The model was applied to a wild rabbit population system, a well-studied pest species in Australia. An independent population genetic data set was used to validate the model.
  • 3
    There was a highly significant association between pairwise population connectivity and the genetic data (Mantel test, r=−0·502, P= 0·002). As predicted, two populations that showed very low connectivity were strongly isolated genetically. These sites appeared to be substantially isolated because of forests, which acted to impede rabbit dispersal. When these sites were excluded from analysis, connectivity indices again explained the pattern of genetic data (Mantel test, r=−0·46, P= 0·037). This showed that both spatial variation in resource quality and forests influenced connectivity in this system. Sensitivity analyses confirmed that the distribution and extent of forests was important in limiting connectivity to some sites. The model was relatively robust to changes in population parameters.
  • 4
    Synthesis and applications. Connectivity among wild rabbit populations in this system was strongly influenced by habitat heterogeneity, rather than factors such as geographical distance or major landscape elements such as rivers, both of which are traditionally considered to influence system dynamics. This may have substantial implications for many pest systems, and suggests that the impact of habitat heterogeneity on connectivity should be considered when planning efficient management strategies.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The internal dynamic processes of local populations are affected by the quality of resources in the area that they occupy (Caughley et al. 1988; Pulliam 1988). When local populations function within a population system, however, their characteristics will also be influenced by interactions with other populations (Andrewartha & Birch 1954; den Boer 1968; Pulliam 1988). As in any population system, connectivity is of fundamental importance to the functioning of pest systems, but the need to consider connectivity has at times been overlooked in management programmes. An isolated pest population that experiences an increase in mortality over natality (e.g. because of local control procedures) may be driven to extinction because it is unlikely to be buffered by a rescue effect (Brown & Kodric-Brown 1977). Conversely, highly connected pest populations are likely to receive substantial numbers of immigrants. Extinction under a local control regime is improbable, and even if a local population is driven to extinction the patch is liable to be quickly recolonized. However, highly connected populations may be vulnerable to broad-scale control measures such as the targeted introduction and transmission of diseases designed to increase mortality, or act as vectors for immunocontraceptive agents. The dynamics of pest population systems are thus critically influenced by long-term patterns of dispersal. Understanding the factors that influence connectivity is vital for the pest's effective management, and there is a need for simple and flexible tools to allow managers to assess connectivity.

Several studies have examined the effects of habitat heterogeneity on connectivity in population systems. They have included examinations of the effects of variation in patch quality (Fleishman et al. 2002; Bonte et al. 2003), matrix composition (Ricketts 2001) and corridor effectiveness (Aars & Ims 2000; Mech & Hallett 2001). These studies have been predominantly set within a metapopulation paradigm, in which small habitat patches are encompassed by a broad region of unsuitable habitat. This approach has proved highly effective in many studies in which the matrix is a far larger component of the landscape than habitat and habitat patches tend to be relatively isolated. However some species, notably pest species, may have the capacity to utilize much of the landscape in which they live. In these systems, interactions among spatially structured local populations occur within a landscape in which resources are spread relatively continuously across large regions, albeit with variation in resource quality. Note that we distinguish here between the relatively continuous distribution of resources on the landscape, and the natural groupings of organisms that form populations that utilize the resources. While much of the landscape could be utilized, a real landscape is also likely to contain structural features that impede free movement of individuals.

In semi-arid regions of Australia the wild rabbit Oryctolagus cuniculus L. provides an example of a species living in such a heterogeneous landscape. Whilst most of the landscape here can support rabbit populations, they cannot establish warrens in forest, and forests also impede dispersal (Stodart & Parer 1988; Myers et al. 1994). Thus while a landscape consists of both suitable and unsuitable patches, many of the suitable patches will be contiguous. Analysing connectivity in such a system using traditional models is therefore difficult.

The aim of this present study was to develop a simple habitat heterogeneity model in order to investigate the hypothesis that connectivity within a broadly spread pest population system can be influenced by the spatial configuration of contiguous suitable patches interspersed with unsuitable patches. The utility of the model was assessed by applying it to a case study of a wild rabbit population system in Australia. The model was validated with population genetic data, which is appropriate because the level of population genetic structuring within a system will be influenced largely by the level of interaction among its constituent populations through time (Wright 1951; Slatkin 1987). The insights gained from this study should not only enable more efficient management of wild rabbits in Australia, but of many other pests that rely on resource patches that are broadly spread across a region.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

model development

Consider a structured population consisting of P local populations inhabiting a heterogeneous landscape of Q patches. The heterogeneous landscape consists of contiguous suitable patches (resource patches) interspersed with unsuitable patches. Patches are considered suitable if they contain resources that can support a local population. Suitable patches consist of a suite of m resources, and the quality of each resource can vary among patches. Resources promote population growth, so the quality of a resource is assessed with respect to its effect on population growth. An index of quality for a resource in any suitable patch (Rp) can be defined as the quality of that resource in the patch (rp) relative to the optimal quality of the resource among all suitable patches (ropt). In a system with a single resource:

  • image(eqn 1 )

The resource index in a patch potentially acts as a reducing fraction on growth in a patch if resources are suboptimal (R1 < 1). The shapes of resource patches reflect the spatial distribution of resources within a region and so tend to be irregular in a real landscape. The difference between the smallest and largest patches can thus vary by orders of magnitude. Unsuitable patches are landscape elements that do not contain resources and thus cannot support a population. However, they do reduce dispersal capacity and this feature is considered below.

Consider now the network of P local populations within the landscape. Local population growth is influenced by the quality of resources in a suitable patch. Each population that is established exhibits dynamics described by a simple function of growth and resource quality. We make the assumption that maximum growth (G) is a constant for the system.

Each of the P local sampling populations has the potential to interact with each of the other populations in the system. We consider interactions across a broad scale, and each of the P sampling populations is separated from all others by a distance beyond the range of a single dispersal event. Interaction between each pair of populations therefore occurs in a stepping stone fashion via a series of patch colonization, population growth and dispersal events through contiguous patches that intervene between the two populations. We define here the terms non-sample patch (NSP), a patch which forms part of the series of patches between sampling populations, and non-sample population, a population that occupies a suitable NSP. The boundaries of each NSP in the following case study are partly delineated by dispersal paths and will be described in more detail below. Each suitable NSP can support a single non-sample population.

We assume that dispersal from a population is a linear function of population size. If we consider the interaction between two sampling populations in two adjacent patches, the number of immigrants entering the second patch can be calculated as:

  • image(eqn 2 )

where N1 is a seeding population introduced into the first patch and d is the dispersal constant. However, the two populations of interest may be separated by p suitable NSP. In this case, each suitable patch that separated the two populations would in turn be colonized, the resultant population would experience growth (modified by resource quality), and a dispersing propagule would be sent into an adjoining patch. These steps would continue until the second population accepted immigrants. Here, the size of the propagule immigrating into the destination patch is:

  • image(eqn 3 )

Unsuitable NSP may also lie on a dispersal path between two sample populations. Populations cannot establish on unsuitable patches, and they also impede dispersal. The probability of successful dispersal through these patches relates to the geographical distance of the dispersal path through the unsuitable patch. An appropriate dispersal resistance function is applied based on the capacity of the model organism to disperse through that landscape element. In effect this imposes some level of mortality on propagules dispersing through the patch. If empirical data for such dispersal probabilities do not exist, a suitable mathematical function (e.g. a logarithmic reduction) based on the unimpeded dispersal distance of a species may be used. For example, experimental evidence shows that the mean unimpeded dispersal distance of the wild rabbit is 1·5 km (Parer 1982). In the following case study, a function was used in which 100% of a dispersing propagule could pass through small tracts of 1·5 km of an unsuitable patch, 10% of the propagule could pass through intermediate tracts of 1·5–3 km of an unsuitable patch, and 1% could pass through large tracts of 3 km or more.

The number of immigrants that finally enter a destination population in any interaction will thus depend on heterogeneity at the resource level (variations in resource quality influencing the size of dispersing propagules) and at the structural level (where unsuitable patches impede dispersal). A simulation model was constructed to evaluate all pairwise interactions among the network of P local populations. For each population pair, one population was initially considered a donor and the other a destination population, after which the roles were reversed. Before each pairwise interaction, all patches in the landscape, including sample population sites, were considered vacant and a seeding population (N1) was introduced into the donor population. The results of each pairwise interaction (total number of successful colonists arriving at a destination population and the donor populations from which they originated) were recorded.

model validation

In this model, dispersal between two local populations would usually be asymmetric. This means that in a pairwise interaction between populations 1 and 2, I1  I2, where I is the number of immigrants entering a population. We considered that connectivity between a pair of populations was the average of the number of individuals immigrating into each of the populations, allowing half matrices of pairwise connectivity to be constructed. In the following case study eight sites in the landscape were selected as sampling populations to coincide with populations for which validation data were available. Two sets of simulations were run. The first, the full model, included all local populations. The second simulation excluded those populations that appeared to be substantially isolated by forest (Polworth and Verniew; Fig. 1). The purpose of the second simulation was to ensure that connectivity within this system was not entirely driven by barrier effects. For each simulation a half matrix of connectivity was constructed.

image

Figure 1. Habitat heterogeneity map for the study area with geographical distribution of wild rabbit populations (stars) and location of the Mitchell township (circle). For clarity, some small soil patches have been excluded. The method used for constructing dispersal paths between two sites is demonstrated in the legend (not to scale).

Download figure to PowerPoint

The mitochondrial haplotype frequency data of Wilson, Fuller & Mather (2002) and Fuller, Mather & Wilson (1996, 1997) were then used to provide an independent assessment of connectivity for each simulation. Note that the haplotype frequencies at one site (Glenlea) were determined by summing four populations occurring on the same large resource patch. There was no significant difference in haplotype frequencies between the four populations (χ2 = 10·4, P= 0·11). We tested the hypothesis that there was no association between pairwise connectivity and a measure of genetic distance (FST values) between pairs of populations in the system using a Mantel test. The significance of the Mantel test result was determined by 9999 random permutations of the matrices. As we would predict that pairwise FST values should decrease as connectivity increases, we would expect a negative association between the connectivity and genetic distance matrices.

sensitivity analysis

For the purposes of sensitivity analysis, we investigated an alternative connectivity measure to summarize interactions. The relative connectivity C of local population α results from the mean of all two-way interactions between α and the n – 1 other populations:

  • image(eqn 4 )

where Nαi is the number of individuals originating from the ith population immigrating into population α; N*αi is the number of individuals from population α immigrating into the ith population; and inline image is the total number of individuals from all other populations immigrating into the ith population. Regression of C against the mean of pairwise FST values provided values that were used for qualitative comparisons in the following sensitivity analyses.

The assumptions governing dispersal resistance were assessed. The small forest tract index was kept constant, and the intermediate forest tract index and the large forest tract index were varied over approximately 1000 combinations. Additionally, approximately 1000 simulations were run using a range of growth and dispersal values that were consistent with those found for the wild rabbit in Australia (Daly 1979; Parer 1982; Hone 1999). Variation in initial population size (from N1 = 10 to N1 = 1000) was also assessed.

model application

The model was applied to a wild rabbit population system in the semi-arid Mitchell region of southern central Queensland, Australia. High-quality population genetic data were available for eight wild rabbit populations in the region (Wilson, Fuller & Mather 2002) to enable validation of the model.

Because of its pest status in Australia, the wild rabbit has been intensively investigated. The resources upon which wild rabbit populations depend are well documented, and many studies highlight soil quality as the major resource that affects rabbit population growth, with rabbit populations in Australia (Myers & Parker 1965; Myers 1970; Myers & Parker 1975; Parker et al. 1976; Parer & Libke 1985), England (Trout & Smith 1995), France (Rogers, Arthur & Soriguer 1994) and Spain (Soriguer & Rogers 1981) attaining different densities in patches of different soil types.

We classified optimal, intermediate and poor quality soils in the study region according to the density of rabbits those soils could support using the study of Myers (1970). Densities were indexed relative to optimal soil, resulting in indices of 1, 0·61, 0·31 for optimal, intermediate and poor soils, respectively (equation 1). Soil patches were defined using a soil map (Gallowway, Gunn & Fitzpatrick 1974) and the appropriate soil index was applied to each patch.

The only known unsuitable habitat type for wild rabbit in this region is forest. Forests in the Mitchell region were defined using aerial photographs and forest patches were superimposed on the soil patch network. The habitat heterogeneity map therefore consisted of a network of contiguous resource patches of optimal, intermediate and poor soils interspersed with forest patches (Fig. 1).

Population parameters for the model were based on empirical data. The growth factor was calculated from 19 years of population census data for rabbit populations in arid Australia (B. Cooke, unpublished data; a subset of the data is available in Cooke 1974). The growth factor was calculated by averaging population increases from trough to peak densities. The resulting growth factor (G) of 7·88 agreed well with the maximum observed annual finite rate of increase of 7·85 calculated for rabbits in Australia by Hone (1999). As validation was conducted with mitochondrial DNA (mtDNA) data (see below), a female dispersal constant (d) was used. This dispersal constant of 0·29 was calculated from the studies of Daly (1979) and Parer (1982).

While forests impede rabbit dispersal, the nature of this resistance has not been quantified. A conservative dispersal resistance function based on the unimpeded dispersal distance of the wild rabbit (described above) was implemented.

Eight population sites were selected on the habitat heterogeneity map to coincide with the populations sampled by Wilson, Fuller & Mather (2002) (Fig. 1). For the purposes of drawing dispersal paths, each site was considered as a circle corresponding to a 6-km radius centred on the site sampled. Five linear dispersal paths were constructed between each pair of sites by connecting points on the circles surrounding them (Fig. 1). For the interaction between any two sampling populations, we considered only those soil and forest patches defined within the outermost pair of linear dispersal paths (Fig. 1) and disregarded the remainder of the landscape. Therefore it was possible for a single large soil patch to be colonized more than once during a single pairwise interaction if intersected by a dispersal path on more than one occasion. There were 280 dispersal paths among all sites (i.e. 140 paths considered in both directions), effectively sampling landscape heterogeneity in the region.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

model validation

There was a highly significant association between pairwise connectivity and genetic distance in the full model of eight local populations (Mantel test, r=−0·502, P= 0·002; Table 1), showing that FST (and thus genetic distance) decreased as connectivity increased. On examination it was apparent that connectivity varied markedly among sites (Table 1, lower half matrix). Bowann, Claravale, Currawong, Glenalba and Thornlee exhibited high levels of connectivity within the system. These sites formed a group within which interaction was high, with each population accepting large numbers of immigrants from other populations within the group. Although still high, Glenlea's connectivity was somewhat lower than this group. In contrast, Polworth and Verniew showed very low connectivity. These two sites contributed and accepted few migrants from other sites within the system, and there was no direct interaction between them. These latter two populations were thus effectively isolated from each other and the rest of the system. Based on these observations, two predictions were made regarding the genetic constitution of the system: (i) the highly connected group would show genetic similarities among sites and (ii) the isolated sites would show divergence from the rest of the system and from each other.

Table 1.  Mean pairwise connectivity (lower half matrix) and pairwise FST values (upper half matrix) for eight wild rabbit populations (*P < 0·05). Sample sizes for each of the eight populations are provided in parentheses (Wilson, Fuller & Mather 2002)
 Bowann (33)Claravale (31)Currawong (33)Glenalba (30)Glenlea (142)Polworth (30)Thornlee (46)Verniew (35)
Bowann    −0·016    0·019   0·015 0·02  0·178*−0·0270·111*
Claravale18545·5     0·084   0·078 0·007  0·263*−0·0050·037
Currawong 1235·52013·5   −0·034 0·08*  0·065   0·0120·266*
Glenalba 5381·58039·513421  0·077*  0·069   0·0090·259*
Glenlea 3714 618 43322540   0·167*   0·0280·069*
Polworth   86 147·5  233·5 51594·5    0·149*0·438*
Thornlee 71605081·512241·54389·52566665·5 0·134*
Verniew  119·5 896·5 10162427   0  0496 

As expected, there was no significant divergence among the highly connected populations (Table 1, upper half matrix). Consistent with somewhat lower connectivity, Glenlea showed significant genetic differences compared with four local populations (Currawong, Glenalba, Polworth and Verniew). Most notably, the pair of populations with very low connectivity showed significant differentiation from the majority of other sites within the system. Polworth showed significant differences compared with all populations except Currawong and Glenalba, and Verniew showed significant differences to all populations except Claravale. This provided evidence of the substantial genetic isolation of Polworth and Verniew.

Unlike other local populations, these two isolated sites were closely associated with large tracts of forest (Fig. 1). There was thus a possibility that isolation because of forest had a disproportionate influence on system dynamics. However, when these two isolated local populations were excluded from the analysis, there was a weaker but still significant relationship between connectivity and genetic distance for the remaining six populations that occupied more continuous habitat (r = −0·46, P= 0·037). This suggested that both spatial variation in resource quality and the spatial distribution of forests influenced connectivity among local wild rabbit populations in the region.

sensitivity analysis

Modifying the dispersal resistance function had substantial impacts on the fit of the model to validation data, confirming the importance of forest in preventing rabbit dispersal (Fig. 2). Model output was affected by dispersal resistance indices applied to both intermediate and large forest tracts, but was most strongly affected by the proportion of dispersers that successfully traversed large forest tracts.

image

Figure 2. Sensitivity of the habitat heterogeneity model output to changes in the dispersal resistance function. The contours represent regression coefficients, and the axes show changes in the indices for intermediate (1·5–3 km) and large (> 3 km) forest tracts. Note the contours fill only half the graph space because the large forest tract index was always equal to or less than the intermediate forest tract index. Regression isoclines with regression values of 0·5 or greater have been labelled.

Download figure to PowerPoint

The model was relatively robust to changes in population parameters, with growth and dispersal parameters having a roughly equivalent impact on model output (Fig. 3). This was not unexpected given the form of equation 3. Model output was also robust to changes in initial population size, with the coefficient of variation varying between 0·75 and 0·83.

image

Figure 3. Sensitivity of habitat heterogeneity model output to changes in growth and dispersal parameters. Refer to Fig. 2 for explanation of the graph and notation.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The results presented here show that connectivity among local populations in a spatially structured pest system can be influenced by habitat heterogeneity. Genetic distance among wild rabbit populations in the Mitchell region increased as inferred levels of interaction based on pairwise connectivity decreased, and, as predicted, populations that exhibited very low connectivity levels were also genetically isolated within the system. This study provides evidence that interactions among local wild rabbit populations in this region were strongly influenced by the spatial distribution of soil quality and forests.

It is important to note that while the distribution of forests strongly isolated two local populations in the system (Polworth and Verniew), the habitat heterogeneity model also explained genetic structuring amongst sites in more continuous habitat when the isolated sites were omitted from analysis. Therefore, the spatial distribution of both suitable and unsuitable patches explains connectivity in this system. For example, Glenlea showed a significant genetic difference to Glenalba and forest was only sparsely distributed between these two sites. However, there was a sequence of intermediate and poor soil patches between these sites that limited the connectivity between the local populations occupying them.

Although geographical distance is often used as an indicator of connectivity within population systems, it does not appear to be a major factor affecting connectivity among the populations studied here. Some populations that are widely separated (e.g. Claravale and Glenalba; Fig. 1) show much higher interaction (Table 1) than adjacent populations (e.g. Claravale and Verniew). This is consistent with the analysis of Wilson, Fuller & Mather (2002), who found that genetic structuring within this system could not be explained by a distance model. Interestingly, Wilson, Fuller & Mather (2002) also found that the Maranoa river, which bisects the study site, did not influence structuring, because rabbits are capable of swimming and the river flow rates are insufficient to restrict rabbit movement. The finding that connectivity in this system is mainly driven by habitat heterogeneity has substantial implications for the management of pest population systems.

Connectivity among local pest populations will influence probabilities of extinction, recolonization and rescue effects (Brown & Kodric-Brown 1977). To be useful as a management tool, a model should identify patterns of connectivity within a system and the critical factors that influence those patterns. In the Mitchell region it is important to know that a sequence of high-quality soil patches between wild rabbit populations will allow high connectivity, even when the populations are separated by tens of kilometres. For instance, attempting to decrease the density of rabbits in one of the highly connected populations (e.g. Bowann) by a local control measure such as baiting may be a poor allocation of resources because, even if an extinction occurred, the site would be rapidly recolonized from within the highly connected group. Conversely, release of a disease such as the Rabbit Calicivirus Disease (RCD) at a highly connected site may be efficient because it is likely to lead to transmission to the other highly connected sites. Transmission to isolated sites (Polworth and Verniew) is much less probable, where local control measures may be suitable. It is also noteworthy that retaining forests may be an important component of a regional rabbit control strategy, as large tracts of forest strongly limit connectivity. These recommendations must be tempered by the recognition that changes in population structure may occur on a temporal scale longer than that of management strategies, and that the efficacy of lethal management techniques will be strongly influenced by the absolute frequency of recolonization events. The frequency of recolonization events should therefore also be considered within an effective management strategy.

While it is possible to construct a model with many parameters in order to emulate closely one species in a particular region, there are significant advantages to constructing a model that captures the major biological processes in a population system without excessive complexity. A simple model maintains generality, allowing it to applied to a variety of systems with minor modifications. As there are few parameters, these parameters can be estimated from empirical data. Importantly, it may be possible to confirm the assumptions on which the model is based by validation with appropriate independent data, which is often difficult for complex models (Kareiva 1990). Indeed, the appropriate validation of an ecological model with a large, high-quality and independent data set, as has been done here, is rare.

The habitat heterogeneity model developed above is not, however, presented as a comprehensive population dynamic model. The model is both simple and deterministic, which allowed the few parameters to be drawn from empirical studies and the model to be clearly validated with independent data. While a simple model was appropriate to investigate the link between habitat heterogeneity and connectivity in this study, further refinements to the model could be considered. Stochasticity is an important component of natural processes and incorporating natural variation into the model, at the population and environmental levels, would undoubtedly provide further insights into connectivity patterns in pest population systems. It may also be fruitful to vary the form of certain parameters in a biologically meaningful way, for instance by varying dispersal according to population density or introducing density dependence. Implementing random walks among populations, rather than linear dispersal paths as used in the current study, is also likely to be useful. One particularly interesting extension would be to consider the effect of any biases in dispersal between males and females. Sex-biased dispersal is a characteristic of many vertebrates and one that is likely to influence connectivity patterns and thus management plans. Simulations are currently being undertaken to determine any effects on connectivity of sex-biased dispersal in rabbits.

Alternative methods such as least cost dispersal (LCD) modelling have been used to assess connectivity in metapopulations (Vos et al. 2001; Schweiger, Frenzel & Durka 2004) and are generally implemented using a geographic information system (GIS). While useful in systems in which a matrix comprises the majority of the landscape, LCD is unlikely to provide any advantage over the current method in systems where it is not necessary to consider explicitly movement cost on most patches. Such models are usually difficult to parameterize (although see Kramer-Schadt et al. 2004 for parameterization of a spatially explicit model using telemetric data). Even for patches that impede dispersal in the rabbit case study (forests), implementation of a simple resistance function proved adequate in the absence of empirical data, with the additional advantages that it was clearly a model assumption, could easily be replaced by an alternative function if appropriate, and was amenable to sensitivity analysis.

We have shown here that habitat heterogeneity can influence connectivity in a pest population system. The modelling approach used in this study could be easily adapted for use in other pest systems in specific regions by identifying and mapping important resources and landscape elements that impede dispersal. It is also important to note that this approach could equally be used in a conservation setting. Resources that affect population growth will differ among target plant and animal species, whether in conservation or control settings, and identification of these critical resources will be a key factor in correct utilization of the model. For example, the European hare Lepus europaeus has been classed as a ‘priority species of conservation concern’ by the UK government, and pastures are suboptimal habitat for this species (Smith et al. 2004). Vegetation type or structure may therefore be one measure of habitat quality here. The suggestions for population control outlined in this study could easily be adapted for this species, for example to increase connectivity among local populations if appropriate.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The efforts of Adam Liedloff in assisting with the development of the simulation program are appreciated. The manuscript was improved by comments from Greg Hood, Severine Vuilleumier and two anonymous referees.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Aars, J. & Ims, R.A. (2000) Population dynamic and genetic consequences of spatial density-dependent dispersal in patchy populations. American Naturalist, 155, 252265.
  • Andrewartha, H.G. & Birch, L.C. (1954) The Distribution and Abundance of Animals. University of Chicago Press, Chicago, IL.
  • Den Boer, P.J. (1968) Spreading of risk and stabilisation of animal numbers. Acta Biotheoretica, 18, 165194.
  • Bonte, D., Lens, L., Maelfait, J.P., Hoffmann, M. & Kuijken, E. (2003) Patch quality and connectivity influence spatial dynamics in a dune wolfspider. Oecologia, 135, 227233.
  • Brown, J.H. & Kodric-Brown, A. (1977) Turnover rates in insular biogeography: effect of immigration on extinction. Ecology, 58, 445449.
  • Caughley, G., Grice, D., Barker, R. & Brown, B. (1988) The edge of the range. Journal of Animal Ecology, 57, 771785.
  • Cooke, B.D. (1974) Food and other resources of the wild rabbit Oryctolagus cuniculus (L.). PhD Thesis. University of Adelaide, Adelaide, Australia.
  • Daly, J. (1979) The ecological genetics of the European wild rabbit Oryctolagus cuniculus (L.) in Australia. PhD Thesis. Australian National University, Canberra, Australia.
  • Fleishman, E., Ray, C., Sjogren-Gulve, P., Boggs, C.L. & Murphy, D.D. (2002) Assessing the roles of patch quality, area, and isolation in predicting metapopulation dynamics. Conservation Biology, 16, 706716.
  • Fuller, S.J., Mather, P.B. & Wilson, J.C. (1996) Limited genetic differentiation among wild Oryctolagus cuniculus L. (rabbit) populations in arid eastern Australia. Heredity, 77, 138145.
  • Fuller, S.J., Mather, P.B. & Wilson, J.C. (1997) Patterns of differentiation among wild rabbit populations Oryctolagus cuniculus L. in arid and semi-arid ecosystems in north-eastern Australia. Molecular Ecology, 6, 145153.
  • Gallowway, R.W., Gunn, R.H. & Fitzpatrick, E.A. (1974) Lands of the Balonne-Maranoa Area. Land Research Series No. 34. CSIRO, Melbourne, Australia.
  • Hone, J. (1999) On rate of increase (r): patterns of variation in Australian mammals and the implications for wildlife management. Journal of Applied Ecology, 36, 709718.
  • Kareiva, P. (1990) Population dynamics in spatially complex environments: theory and data. Philosophical Transactions of the Royal Society of London, Series B, 330, 175190.
  • Kramer-Schadt, S., Revilla, E., Wiegand, T. & Breitenmoser, U. (2004) Fragmented landscapes, road mortality and patch connectivity: modelling influences on the dispersal of Eurasian lynx. Journal of Applied Ecology, 41, 711723.
  • Mech, S.G. & Hallett, J.G. (2001) Evaluating the effectiveness of corridors: a genetic approach. Conservation Biology, 15, 467474.
  • Myers, K. (1970) The rabbit in Australia. Proceedings of the Advanced Study Institute on Dynamics of Numbers in Populations (eds P.J.Den Boer & G.R.Gradwell), pp. 478506 Centre for Agricultural Publishing and Documentation, Wageningen, the Netherlands.
  • Myers, K. & Parker, B.S. (1965) A study of the biology of the wild rabbit in climatically different regions of eastern Australia. I. Patterns of distribution. CSIRO Wildlife Research, 10, 132.
  • Myers, K. & Parker, B.S. (1975) A study of the biology of the wild rabbit in climatically different regions of eastern Australia. VI. Changes in numbers and distribution related to climate and land systems in semi-arid north-western New South Wales. Australian Wildlife Research, 2, 1132.
  • Myers, K., Parer, I., Wood, D. & Cooke, B.D. (1994) The rabbit in Australia. The European Rabbit: The History and Biology of a Successful Coloniser (eds H.V.Thompson & C.M.King), pp. 108157. Oxford University Press, New York, NY.
  • Parer, I. (1982) Dispersal of the wild rabbit Oryctolagus cuniculus, at Urana in New South Wales. Australian Wildlife Research, 9, 427441.
  • Parer, I. & Libke, J.A. (1985) Distribution of rabbit (Oryctolagus cuniculus) warrens in relation to soil type. Australian Wildlife Research, 12, 387405.
  • Parker, B.S., Hall, L.S., Myers, K. & Fullagar, P.J. (1976) The distribution of rabbit warrens at Mitchell, Queensland, in relation to soil and vegetation characteristics. Australian Wildlife Research, 3, 129148.
  • Pulliam, H.R. (1988) Sources, sinks, and population regulation. American Naturalist, 132, 652661.
  • Ricketts, T.H. (2001) The matrix matters: effective isolation in fragmented landscapes. American Naturalist, 158, 8799.
  • Rogers, P.M., Arthur, C.P. & Soriguer, R.C. (1994) The rabbit in continental Europe. The European Rabbit: The History and Biology of a Successful Coloniser (eds H.V.Thompson & C.M.King), pp. 2263. Oxford University Press, New York, NY.
  • Schweiger, O., Frenzel, M. & Durka, W. (2004) Spatial genetic structure in a metapopulation of the land snail Cepea nemoralis (Gastropoda: Helicidae). Molecular Ecology, 13, 36453655.
  • Slatkin, M. (1987) Gene flow and the geographic structure of natural populations. Science, 236, 787792.
  • Smith, R.K., Jenning, R.V., Robinson, A. & Harris, S. (2004) Conservation of European hares Lepus europaeus in Britain: is increasing habitat heterogeneity in farmland the answer? Journal of Applied Ecology, 41, 10921102.
  • Soriguer, R.C. & Rogers, P.M. (1981) The European wild rabbit in Mediterranean Spain. Proceedings of the World Lagomorph Conference (1979) (eds K.Myers & C.D.McInnes), pp. 600613. University of Guelph, Ontario, Canada.
  • Stodart, E. & Parer, I. (1988) Colonisation of Australia by the Rabbit. Division of Wildlife and Ecology Project Report No. 6. CSIRO, Canberra, Australia.
  • Trout, R.C. & Smith, G.C. (1995) The reproductive productivity of the wild rabbit (Oryctolagus cuniculus) in southern England on sites with different soils. Journal of Zoology, 237, 411422.
  • Vos, C.C., Antonisse-de Jong, A.G., Goedhart, P.W. & Smulders, M.J.M. (2001) Genetic similarity as a measure of connectivity between fragmented populations of the moor frog (Rana arvalis). Heredity, 86, 598608.
  • Wilson, J.C., Fuller, S.J. & Mather, P.B. (2002) The formation and maintenance of discrete wild rabbit (Oryctolagus cuniculus L.) population systems in arid Australia: habitat heterogeneity and management implications. Austral Ecology, 27, 183191.
  • Wright, S. (1951) The genetical structure of populations. Annual Eugenics, 15, 323354.