1. Fire is a major driver of ecosystem structure and function worldwide. It is also widely used as a management tool to achieve conservation goals. A common objective is the maintenance of ‘fire mosaics’ comprising spatially heterogeneous patches of differing fire history. However, it is unclear what properties of fire mosaics most enhance conservation efforts. Here we focus on the spatial and temporal properties of fire-prone landscapes that influence the distribution of small mammals.
2. We surveyed small mammals in 28 landscapes (each 12·6 km²) representing a range of fire histories in the Murray Mallee region (104 000 km²) of semi-arid Australia. Generalised linear mixed models were used to examine the influence of five landscape properties on the capture rate of individual species and the species richness of native small mammals. We investigated the influence of the proportional extent of fire age-classes, the diversity of fire age-classes, the extent of the dominant vegetation type, rainfall history and biogeographic context.
3. Three of four study species were associated with the spatial extent of fire age-classes. Older vegetation was found to provide important habitat for native small mammals. Overall, however, rainfall history and biogeographic context were dominant influences: for example, the species richness of native mammals was positively associated with above-average rainfall. There was little evidence that the diversity of fire age-classes influenced either the capture rate of individual species or species richness.
4.Synthesis and applications. In fire-prone environments, habitat availability can change markedly over short time-scales. Sufficient habitat at a suitable seral stage within the landscape is a key requirement for species conservation. In mallee ecosystems, the retention of older vegetation is recommended to create more desirable fire mosaics for native small mammals. In addition to such spatial properties of mosaics that are amenable to manipulation, an understanding of how ecological processes affect the biota (such as variation in rainfall-driven productivity) is also essential for informed conservation management.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Several mechanisms have been proposed, which directly link the distribution and abundance of fauna with the properties of fire mosaics (Bradstock et al. 2005; Parr & Andersen 2006). First, animal species may depend on resources that vary temporally in response to time-since-fire (or fire history) (e.g. Fox 1982). Under this hypothesis, changes in the spatial extent of different fire age-classes will modify the total amount of these resources available in the landscape and thereby influence the distribution and abundance of species. Secondly, because different taxa exhibit different responses to fire history, it is argued that heterogeneous mosaics composed of patches of differing fire history will provide a range of resources that will enable the persistence of a species-rich biota in the landscape (e.g. Briani et al. 2004). This hypothesis has been encapsulated by the expression ‘pyrodiversity begets biodiversity’ (Parr & Andersen 2006). Thirdly, individual species may require resources present in multiple components of the landscape: for example, patches of recently burnt vegetation adjacent to older vegetation (e.g. Firth et al. 2010). The composition of a fire mosaic may determine whether or not such species can occur. Finally, during a fire event, unburnt patches in the landscape can provide refuges for animals. Following a fire, these patches can provide a source for recolonisation into disturbed areas (Bradstock et al. 2005).
Studies of small mammals have been influential in the development of fire ecology. Such studies have, for example, provided insights into post-fire succession in animal communities (Fox, Taylor & Thompson 2003), the influence of fire frequency on fauna (Pardon et al. 2003) and the interactions between fire, rainfall and other ecological processes (Letnic & Dickman 2005). However, the relative influence of different spatial properties of fire-prone landscapes on small mammals, and other faunal groups, is poorly known. This is because data are primarily collected at a site level and do not allow direct comparison and understanding of how the combinations of spatial features that make up the whole landscape – the extent, composition and configuration of elements – influence fauna. To date, a whole-of-landscape approach (where the unit of study is the landscape) has largely been restricted to studies of fragmented agricultural systems (Bennett, Radford & Haslem 2006; Mortelliti et al. 2011).
We conducted a broad-scale natural experiment in the Murray Mallee region of semi-arid southern Australia. ‘Mallee’ shrubland and woodland offers a model system for investigating fire mosaics. This eucalypt- and spinifex-dominated vegetation is highly flammable, landscape structure is strongly associated with fire history and large tracts of native vegetation occur in conservation reserves. We surveyed small mammals in 28 landscapes (each 12·56 km²) selected to represent a range of fire histories and therefore different fire mosaics. In particular, we examined five landscape properties: the proportional extent of fire age-classes, the diversity of fire age-classes, the extent of the dominant vegetation type, rainfall history and the biogeographic context of the landscape. We sought to establish the relative influence of each of these properties on the distribution of small mammals in the region.
Materials and methods
The Murray Mallee region (104 000 km2) encompasses an extensive system of reserves managed primarily for biodiversity conservation (Fig. 1). The landscape is of low elevation (≤100 m above sea level), with moderate topographic variation provided by undulating dune and swale systems. Large tracts of native vegetation are characterised by stands of ‘tree mallee’ (i.e. vegetation dominated by Eucalyptus spp. with a multistemmed growth form). Three types of mallee vegetation have been identified and mapped across the study region: Triodia Mallee, Heathy Mallee and Chenopod Mallee (Haslem et al. 2010).
Mallee vegetation is fire-prone: an area equivalent to 40% of tree mallee vegetation in the region was burnt between 1972 and 2007 (S.C. Avitabile, unpublished data). Lightning is the major source of ignition. Mallee eucalypts and the perennial hummock grass, Triodia scariosa N.T. Burbidge, are the primary sources of surface fuel (Bradstock & Cohn 2002). Wildfires exceeding 100 000 ha typically occur in the region every 10–20 years, although longer fire intervals are common for any given location (Clarke et al. 2010). Smaller fire events occur more frequently.
The climate of the region is semi-arid; mean annual rainfall ranges from 218 mm in the north to 329 mm in the south. Rainfall typically is non-seasonal, and the interannual variability is high. Mean daily maximum temperatures in summer range from 30–33 °C, and temperatures >40 °C are common. Winters are mild, with mean daily maximum temperatures ranging from 15–18 °C.
Fire History Mapping
Landsat imagery was used to map the fire history of the study region from 1972 to 2007. Imagery was acquired from Landsat Multispectral Scanner (MSS) (1972–1988), Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) (1989–2005). Fire events were digitised using composites of 15 satellite images at a resolution of 25 m² (TM and ETM+) to 50 m² (MSS) (envi version 4.2). Fire maps were then exported to ArcView 9.2 for data checking and the addition of attribute data. Fires initially were dated to the intervals of available satellite images (e.g. 1977–1980). We obtained the year of each fire by consulting spatial data prepared by land management agencies and by contacting local experts. All data were compiled in a geographic information system (GIS) to form a complete regional fire history map (spanning 35 years), which provided a basis for selecting study landscapes.
Landscape and Site Selection
Twenty-eight landscapes were selected for study, each a circular area 4 km in diameter (12·56 km2). Landscapes were selected to represent variation in two key properties: the proportional extent of vegetation aged >35 years post-fire and the diversity of post-fire ages present. Landscape selection was stratified to ensure spatial coverage of the Murray Mallee region, with 14 landscapes located in each of the southern and northern halves of the study area. The size of study landscapes represented a scale large enough to encompass a range of post-fire histories and a scale appropriate to the distribution and ecology of small mammal populations. Neighbouring landscapes were non-overlapping and separated by at least 2 km. In summary, study landscapes ranged from those largely covered by recently burnt vegetation, to those containing several post-fire age-classes, to those mainly composed of older vegetation.
Ten survey sites were established in each landscape. Site selection was stratified according to the proportional extent of post-fire age-classes of vegetation in the landscape (proportional extent of 0·02–0·10 = one site, 0·10–0·20 = two sites, 0·20–0·30 = three sites, etc.). Within each post-fire age-class, sites were located to represent the range of vegetation types and topographic positions (dune, swale, flat) present. Potential sites were first chosen using fire history and vegetation maps, and later checked for suitability in the field. Wherever logistically possible, sites were located in all four quadrants of the landscape (to ensure appropriate spatial coverage) and were typically >200 m apart (range 175–685 m).
Small Mammal Surveys
At each survey site, we established a line of pitfall traps comprising ten 20-L plastic buckets, spaced 5 m apart, connected by a continuous 300-mm-high flywire drift fence. Small mammal surveys were conducted at each of these 280 sites (i.e. 28 landscapes × 10 sites) four times starting in October–November 2006 (Spring), January–February 2007 (Summer), October–November 2007 (Spring) and January–February 2008 (Summer). Each survey period consisted of five consecutive nights of trapping and traps were checked daily. Elliott aluminium box traps (33 × 10 × 10 cm) were used to complement pitfall trapping in Spring, but not in Summer (to ensure the welfare of animals during high summer temperatures). In Spring surveys, five Elliott traps were placed adjacent to the pitfall line at each site. The species, mass, sex, age-class and reproductive status of each animal were recorded, and hair clipping was used to mark individuals to identify recaptures during each survey period. Nomenclature for mammals follows van Dyck & Strahan (2008).
A large wildfire modified three landscapes during Spring 2006. Because of changes to mosaic structure, the spring 2006 data from these three landscapes were excluded. The fire history of these landscapes was remapped, and small mammals were subsequently surveyed twice in Spring 2007. In total, we completed 56 000 pitfall trap nights and 14 000 Elliott trap nights. Prior to statistical analysis, data from the 10 sites established in each landscape were pooled. We also grouped data from small mammal surveys to represent two distinct time periods: Spring/Summer 2006/20007 and Spring/Summer 2007/2008. This grouping, based on the warm-season breeding period of several species, enabled an examination of the effects of rainfall on animal populations during the study.
Study landscapes were characterised by variables representing five mosaic properties. The extent of post-fire age-classes was represented by three subgroups: (i) the extent of vegetation <10 years post-fire, (ii) the extent of vegetation 10–35 years post-fire and (iii) the extent of vegetation >35 years post-fire. The diversity of post-fire age-classes was represented by the Shannon–Wiener diversity index of the proportion of each individual fire year in the landscape. The proportional extent of Triodia Mallee was used to describe the vegetation types occurring in the landscapes: those mosaics with less Triodia Mallee typically contained increasing proportions of Chenopod Mallee. Variables relating to fire history and vegetation type were calculated in fragstats version 3.3 (McGarigal et al. 2002).
The recent rainfall history of each landscape was represented by the amount of rainfall (mm) summed over a 12-month period. Monthly rainfall data were obtained from the closest weather station (Australian Bureau of Meteorology, http://www.bom.gov.au/) to each landscape [n = 9, mean distance to closest station = 25 km (range 4–50 km)]. Summed rainfall totals were calculated for 12-month periods prior to small mammal surveys at four time-lags: 3, 6, 9 and 12 months. Each time-lag was calculated as of November 2006 and November 2007 (the midpoint of the first survey during each grouped Spring/Summer trapping period).
Study landscapes were arrayed along a 200-km north-south axis (Fig. 1). Temperature increases and precipitation decreases from south to north (Pausas & Bradstock 2007). Northing of each landscape was included as an explanatory variable because it captured much of the biogeographic variation present in the study region. Summary data for each explanatory variable are shown in Table S1 in Supporting information.
We used generalised linear mixed models (GLMMs) to investigate small mammal responses to explanatory variables. GLMMs are a useful extension of generalised linear models and allow sources of correlation structure in the data to be modelled as random effects (Zuur et al. 2009). For models of individual species, the response variable was the capture rate of a species at a landscape, for each grouped survey period: Capture rate = total captures of individual animals/total trap nights. Recaptured animals were excluded from this calculation. For marsupials, which were not commonly recorded in Elliott box traps, capture rate was calculated from pitfall trap effort only. For rodents, which were captured in both pitfall traps and Elliott traps, capture rate was calculated using the combined trap effort of both survey techniques. For statistical analyses, capture rates were expressed as captures per 1000 trap nights and rounded to the nearest whole number. For the model of species richness, the response variable was the total number of native small mammal species captured in a landscape, for each grouped survey period.
Prior to fitting the regression models, we carried out graphical exploration of relationships between response and explanatory variables using scatter plots. This served two purposes: first, to identify the presence of nonlinear relationships and secondly, in the case of landscape variables with several correlated subgroups, to select (for each response variable) one subgroup from each set to include in regression models. Thus, for landscape variables relating to the extent of post-fire age-classes and rainfall history, only the subgroup with the strongest relationship with the response variable was included in further analyses. Selection of subgroups was also guided by our knowledge of the ecology of each study species. Landscape variables included in final models were not highly correlated (Pearson correlation coefficients < 0·5).
Generalised linear mixed models were implemented with a log-link function and Poisson distribution (Venables & Ripley 2002). Landscape variables were standardized by dividing by 2 standard deviations. This allowed effect sizes of variable coefficients to be compared (Gelman & Hill 2007). Explanatory variables were entered as linear fixed effects, as exploratory analyses identified no clear nonlinear relationships. Landscape was entered as a random effect (specifically, a random intercept) to account for possible temporal correlation of sites measured repeatedly (i.e. two grouped survey periods for each landscape, and therefore n = 56 for all regression analyses). GLMMs were fitted as full models to allow for the comparison of standardised variable coefficients. We interpreted landscape variables as having a significant influence on response variables where the standard errors of variable coefficients multiplied by two did not overlap zero. Model fit was assessed by Pearson’s correlation of the observed vs. predicted values, providing an indication of how closely the two sets of values agree. Pearson residuals were tested for spatial autocorrelation using spline correlograms (Zuur et al. 2009). All statistical analysis was undertaken in the R statistical package version 2.13.1 (R Development Core Team 2010). GLMMs were fitted using the package mass version 7.3-14 (Venables & Ripley 2002). We used this package because it can correct for overdispersion (commonly detected in count data) in Poisson GLMMs. Spline correlograms were produced in the extension package ncf version 1.1-3 (Bjornstad 2009).
Faunal surveys resulted in 1265 captures of seven mammal species. The small mammal assemblage was composed of two species of dasyurid marsupial, the mallee ningaui Ningaui yvonneae (554 captures) and common dunnart Sminthopsis murina (290 captures); two marsupial burramyids, the western pygmy possum Cercartetus concinnus (183 captures) and the little pygmy possum (18 captures); and three rodents, Bolam’s mouse Pseudomys bolami (30 captures), Mitchell’s hopping mouse Notomys mitchelli (11 captures) and the introduced house mouse Mus musculus (179 captures). The average species richness of native small mammals in a landscape, per grouped survey period, was 2·8 ± 1·0 SD (range 1–5).
Further details of small mammal captures are provided in Table S2 in Supporting information. We present statistical analyses of the four most commonly encountered species and the species richness of native small mammals. Spline correlograms provided no evidence of spatial autocorrelation in the residuals of the models presented (Fig. S1 in Supporting information).
The capture rate of C. concinnus was associated with the fire and rainfall histories of the study landscapes. Rainfall had the strongest influence: captures of C. concinnus were positively associated with 12-month rainfall totals at a time-lag of 6-months (Table 1). For example, the highest capture rate of C. concinnus (54 captures per 1000 trap nights) occurred at a landscape that had experienced above-average rainfall 6 months prior to trapping (Fig. 2a). Cercartetus concinnus was also positively associated with the proportional extent of vegetation aged >35 years post-fire (Fig. 2b) and with landscapes located in northern areas of the study region (Table 1). The diversity of post-fire age-classes and the cover of vegetation types had no clear impact on the capture rate of C. concinnus (Table 1). A correlation between observed and model-predicted capture rates indicated good model fit (r = 0·93).
Table 1. Results of generalised linear mixed models for individual species of small mammal and species richness of native small mammals
Species richness of native mammals
Species richness calculations excluded the introduced M. musculus. Values are standardised variable coefficients ± one standard error. Bold values indicate where coefficient standard errors multiplied by two do not overlap zero. Model fit was assessed using a Pearson correlation (r) of observed vs. model-predicted values (95% confidence intervals of correlations are shown in parentheses).
*Extent of post-fire age-class represented by subgroup: <10 yearsa, 10–35 years and >35 yearsb.
Although widespread across the study region, biogeographic context had a substantial influence on capture rates of N. yvonneae (Table 1). Ningaui yvonneae were more common in northern, more arid, locations (Fig. 2c). Fire history was also a significant influence. Landscapes characterised by high proportional cover of recently burnt vegetation (<10 years) contained lower numbers of N. yvonneae (Fig. 2d). Finally, capture success of N. yvonneae was positively associated with rainfall totals (at a 9-month time-lag) (Table 1). There was a moderate correlation between observed and predicted capture rates (r = 0·57) for this species.
We identified no significant relationships between S. murina and landscape variables, and model predictions were not correlated with observed capture rates (r = 0·11). The species was widespread, occurring in all 28 study landscapes, and the capture rates were relatively high (Table S2). These results demonstrate that this species occurs in landscapes with a broad range of fire histories under a variety of environmental conditions.
Regression modelling showed that capture rates of M. musculus were positively associated with the amount of recently burnt vegetation (<10 years) (Fig. 3a). Captures of these species also increased in landscapes with higher recent rainfall totals (at a 6-month time-lag) (Fig. 3b) and in more northern locations (Table 1). We found no clear patterns between capture rates of M. musculus and the diversity of post-fire age-classes or the extent of vegetation type (Table 1). A correlation between observed and predicted capture rates indicated good model fit (r = 0·78).
Rainfall was a substantial influence on the species richness of native small mammals (Table 1). Species richness was positively associated with 12-month rainfall totals (at a time-lag of 6 months) (Fig. 4a); higher richness occurred in landscapes where rainfall was above the long-term mean. Northing (biogeographic context) was also a positive influence (Fig. 4b). Regression modelling showed no evidence that the extent of older vegetation, the diversity of post-fire age-classes or the cover of Triodia Mallee were strong influences on the richness of native small mammals in these landscapes (Table 1). There was a moderate correlation between observed and predicted species richness (r = 0·58).
We surveyed small mammals in 28 landscapes selected to represent a range of fire histories, and therefore different fire mosaic properties. This approach – whereby the response and predictor variables characterise whole landscapes – allows a direct comparison of the influence of mosaic properties on faunal distributions. To our knowledge, this is the first application of this approach to investigate the responses of mammals to fire mosaic properties. We identified the total extent of habitat, particularly that of a suitable seral stage, as a key spatial property of landscape mosaics in fire-prone environments. Specifically, we found that areas of older post-fire vegetation provide important habitat for native small mammals. However, for most study species, the influence of recent rainfall and the biogeographic context of the landscape were dominant influences on capture rates.
Influence of Landscape Properties
Fire history and vegetation type
A primary mechanism proposed to link the distribution and abundance of fauna with the properties of fire mosaics relates to the amount of suitable habitat in a landscape. Under this hypothesis, animal species depend on resources that vary temporally in response to fire regimes (e.g. Fox 1982; Briani et al. 2004). Changes in the spatial extent of different fire histories will modify the total amount of these resources available in the landscape and thereby influence the abundance and distribution of species.
The extent of recently burned vegetation (<10 years) negatively influenced the capture rate of N. yvonneae. This species does not occur at sites immediately after they are burned (Kelly et al. 2010). The distribution of N. yvonneae is closely associated with cover of the hummock grass, Triodia (Bos, Carthew & Lorimer 2002), used for shelter and foraging. Triodia hummocks are highly flammable and are consumed by fire, and then recover over subsequent decades (Haslem et al. 2011). The proportional extent of older vegetation (>35 years) positively influenced the capture rate of C. concinnus. This preference is consistent with the species’ association with large trees and extensive cover of hummock grass (L.T. Kelly, unpublished data). The only species positively associated with recently burnt vegetation was the introduced M. musculus– a disturbance specialist.
The total amount of suitable habitat in a landscape may also be determined by the proportional extent of a particular vegetation type, irrespective of fire age. The lack of association of N. yvonneae with the proportional extent of Triodia Mallee in the study landscapes was surprising, given that the presence of hummock grass is a key determinant of its distribution (Bos, Carthew & Lorimer 2002). However, several landscapes dominated by Triodia Mallee were also those with large amounts of recently burnt vegetation. Although we captured too few N. mitchelli to undertake statistical analyses, this species showed a clear preference for landscapes dominated by Heathy Mallee. Six of 11 captures of N. mitchelli were from landscapes largely comprised of this relatively uncommon vegetation.
A second mechanism that potentially links the distribution and abundance of fauna with properties of fire mosaics relates to fire-induced habitat heterogeneity. Because different taxa exhibit different responses to fire history, heterogeneous mosaics composed of patches of differing fire history may provide a greater range of resources that facilitate a more species-rich biota than in homogeneous landscapes (Masters 1993; Briani et al. 2004). We found little evidence that ‘pyrodiversity begets biodiversity’ (Parr & Andersen 2006) at the scale of this study, despite deliberately selecting study landscapes with 1–6 post-fire age-classes to represent a broad range in pyrodiversity. This is consistent with our contemporaneous work on birds in these landscapes (Taylor et al. 2012). There was a trend for increasing species richness of native mammals with increasing diversity, but it failed to reach statistical significance (Table 1). The lack of such response may be because these species do not have markedly different requirements relating to fire history. Those species that do show a preference for particular seral stages typically favour longer unburnt areas (N. yvonneae and C. concinnus). Further, most species occur in a wide range of post-fire age-classes, albeit with differing probabilities of occurrence (Kelly et al. 2011); and for some, such as S. murina, there was little or no relationship between occurrence and post-fire vegetation age at trap sites (Kelly et al. 2011).
The diversity hypothesis also relates to individual species. Under this hypothesis, individual species may require complementary resources present in different components of the landscape: for example, patches of recently burnt vegetation adjacent to older vegetation (Firth et al. 2010). In the present study, this would be manifest by a significant influence of mosaic diversity (e.g. landscapes containing a mix of recently burnt and older vegetation) on capture rates of individual species. We found no evidence that the diversity of fire age-classes influenced individual species. This may be because the scale of heterogeneity sampled was not at the spatial scale of individual home ranges of these species. Rather, our focus was on larger-scale mosaics typical of the fire-induced heterogeneity that presently occurs in the region. Landscape complementation effects (at the scale of this study) are most likely to be observed for species with larger home ranges, such as the western grey kangaroo Macropus fuliginosus (Coulson 1993).
A fourth hypothesis relates to the presence of unburnt patches that provide refuges for animals. Following a fire, these patches can provide a source for recolonisation into disturbed areas (Bradstock et al. 2005). While we were unable to specifically test this hypothesis (landscape selection was focussed on other properties of fire mosaics), it seems likely that the presence of unburnt patches will benefit individual species. Immediately following a wildfire that affected three of our landscapes, there was a marked reduction in the occurrence of N. yvonneae and in the species richness of small mammals (Kelly et al. 2010). The Murray Mallee region experiences very large fires (>100 000 ha) within the region on a regular basis (c. 20 year intervals), and in the absence of unburnt refuges, recolonisation of burnt areas at this spatial scale may be slow.
Biogeographic context and rainfall
The geographic scale at which studies must be undertaken when ‘whole landscapes’ are the unit of study means that environmental gradients across a study region often influence the distribution of species or composition of assemblages (Bennett, Radford & Haslem 2006; Mortelliti et al. 2011). Here, the biogeographic context of the landscape was an important influence on three species. For example, N. yvonneae was more common in landscapes in the northern, more arid, part of the study area. Differing environmental histories or bioclimatic influences between the northern and southern landscapes may determine specific food and shelter resources. Biogeographic context also was an important influence on several less-common species. For example, P. bolami was restricted to the northern half of the region, while N. mitchelli was recorded only in the south.
The distribution and abundance of species are influenced not only by the spatial properties of landscapes, but by a range of processes that may be associated with, or independent of, particular spatial patterns. These include processes such as competition, predation and climatic variation (Meserve et al. 2003). In this study, spatial and temporal variation in rainfall influenced several species. Recent rainfall history was a particularly strong influence on C. concinnus with higher capture rates recorded 6 months after a year of high rainfall. Similarly, M. musculus, N. yvonneae and the species richness of native mammals were positively associated with rainfall totals. This association of small mammals with rainfall is likely to be driven by changes in primary productivity, and therefore the abundance of food resources (e.g. plants, seeds, invertebrates and pollen). This is consistent with other studies from the Australian arid zone, where rainfall-driven processes have a large influence on small mammal population dynamics (Southgate & Masters 1996; Letnic & Dickman 2010). We propose that higher capture rates of small mammals following high rainfall represent an increase in the abundance of regional populations, as well as a potential increase in the ‘trappability’ of individuals.
Synthesis and Applications
A primary management goal in many fire-prone environments is to protect, maintain and enhance biodiversity. Patch mosaic burning (i.e. introducing heterogeneity in burning patterns) is widely advocated to achieve such goals (Parr & Andersen 2006). This is the case in mallee ecosystems, where fire plays an important role in ecosystem function and is widely used as a management tool (Bradstock & Cohn 2002). This includes using fire to maintain a range of post-fire ages in the landscape to maintain vegetation communities and provide habitat for animals, and to create areas with reduced fuel loads to avoid large, reserve-scale fires (Sandell et al. 2006). In combination with ‘natural’ fire regimes, management practices will create a range of fire mosaics.
Management of fire mosaics for biodiversity conservation will need to consider both the characteristics of the ecosystem under consideration and the species that are a priority for regional conservation efforts. In this study, we identified the proportional extent of older vegetation as an important spatial property of landscape mosaics for native small mammals in mallee ecosystems. Older vegetation is heterogenous; it includes dense, patchy cover of spinifex, complex ground litter and larger trees, which provide high-quality habitat components for several small mammal species. Given that many important habitat attributes take decades to develop (Haslem et al. 2011), the retention of older mallee vegetation (e.g. >35 years) in the landscape should be encouraged.
Although we found little support for the ‘pyrodiversity hypothesis’, several important considerations remain. First, we focused here on the direct effects of fire mosaics on small mammals (i.e. how the present mosaic influences species distributions). Fire management will also need to consider the indirect effects of fire, such as the role of fuel reduction in protecting and retaining unburnt vegetation in the landscape. Secondly, at a regional scale, there clearly is a need for a range of fire age-classes to ensure an ongoing supply of successional stages. Thirdly, we modelled the responses of relatively common species. Several other species were relatively uncommon in the region (e.g. P. bolami and N. mitchelli), and further research is required to identify the influence of landscape-scale properties on these species. Finally, this study was undertaken in an ecosystem that has been modified over the last 150 years. One-third of the historical mammal assemblage is regionally extinct (Bennett, Lumsden & Menkhorst 1989), with declines greatest among small- to medium-sized mammals (e.g. bandicoots, bettongs and stick-nest rats). Any re-introductions of these species will need to consider their possible requirement for fine-scale heterogeneity. The present day mammal assemblage may represent the most resilient species.
Finally, this work highlights the importance of the spatial properties of fire mosaics, which are amenable to manipulation and management, and other ecological processes that also determine habitat quality, in this case, spatial and temporal variation in rainfall-driven productivity. Thus, while the management of fire regimes can be used as a tool to enhance biodiversity, it is critical to recognise that other ecological processes interact with fire-generated spatial patterns to determine the distribution of species and composition of assemblages. Future studies of fire mosaics will benefit from considering the interactions with spatial patterns of other processes, particularly those such as predation and grazing by herbivores that managers can influence.
Funding and support were provided by Land and Water Australia, Department for Environment and Heritage (SA), Parks Victoria, Department of Sustainability and Environment (Vic.), Mallee Catchment Management Authority, NSW National Parks and Wildlife Service, Department of Environment and Climate Change (NSW), Lower Murray–Darling Catchment Management Authority, Natural Heritage Trust, Birds Australia, Australian Wildlife Conservancy, and the Murray Mallee Partnership. We are grateful to the Doyle and Barnes families for access to Petro and Lethero Stations, respectively. Thanks to Sarah Avitabile, Lauren Brown, Kate Callister, Angie Haslem and Sally Kenny; and the many volunteers who assisted with fieldwork. Three anonymous referees made useful comments on the manuscript.