Conservation management of eastern Australian farmland birds in relation to landscape gradients


  • Jan Hanspach,

    Corresponding author
    1. Fenner School of Environment and Society, The Australian National University, Canberra ACT 0200, Australia
    2. UFZ, Helmholtz Centre for Environmental Research – UFZ, Department of Community Ecology, Theodor-Lieser-Strasse 4, 06120 Halle/S., Germany
      Correspondence author. E-mail:
    Search for more papers by this author
  • Joern Fischer,

    1. Fenner School of Environment and Society, The Australian National University, Canberra ACT 0200, Australia
    Search for more papers by this author
  • Jenny Stott,

    1. Fenner School of Environment and Society, The Australian National University, Canberra ACT 0200, Australia
    Search for more papers by this author
  • Karen Stagoll

    1. Fenner School of Environment and Society, The Australian National University, Canberra ACT 0200, Australia
    Search for more papers by this author

Correspondence author. E-mail:


1. Birds inhabiting farmland are of conservation concern around the world. In Australia, conservation management has focused primarily on woodland environments. By contrast, semi-natural open areas have received less attention. We argue that long-term conservation strategies should consider broad gradients of environmental conditions. Otherwise, there is a risk that semi-natural open areas will degrade through ‘benign neglect’, and currently common species using these areas will become uncommon.

2. We examined how birds responded to three environmental gradients in an Australian livestock grazing landscape: tree density, grazing intensity and nutrient enrichment. First, we investigated changes in species composition across the environmental gradients in multivariate space. Secondly, we modelled species richness and the response of selected individual species in relation to the gradients. Thirdly, we examined if there were patterns in guild composition and body mass distribution.

3. Tree density was the primary driver of virtually all patterns observed. Species richness peaked at moderately high tree densities. With increasing tree density, species composition changed, foraging guild composition changed and the median body mass of bird species decreased. Small insectivores were more likely to occur in areas with high tree densities, whereas large granivores were more likely to occur in areas with relatively low tree densities. Grazing intensity and nutrient enrichment were less strongly related to bird distribution patterns, although the indirect effects of these gradients may be substantial because they affect tree regeneration.

4. sSynthesis and applications. Relatively dense woodland patches were important for species already of conservation concern, lending support to their active conservation management, for example through livestock exclusion. However, semi-natural open areas also were used by many birds, which represented a different mix of body sizes and foraging guilds. Scattered trees occurring at a range of densities are key habitat elements in semi-natural open areas. However, many scattered trees are dying and are not being replaced by natural regeneration or tree planting. If areas with scattered trees continue to degrade, there is a risk that currently common farmland birds will decline. Management strategies aiming to maintain scattered trees therefore are important, including the planting of individual trees and the adoption of grazing practices that allow for natural tree regeneration.


Birds inhabiting farmland are of conservation concern around the world (Hughes, Daily & Ehrlich 2002; Donald et al. 2006; Ranganathan et al. 2008). They are valued in their own right, can help to engage local people in conservation issues (Lindenmayer & Fischer 2002), and also provide valuable ecosystem services such as pest control (Kirk, Evenden & Mineau 1997; Sekercioglu 2006). In Europe, conservation activities focus primarily on open-country species, which are declining due to agricultural intensification and landscape homogenization (Donald et al. 2006; Butler, Vickery & Norris 2007). By contrast, conservation activities in recently cleared landscapes typically focus on the species characteristic of the indigenous ecosystem that existed prior to clearing. For example, in Australia, woodland birds are a major conservation concern (Barrett, Ford & Recher 1994; Ford et al. 2001; Lindenmayer et al. 2010), but open-country species receive little attention, and anthropogenic farmland environments are not a primary target of conservation policies (Attwood et al. 2009). Ecological studies in Australia traditionally also have focused primarily on woodland patches (Barrett, Ford & Recher 1994; Watson, Freudenberger & Paull 2001) and corridors (Kavanagh, Stanton & Herring 2007; Lindenmayer et al. 2010). Recently, however, there has been a significant shift in research foci, with increasing interest in open environments such as pastures with scattered trees (see Manning, Lindenmayer & Barry 2004; Haslem & Bennett 2008b; Maron et al. 2010).

An exclusive focus on woodland birds or woodland environments runs the risk of landscape-wide degradation of semi-natural farmland environments through their ‘benign neglect’. For example, scattered trees may disappear due to senescence and a lack of natural regeneration (Gibbons et al. 2008), and native pastures may be lost through fertilizer application or heavy livestock grazing (McIntyre & Lavorel 2007). Such changes may have unintended negative consequences for a wide range of birds, including woodland species which occasionally use open environments, but also a suite of currently common species whose core habitat is in semi-natural, open environments. Developing long-term conservation strategies that cater for a broad range of species requires an understanding of how different bird species respond to a wide range of environmental conditions, from intensively used, treeless pastures to intact woodlands. Studying environmental gradients is particularly important in Australian farmland, because gradual changes in tree density and land use intensity are widespread (McIntyre & Barrett 1992).

In this study, we investigated the response of birds to three environmental gradients that are directly influenced by farm management: (i) tree density; (ii) nutrient enrichment; and (iii) grazing intensity. Tree density is a major driver of many ecological processes (Wilson 2002; Eldridge & Freudenberger 2005), but few studies have explicitly designed studies with reference to tree density (Lumsden & Bennett 2005). Farmers influence tree density directly through planting or clearing trees, but also indirectly through encouraging or preventing natural tree regeneration (Dorrough, Vesk & Moll 2008b). Nutrient enrichment is widely recognized as a proxy for land use intensity in agricultural landscapes (McIntyre & Lavorel 2007; Kleijn et al. 2009; Stoate et al. 2009), but few studies on vertebrates have considered nutrient gradients (but see Landsberg & Wylie 1983; Dorrough et al. 2008a; Oldland, Taylor & Clarke 2009). Finally, livestock grazing is recognized to affect fauna, but most authors have classified grazing intensity into a small number of broad categories (Martin & Possingham 2005; Martin & McIntyre 2007; Eyre et al. 2009). Moreover, the degree of livestock rotation, and hence the duration of rest periods between grazing events, has received little attention to date (Bock & Bock 1999; Dorrough et al. 2008a; Fischer et al. 2009).

The overarching goal of our work was to understand bird distribution in relation to these gradients, and from this understanding, to derive appropriate guidelines relevant to farm management. First, we investigated changes in species composition across the environmental gradients in multivariate space. Secondly, we modelled species richness, and thirdly, the response of selected individual species, in relation to the gradients. Finally, we examined if there were patterns in guild composition and body mass distribution across the gradients. Our findings demonstrate that farmland bird diversity depends on the maintenance of a broad range of tree densities, including both woodland areas and areas of scattered trees.

Materials and methods

Study area and experimental design

We studied a 1 000 000 ha area in the Upper Lachlan Catchment of New South Wales, Australia (148·7°E, −33·9°N). The region is undulating, receives c. 600 mm of rainfall annually, and the long-term mean of daily maximum temperature is c. 20 °C. Prior to the 1800s, the vegetation was dominated by temperate woodland in the valleys (e.g. Eucalyptus blakelyi, E. melliodora) and dry forest on the hilltops (e.g. E. macrorhyncha) (Gibbons & Boak 2002). Understorey is largely absent, especially in areas grazed by livestock. We worked on 33 farms (ranging in size from 236 to 3,036 ha; median 900 ha) that were dominated by sheep and/or cattle grazing [80%± 4% of land grazed per farm (mean ± SE)]. We only considered locations that had been under broadly the same grazing regime since at least mid-2002.

To capture the gradient in tree density, we established 140 2-ha survey sites, including 38 ‘paddock sites’ (c. 10 or fewer crowns discernible on aerial photographs), 37 ‘scattered tree sites’ (c. 10–40 crowns), 33 ‘grazed woodland sites’, 18 ‘ungrazed woodland sites’ (both with dense crown cover) and 14 ‘revegetation sites’ planted within the last 20 years (e.g. Freudenberger, Harvey & Drew 2004; Lindenmayer et al. 2010). Almost all sites were in separate paddocks, and were separated by several hundred metres or more. Broad classes of tree density were refined by measuring the actual density of trees within a given site, where each stem taller than 130 cm was measured (Fischer et al. 2009). In addition, we calculated the proportion of tree cover within concentric circles around the centre of the site, using radii of 100 m, 200 m, 400 m and 800 m. The proportion of tree cover was based on remotely sensed data classified from SPOT imagery with a 10 m resolution, and was highly correlated with actual tree density (for details, see Fischer et al. 2010).

To capture gradients in grazing intensity, we strategically selected sites to cover the full range of grazing regimes in the study area, keeping independent stocking rate and extent of livestock rotation (Fischer et al. 2009). For each site, we obtained information from farmers about its annual mean stocking rate (measured in DSE = dry-sheep-equivalent; a 48–50 kg wether), and the number of days a site was grazed per year.

To capture the gradient in nutrient enrichment, we measured available topsoil phosphorus at every site via the Colwell method (Rayment & Higginson 1992, p. 64). At 32 locations distributed throughout each site, we took 2 cm diameter cores to a depth of 7 cm. All cores for each site were mixed and dried prior to analysis. Further details on soil sampling and analysis are provided by Fischer et al. (2009). We acknowledge that available phosphorus varies naturally between soil types. However, previous work has found that available phosphorus was a good proxy of the history of fertilizer use at a site (Dorrough & Moxham 2005). Additional covariates visually assessed at each site included the percent cover of logs and shrubs.

Bird surveys

Each site was surveyed for birds in four separate 20 min area searches, following Barrett et al. (2003). Two observers (J.F. and J.S.) each searched every site once in spring 2007 and again in spring 2008, on fine mornings. In each year, due to logistic constraints, all sites within a farm were surveyed on the same day by both observers, but the order that farms were visited was changed between years. We acknowledge that detectability of birds may differ between sites (Johnson 2008). Although we cannot rule out such differences, they are likely to be small because all sites were relatively open in structure and observers were experienced. Hence, we consider our estimates of species presence to be reliable indices of true presence (Johnson 2008).

Data analysis

An initial multivariate analysis aimed to elicit differences in bird community composition between sites. For each site, presence/absence data were pooled across the four 20 min surveys. Non-metric multidimensional scaling (Kruskal 1964) based on a Jaccard dissimilarity matrix was used to visualize differences in bird species composition between sites; and vectors describing continuous environmental variables were superimposed. Only vectors significantly correlating with the ordination in a permutation test (< 0·05 after 10 000 permutations) were plotted. The multivariate analysis was performed using the R-package vegan (Oksanen et al. 2009).

To model species richness, we excluded revegetation sites because they were not remnant vegetation and therefore substantially different in vegetation structure and species composition. For the remainder of sites (n = 126), we considered the bird richness recorded in each 20 min survey (126 × 4 = 504). Species richness was modelled using generalized linear mixed effects models (Pinheiro & Bates 2000) with Laplace approximation, Poisson error, natural logarithm as link function, and random effects initially specified as ‘farm’ and ‘site nested in farm’. Because the variance component associated with the farm level was low, only the single random effect of ‘site’ was retained. The following fixed effects were included as explanatory variables: survey year, tree density (both linear and quadratic terms), percentage cover of shrubs and of logs, available phosphorus, as well as number of grazing days, stocking rate and the interaction between the number of grazing days and stocking rate. We simplified the model by deleting non-significant terms and comparing the resulting model with the more complex one using analysis of variance. Prior to the analysis, environmental variables were log transformed (except number of grazing days) and scaled by subtracting the mean and dividing by the standard deviation. This was done to remove the skew of variables and allowed us to directly compare the model coefficients of different variables. To assess the spatial scale of the tree density gradient most strongly related to birds, we compared species richness models using tree density at the site level against alternative models using remotely sensed tree cover at a series of increasingly large radii (see above; Ricketts et al. 2001). Tree density at the site level provided the best fit (as measured by the Akaike Information Criterion), and hence remotely sensed tree cover was not used in subsequent analyses.

Preliminary models showed that species richness was negatively affected by the presence of the noisy miner Manorina melanocephala. The noisy miner is a colonially breeding native honeyeater that thrives in agricultural landscapes and aggressively excludes other species from its territory (Piper & Catterall 2003; Clarke & Oldland 2007). Models accounting for tree density showed that presence of the noisy miner reduced overall species richness significantly by 0·8 ± 0·05 (SE) species per 20 min; and reduced richness of species smaller than the noisy miner (<60 g) by 0·5 ± 0·06 (SE) species per 20 min. Because we were not primarily interested in the effect of the noisy miner, we excluded 105 survey periods where the noisy miner was present from further analyses; this left 399 separate 20 min periods. There were 14 sites where the noisy miner was detected in all four surveys. Removal of these sites from the analysis did not affect the continuity of the environmental gradients because the noisy miner occurred only at a subset of sites, most of which had intermediate tree densities.

To investigate selected individual species, we modelled eight of the 21 most widespread species (i.e. occurring in more than 10% of the 20 min periods). The species were selected to exhibit a representative range of responses to the environmental gradients (Table S1, Supporting information). As for species richness, we used GLMM with Laplace approximation, but with a binomial error and logistic link function.

To describe the possible effects of the tree density gradient on species traits, we collated data on foraging guild and body mass for all species (HANZAB 1990–2007; see Table S1). For each site, we calculated the proportion of species belonging to a particular foraging guild. Guild composition (with individual guilds summing to 100% at each site) was then modelled as a multinomial response variable using a vector GLM (Yee 2010), with linear and quadratic terms of tree density as explanatory variables. Water birds and birds feeding primarily on vertebrates were excluded from this analysis because they were rare and their occurrence was highly idiosyncratic. Finally, we calculated the median body mass of all bird species at a given site and fitted a local polynomial regression (Cleveland, Grosse & Shyu 1992) to visualize the trend of bird body mass along the tree density gradient. The nature of the trait analyses was exploratory, and we did not account for the grouping of experimental units.


Site types differed in bird species composition (Fig. 1 and Fig. S1, Supporting information). Bird community composition changed along the tree density gradient from open paddock sites, through scattered tree sites and grazed woodland sites, to ungrazed woodland sites (Fig. 1). Stocking rate, number of grazing days and soil phosphorus tended to be higher at paddock and scattered tree sites, whereas shrub cover and log cover tended to be higher in grazed and ungrazed woodland sites (Fig. 1).

Figure 1.

 Non-metric multidimensional scaling of bird species composition based on a Jaccard dissimilarity matrix of 126 sites containing 106 species (three axes; stress = 16·9; using only two axes yielded stress values >20). The directions of the vectors indicate correlations between the continuous variables and the ordination configuration, and the relative lengths of the vectors represent the strength of the correlation. (a) first and second axis; (b) first and third axis. Environmental variables were superimposed on the ordination plot. (Site types: P = paddocks, S = scattered trees, GW = grazed woodland, UW = ungrazed woodland; environmental variables: tree = tree density, tree.2 = tree density as a quadratic term, shrub = shrub cover, log = log cover; dse = dry-sheep-equivalent (stocking rate), d.graz = days grazed per year, p = available phosphorus).

Species richness was significantly related to tree density and survey year, but not to any other variables (Table 1; Fig. 2). Individual species differed in their response to tree cover, exhibiting positive and negative linear or quadratic responses (Table 1; Fig. 3). Additional variables significantly related to some individual species were stocking rate, number of grazing days, shrub cover and log cover (Table 1).

Table 1.   Summary of significant fixed effects in generalized linear mixed effects models for species richness and individual species (Trees = tree density; DSE = dry-sheep-equivalent (stocking rate). Note that tree density, DSE, shrub and log cover were log transformed and scaled prior to analysis. For a given parameter, the top row shows the estimate, the middle row the SE, and the bottom row the P-value. The column ‘Site’ contains the variance associated with the random effect. See Table S1 in Supporting information for scientific names of species
ResponseTreesTrees squaredDays grazedDSEShrubsLogsYear 2007Year 2008SiteAIC
Species richness0·492−0·259    1·9832·1420·081499·7
0·0390·041    0·053   
<0·001<0·001    <0·001<0·001  
Richard’s pipit−2·361 −0·655   −4·435−3·9322·918209·5
0·588 0·349   0·6200·590  
0·000 0·061   0·0000·000  
Australian magpie0·691   −0·661 −0·245−0·1190·542524·1
0·158   0·162 0·1670·168  
0·000   0·000 0·1430·480  
Noisy friarbird1·018     −2·133−1·8120·442332·2
0·175     0·2280·228  
0·000     0·0000·000  
Striated pardalote1·560−1·348 0·780  0·4770·5333·540427·2
0·3260·310 0·285  0·3750·379  
0·0000·000 0·006  0·2030·159  
Eastern rosella −1·185    0·5600·8601·615472·2
 0·208    0·2730·278  
 0·000    0·0400·002  
Crimson rosella1·299−1·128    −0·664−0·7123·088400
0·3300·326    0·3590·362  
0·0000·001    0·0650·049  
Red wattlebird1·414−0·888    −1·477−1·5762·427339·9
0·3690·338    0·3560·362  
0·0000·009    0·0000·000  
Rufous whistler2·406    0·701−3·613−3·5813·528234·9
0·471    0·3300·5550·555  
0·000    0·0340·0000·000  
Figure 2.

 Response of bird species richness to tree density (number of trees per 2 ha). Over-plotting of multiple points with the same coordinates is visualized by increasingly darker shades of grey. Fitted values (solid line) and 95% confidence interval (dotted lines) correspond to the GLMM summarized in Table 1 (for the year 2008).

Figure 3.

 Probability of detection of selected bird species in response to tree density (number of trees per 2 ha) for the year 2008 derived from generalized linear mixed effects models (Table 1). Species abbreviations: AM = Australian magpie, CR = Crimson rosella, ER = eastern rosella, NF = noisy friarbird, ReW = red wattlebird, RP = Richard’s pipit, RuW = Rufous whistler, StP = striated pardalote. See Table S1, for scientific names.

With an increase in tree density, guild composition changed (Fig. 4a, b; see Table S1 for the guild and body mass of each species) and the median body mass of birds declined (Fig. 4c). At sites with low tree density, granivores and ground insectivores accounted for a large proportion of species, whereas sites with high tree density were characterized by a larger proportion of arboreal and shrub insectivores (Fig. 4a, b).

Figure 4.

 Association of bird species’ traits with tree density (number of trees per 2 ha). See Table S1 for the traits of individual species. (a) Observed proportion of foraging guilds by decantiles of tree density (N = nectar and plant feeding, aeI = aerial insectivores; arI = arboreal insectivores; sI = shrub insectivores; gI = ground insectivores; G = granivores). (b) Predicted proportion of foraging guilds as a function of tree density as estimated by a multinomial model. (c) Median body mass of birds in relation to tree density. A local polynomial regression was fitted to visualize the trend (solid line = fitted values, dotted lines = SE).


We investigated the effects of three environmental gradients on bird distribution patterns: tree density, grazing intensity and nutrient enrichment. All three gradients are directly amenable to management, with tree density by far the most important gradient for birds, underlying virtually all patterns observed.

Tree density

Nearly twenty years ago, McIntyre & Barrett (1992) proposed that Australian grazing landscapes were best conceptualized as ‘variegated’, characterized by gradients of tree density and land use intensity. Gradient approaches are increasingly used in Europe (Haenke et al. 2009; Breitbach et al. 2010; Smith et al. 2010) and North America (Bridges, Crompton & Schaefer 2007; McGarigal, Tagil & Cushman 2009), and are also beginning to be used in Australia (Price et al. 2009). They offer a useful perspective of landscapes that complements more traditional mosaic approaches (Forman 1995; Bennett, Radford & Haslem 2006; Fischer & Lindenmayer 2006). Our focus on a broad gradient of tree densities allowed us to understand changes in (i) bird community composition; (ii) species richness; (iii) individual species occurrence; and (iv) guild composition and body mass patterns.

With increasing tree density, structural attributes such as logs and shrubs tended to increase (Fig. 1). Logs and shrubs are significantly related to some woodland birds (Seddon, Briggs & Doyle 2003; Mac Nally & Horrocks 2007), and the gradient in tree density therefore described an important structural gradient of landscape modification. As tree density increased, bird community composition shifted from grassland species such as Richard’s pipit, through scattered tree sites with parrots such as the eastern rosella and superb parrot, to woodland and forest specialists such as the eastern yellow robin and brown treecreeper (Fig. 1 and Fig. S1; see Table S1 for scientific names; see also Fisher 2001). Species richness was highest at sites with relatively high tree density, which probably resembled most closely natural woodland conditions prior to European landscape modification. Species richness slightly declined at extremely high tree densities (Fig. 2), such as in regrowth stands of E. macrorhyncha, which can exhibit a relatively uniform age profile (J. Fischer, pers. obs.). Tree cover at spatial scales larger than 2 ha was less strongly related to species richness than tree density within a given 2 ha site. Context effects hence were relatively less important than site-level effects (see also Martin et al. 2006; Bowen et al. 2009). While context effects significantly influence animals in some situations (Lindenmayer et al. 1999), site-level conditions are potentially more important for highly mobile species such as birds (but see Lindenmayer et al. 2010).

The response of species was highly individualistic, suggesting that what constituted suitable habitat for one species may not constitute suitable habitat for other species (Hall, Krausman & Morrison 1997; Lindenmayer & Fischer 2007). Such species–specific differences are to be expected and have led to a central criticism of patch-based landscape classifications – often, there may be no obvious way to define universally applicable ‘patches’, because different species peak in different parts of the landscape (Fischer, Lindenmayer & Fazey 2004; Fig. 3). In this study, all densities of tree cover were used by some species, including locations where trees were highly isolated (Law, Chidel & Turner 2000; Fischer & Lindenmayer 2002a). However, the composition of foraging guilds changed along the gradient of tree density, from sites with a large proportion of granivores through to sites with increasingly large proportions of shrub and arboreal insectivores (Fig. 4a, b). The unique responses displayed by individual species and changes in foraging guild composition suggest that landscape heterogeneity as a whole is important to maintain the beta diversity of birds. This finding is mirrored by many studies around the world (Luck & Daily 2003; Tscharntke et al. 2005; Haslem & Bennett 2008 a,b; Ranganathan et al. 2008), suggesting landscape heterogeneity is one of the key determinants of farmland biodiversity (Benton, Vickery & Wilson 2003).

Finally, tree density described a gradient in landscape texture. A fine texture with high tree density, in turn, provided suitable habitat for small-bodied species, whereas a coarse texture was used primarily by larger species (Fig. 4c). This finding supports one facet of the ‘textural discontinuity hypothesis’ (Holling 1992), namely that increasingly fine landscape texture is associated with increasingly smaller-bodied animals. As with landscape heterogeneity, the effect of landscape texture appears to be generic, with predictable links between body sizes and landscape texture demonstrated in different ecosystems around the world (see also Gunnarsson 1992; Tellería & Carrascal 1994; Fischer, Lindenmayer & Montague-Drake 2008).

Grazing and nutrient enrichment

Globally, landscape change in grassy ecosystems is driven by livestock grazing and nutrient enrichment (McIntyre & Lavorel 2007). Both fundamentally alter vegetation composition (Dorrough et al. 2006; McIntyre & Lavorel 2007) and can inhibit natural tree regeneration (Dorrough & Moxham 2005; Fischer et al. 2009). In addition, livestock grazing can lead to the simplification of woodland structure (Smith et al. 1996; Tasker & Bradstock 2006), thereby negatively affecting some woodland species (Maron & Lill 2005; Martin & Possingham 2005; Martin & McIntyre 2007).

Our findings on grazing intensity are broadly consistent with previous work in Australia. In ungrazed woodlands, we found several species of conservation concern, including the brown treecreeper and eastern yellow robin (Fig. 1 and Fig. S1; Barrett et al. 2007; Ford et al. 2009). By contrast, ground-foraging species such as the crested pigeon and magpie-lark were predicted in previous work to increase with grazing pressure (Martin & Possingham 2005), and our findings were consistent with such predictions (Fig. 1 and Fig. S1). Perhaps surprisingly, relationships with stocking rate were relatively uncommon in our analyses, compared to the more dominant gradient of tree density (Table 1). Few significant effects were found for effects of grazing duration, providing only weak evidence that rotational grazing directly benefitted bird conservation; only Richard’s pipit, a ground-foraging and ground-nesting species, responded positively to the rest periods provided by rotational grazing (Table 1). Similarly, no direct relationships were found with soil nutrient levels. Despite a lack of strong evidence for direct benefits of reduced fertilizer use and rotational grazing for birds, the indirect benefits of these practices may be substantial, because tree regeneration is significantly enhanced under low-input rotational grazing (Fischer et al. 2009). Hence, these practices may help to maintain regional scale tree cover in the long term.

Conservation implications

We demonstrated that different tree densities are used by different species representing a range of guilds and body sizes. From a conservation management perspective, it is therefore important that a variety of tree densities is maintained and perpetuated into the future.

Dense areas of trees, either in patches (Watson, Freudenberger & Paull 2001; Seddon, Briggs & Doyle 2003) or linear strips (Bentley & Catterall 1997; Kavanagh, Stanton & Herring 2007; Lindenmayer et al. 2010), traditionally have received most attention in conservation management in Australia. Many declining woodland birds have their core habitat in such areas (Watson, Freudenberger & Paull 2001; Seddon, Briggs & Doyle 2003; this study). The value of dense areas of trees can be significantly enhanced by excluding livestock grazing, which improves structural complexity and facilitates natural tree regeneration (Seddon, Briggs & Doyle 2003; Spooner & Briggs 2008).

Open areas are unlikely to provide core habitat for woodland specialists. However, woodland specialists may use scattered trees in such areas as supplementary habitat or as stepping stones (Fischer & Lindenmayer 2002 a,b). Perhaps more importantly in the long term, many other bird species currently not of conservation concern extensively use semi-natural vegetation in otherwise open areas (e.g. scattered trees). This situation is analogous to Western Europe, where structural elements such as vegetated field margins or hedgerows are critically important in maintaining biodiversity in otherwise open farmland (Benton, Vickery & Wilson 2003; Tscharntke et al. 2005). Unless such elements are maintained in Australian systems throughout the landscape, it is possible that Australia will follow Europe’s trajectory, with currently common farmland birds set to decline over the coming decades. To prevent such a trajectory, it is important to understand which conditions are suitable for which species of birds, and to maintain semi-natural open areas that are deemed valuable.

Maintaining semi-natural vegetation in open areas requires a mix of management approaches. Planting scattered trees with individual tree guards is practiced relatively rarely to date but should be considered more widely, and may need to be supported through relevant state or federal government grants. More generally, conditions for natural tree regeneration can be enhanced by transitioning from continuous livestock grazing to rotational grazing with prolonged rest periods, and by reducing fertilizer use (Fischer et al. 2009).


We gratefully acknowledge funding by the Australian Research Council to J.F. We also thank all participating landholders who granted us access to their properties. R. Forrester provided valuable advice on the experimental design. Three anonymous referees and the Associate Editor provided constructive comments that helped to improve the manuscript.