Landscape Ecol (2014) 29:1249–1259 DOI 10.1007/s10980-014-0065-4

RESEARCH ARTICLE

Effects of landscape composition and connectivity on the distribution of an endangered in agricultural landscapes

Simon J. Watson • David M. Watson • Gary W. Luck • Peter G. Spooner

Received: 5 September 2013 / Accepted: 30 June 2014 / Published online: 14 July 2014 Ó Springer Science+Business Media Dordrecht 2014

Abstract The extent and connectivity of individual extent of semi-arid woodlands dominated by Euca- habitat types strongly affects the distribution and lyptus largiflorens. Connectivity between E. camal- abundance of organisms. However, little is known of dulensis forest (principal nesting habitat) and mallee how the level of connectivity and the interactions (preferred feeding habitat) was a strong predictor of between different habitat types influences the distri- nest locations. Our study shows that the suitability of bution of species. Here, we used the geographically fragmented agricultural landscapes for supporting restricted and endangered regent parrot species can be greatly affected by connectivity and anthopeplus monarchoides as a case study to examine interactions between preferred and non-preferred the importance of composition and connectivity habitats. For species that require complementary between different elements in 39 complex landscape habitats such as the regent parrot, conservation mosaics (each 10 km radius). We compiled a database management activities may be ineffective if they of 674 regent parrot nesting records, regional vegeta- simply focus on a single core habitat type or the tion maps and measures of multipath connectivity impacts of human land uses without regard to the between core vegetation types under different scenar- interrelationships among landscape elements. While ios of resistance to movement provided by landscape increasing the amount of primary preferred habitat elements. The occurrence of regent parrot nests was should remain a cornerstone goal, increasing the strongly affected by landscape composition, being extent and improving connectivity with alternative positively related to the extent camaldul- landscape elements also should be priority manage- ensis riverine forest, but negatively related to the ment objectives.

Keywords Connectivity Habitat fragmentation Electronic supplementary material The online version of this article (doi:10.1007/s10980-014-0065-4) contains supple- Landscape complementarity Landscape mosaic mentary material, which is available to authorized users. Mallee Regent parrot Polytelis anthopeplus Eucalyptus camaldulensis S. J. Watson (&) D. M. Watson G. W. Luck P. G. Spooner Institute for Land, Water and Society, Charles Sturt University, Albury, NSW 2640, e-mail: [email protected] Introduction

S. J. Watson Department of Zoology, La Trobe University, Bundoora, Spatial heterogeneity of ecosystems is fundamental VIC 3086, Australia for biodiversity and promotes species richness and 123 1250 Landscape Ecol (2014) 29:1249–1259 stability via increased persistence, reduced dominance However, regent make regular movements to and decreased sensitivity to disturbance (Tews et al. and from breeding sites to forage and roost, usually in 2004; Allouche et al. 2012). Landscape heterogeneity mallee shrublands (Higgins 1999), which consist of is especially critical for those species reliant on low multi-stemmed Eucalyptus vegetation on nutrient multiple landscape elements for survival (Dunning poor soils (Haslem et al. 2010). While originally et al. 1992a; Forman and Godron 1986); for example, extensive, widespread clearing has reduced the extent using one vegetation type for foraging and another for of foraging habitats greatly, and this is considered a key shelter (Law and Dickman 1998; Joyal et al. 2001). driver of range-wide, regent parrot declines (Baker- While conservation management is increasingly Gabb and Hurley 2011). The regularity of movements focussed on promoting heterogeneous landscapes, a between habitat types by regent parrots represents an better understanding of the spatial properties of interesting case, because it contrasts with the more diverse mosaics which affect biota is needed to ensure commonly recognised effects of fragmentation on management actions improve conservation outcomes disrupting less regular dispersal patterns and geneflow (Lindenmayer et al. 2008; Fahrig et al. 2011). among populations (e.g. Harrisson et al. 2012). Con- Most advances in understanding the effects of serving regent parrots requires knowledge of how landscape spatial properties on biota have arisen through landscape properties (e.g. the extent, composition and studies of binary landscapes (i.e., simple patchworks of configuration of landscape elements; Bennett et al. ‘habitat’ and ‘non-habitat’; (Haila 2002; Fischer and 2006) and connectivity between landscape elements Lindenmayer 2006). Yet, approaches are needed to interact to affect species persistence. determine how species are affected by the composition In this study, we determined which properties of and configuration of landscape elements along a con- heterogeneous landscapes affect landscape suitability tinuum of habitat suitability, clarifying the individual for biota, using the regent parrot as a test case. and interacting effects of habitat extent and configura- Specifically, we examined the effects of three land- tion (Fischer and Lindenmayer 2006; Fahrig et al. 2011; scape characteristics on the occurrence of regent Tscharntke et al. 2012). These knowledge gaps impede parrot nesting locations. First, we tested whether on-ground conservation management because attempts nesting was more common in landscapes with a to improve the extent and connectivity of habitat may be greater extent of E. camaldulensis riverine forest, the ineffective if they ignore an organism’s requirements for core breeding habitat for regent parrots (Higgins different landscape elements (the composition of land- 1999). Second, because regent parrots are known to scape elements; Dunning et al. 1992b; Bennett et al. make regular forays away from nest locations (Higgins 2006). For example, Eurasian curlews Numenius arqu- 1999), we predicted that the composition of non- ata nest in moorland, but feed in adjacent farmland riverine landscape elements would affect regent parrot (Dallimer et al. 2012). Improvements to the quality or nesting. Third, recognizing that regent parrots use extent of moorland may have little effect on the multiple landscape elements, we predicted that land- movement, foraging success or, ultimately, population scape suitability would be reduced with lower con- persistence of the curlew if not considered in the context nectivity between key landscape elements. of the surrounding farmland. As movements between complementary habitats need to occur frequently, reduced connectivity between these habitats may affect Methods species occupancy (Dallimer et al. 2012). The regent parrot (eastern subspecies) Polytelis Vegetation and regent parrot data collation anthopeplus monarchoides is a threatened of inland south-eastern Australia that has suffered a Our study was conducted in fragmented agricultural significant population decline and range reduction over landscapes of south-eastern Australia, centred on the the past century (Higgins 1999; Garnett and Crowley intersection of the states of (NSW), 2000). This species depends on Eucalyptus camaldul- (VIC) and (SA; Fig. 1). ensis forest as primary breeding habitat (Higgins Native vegetation in the region varies according to soil 1999), nesting in old, hollow-bearing trees that are type and proximity to major waterways. Eucalyptus largely restricted to riparian areas (Keith 2004). camaldulensis forests occur along waterways, which 123 Landscape Ecol (2014) 29:1249–1259 1251

Fig. 1 The location of the study area in Australia (inset), and shows how landscape resistance layers were generated from each of the 39 study landscapes. Study landscapes were located vegetation types (here, Scenario 6 is shown, see Table 1) and along the major waterways in the region which support regent highlights each of the presence and absence patches for regent parrot breeding habitat (E. camaldulensis forest). The main parrot (RP) nesting and also large mallee blocks ([100 ha). The image shows the extent of treed vegetation (green) and open/ right image provides an example of a current map between the agricultural land (light grey). The enlargements above show an regent parrot presence patch and large mallee blocks generated example landscape. The left image highlight the site selection using Circuitscape (see ‘‘Methods’’ section). The amount of process, whereby all the E. camaldulensis forest within a 2 km current represents how organisms would flow across the radius (black circles) of regent parrots was selected as a landscape based on the overall resistance (from resistance presence patch (patches outlined in yellow), and absence surfaces) between patches. Resistance values can be generated (availability) patches were selected where the river intersected which represent the level of connectivity between patches. a quadrant (patches outlined in black). Also shown is each of the (Color figure online) different vegetation types in the landscape. The centre image have historically experienced regular flooding (Keith with a distinctive multi-stemmed form, and small 2004). Upper floodplain areas support Eucalyptus areas of dry-land woodlands. Agriculture in the area is largiflorens woodlands, often with a sparse layer of predominantly cereal cropping and grazing of low chenopods (Keith 2004). Farther from waterways, open shrublands, with irrigated horticulture (primarily aeolian soils support mallee shrublands, which form a citrus, almond and olive orchards) restricted to the low open canopy (6–12 m) of Eucalyptus sp. that grow deeper floodplain soils adjacent to the major rivers.

123 1252 Landscape Ecol (2014) 29:1249–1259

We combined four existing vegetation datasets to breeding. Sites represented all E. camaldulensis forest create a single spatial dataset of comparable vegeta- within a 2 km radius. The location of sites was tion types across the study region (Fig. 1): (1) the selected using two criteria. First, we selected sites ecological vegetation class dataset (VIC); (2) the centred on known regent parrot breeding sites; where vegetation formations dataset (NSW); (based on Keith breeding locations were closer than 2 km to each 2004); (3) the native vegetation cover dataset (SA); other, the central point of the cluster was used. Second, and (4) the mallee vegetation dataset from Haslem ‘absence’ sites were selected by dividing landscapes et al. (2010). Since vegetation classifications differed into 16 quadrants using 5 km spaced lines and between datasets, we reclassified the vegetation cat- situating sites where the river intersected a quadrant egories of each state into one of five broader vegeta- line and there were no regent parrot records within a tion categories: (1) mallee; (2) E. camaldulensis 2.5 km radius (half quadrant width; see outset land- forest; (3) E. largiflorens woodland; (4) dry-land scape in Fig. 1). woodlands dominated by Casuarina pauper, Callitris sp., or Alectryon oleifolius; and (5) open/agricultural Statistical analysis land. To delineate the extent of mallee, we used the dataset of Haslem et al. (2010) because it encom- For each study landscape, we measured the extent of the passed the broadest area and was the only dataset that five broad vegetation types. Three study landscapes distinguished mallee from other dry-land treed vege- were excluded from the analysis as outliers because they tation in NSW (see Appendix S1, Table A1 for each contained a far greater extent of specific vegetation vegetation classifications). types than the other landscapes. Two landscapes For regent parrots, we collated existing data from contained 85 and 160 % more area of E. camaldulensis six reports and two surveys commissioned by state forest than the next highest landscape and one landscape government agencies between 1997 and 2011, which contained 100 % more mallee than the next highest contained location records for 674 incidences of landscape, representing data points more extreme than regent parrot nesting across the region over 14 years. 29 the interquartile range. These landscapes were not With the exception of 59 records collected in 1997, all included in the analysis because doing so could result in other data was collected over a 7 year period between unstable models due to the excessively high influence on 2004 and 2011, however, the 1997 data was included model estimates of extreme data-points (Quinn and because regent parrots were known to be currently (as Keough 2007). A further five landscapes either did not of 2011) breeding in this location, but there had not have E. camaldulensis forest at points where quadrants been any more recent published surveys recording the intersected the river or did not contain any large mallee breeding locations (see Appendix S2, Table A2). We blocks ([100 ha). In one landscape, only two points developed a GIS layer of these data using ArcMap 10 supported E. camaldulensis forest where the river (Environmental Systems Resource Institute 2010). intersected a quadrant. Consequently, analysis was conducted on data from 92 sites in 31 landscapes. Site selection We tested for correlations among the percent cover of vegetation types using Spearman’s rank correlation. To examine the influence of landscape composition Eucalyptus largiflorens and dry-land woodlands were and connectivity on the occurrence of regent parrot both negatively correlated with open/agricultural veg- nesting, we selected 39 9 10 km radius study land- etation (RhoS =-0.61 and -0.65, respectively), but scapes. These landscapes encompassed the entire positively correlated with each other (RhoS = 0.22). known distribution of regent parrot breeding records Hence, the former two variables were combined on the along the major Murray-Darling Basin waterways that basis of their woodland structure to form one variable support nesting habitat for this species (E. camaldul- representing ‘semi-arid woodlands’. The extent of ensis forests) (Fig. 1). In each landscape, we selected semi-arid woodlands was strongly negatively corre- three sites (patches of E. camaldulensis woodland) lated with open/agricultural vegetation (RhoS = representing (a) presence patches, representing breed- -0.88) and negatively correlated with the extent of ing locations of regent parrots, and (b) ‘absence’ mallee (RhoS =-0.26). Consequently, of these vari- patches, where there were no records of Regent Parrot ables only the extent of semi-arid woodlands and the 123 Landscape Ecol (2014) 29:1249–1259 1253

Table 1 Six scenarios of varying resistances for vegetation types Vegetation type Scenario 1 (proximity) Scenario 2 Scenario 3 Scenario 4 Scenario 5 Scenario 6

Eucalyptus camaldulensis forest 1 1 1111 Mallee 1 1 1111 Semi-arid woodland 1 2 2123 Agricultural/open vegetation 1 3 2212 Higher values represent greater resistance to movement. A null model has equal resistance across all vegetation types, thus measured resistance between nodes is equivalent to proximity extent of mallee were included in models. Observed resulted in six different resistance surfaces represent- relationships with semi-arid woodland equate to the ing a spectrum of vegetation resistance scenarios inverse relationship with open/agricultural vegetation. (Table 1). For each of the different resistance scenar- We used Circuitscape (McRae and Shah 2011)to ios we calculated the total level of resistance between measure the level of connectivity between sites a site (regent parrot presence and absence sites) and (nesting vegetation in E. camaldulensis forest) and large mallee blocks in the landscape. large patches ([100 ha) of foraging habitat (mallee To compare the relative effect on breeding loca- shrubland; Burbidge 1985) and to ascertain which tions of the composition of the landscape, and of the vegetation types represented a greater resistance to connectivity between E. camaldulensis and mallee movement and thus reduced connectivity between vegetation, we used a two-step process. First, we these habitat types. Circuitscape is based on electronic examined the effects of each of the different resistance circuit theory and calculates the level of connectivity scenarios on regent parrot nesting locations using as the amount of ‘‘current’’ (density of organisms) that generalized linear mixed models (GLMMs). For each pass between nodes (or patches) in a landscape. This resistance scenario, we built a univariate binomial measure can be represented as the total ‘resistance’ to model using the lme4 package (Bates et al. 2012)inR. movement across the landscape between patches, The response variable was specified as presence or taking into account all possible pathways between absence of a regent parrot at a site and the predictor those patches and can be expressed as a landscape variable was the resistance between the site and mallee ‘resistance’ layer (Fig 1). Because Circuitscape is able vegetation. Additionally, we included a categorical to take in all-possible pathways, it is considered an variable representing each study landscape as a advance on traditional metrics such as least-cost path random effect to account for the multiple (3) sites analysis (McRae et al. 2008). Because our sites (breeding and availability locations) within each represented patches of vegetation, we defined nodes landscape. We then compared the explanatory power as ‘‘regions’’, rather than single point locations in the of each model (resistance scenarios) on the presence of landscape (McRae and Shah 2011). For each breeding regent parrot breeding locations using Akaike’s site and availability site, we obtained a measure of Information Criterion corrected for small sample sizes resistance between that site and large patches (AICc). ([100 ha) of foraging vegetation (mallee shrubland). For the second stage of our analysis, we built a new Because we wished to determine the relative set of GLMMs comparing the relative explanatory contribution of different vegetation types to connec- power of connectivity with the composition of the tivity between E. camaldulaensis forests and mallee landscape. In this case, we built 16 models represent- we defined a range of resistance surfaces (scenarios) ing all possible subsets of four predictor variables: (1) where different vegetation types represented different connectivity (resistance) between sites (E. camaldul- resistances to movement. To define the resistance ensis forest) and foraging locations (mallee[100 ha), surfaces, we held E. camaldulensis forest and mallee using the resistance scenario as being identified as best at the lowest resistance level and varied the resistance model from the first stage of the analysis; (2) the extent of the intervening landscape elements (semi-arid of E. camaldulensis forest; (3) the extent of mallee woodland and open/agricultural land). This process vegetation; and (4) the extent of semi-arid woodlands.

123 1254 Landscape Ecol (2014) 29:1249–1259

2 All models also included study landscape as a random (Rm = 0.46–0.62, coefficient ±2 SE does not overlap effect. We then compared models using an informa- with 0, Table 2). The model with the most support tion theoretic approach based AICc to assess the level ranked semi-arid woodland as the highest resistance of support for each model across all sixteen models of and agricultural/open land as moderate resistance all combinations of variables (Burnham and Anderson (Table 2). However, there was considerable uncer- 2002). For both stages of the analysis, we considered a tainty around the level of improvement in explanatory model to have substantially stronger support where the power of different scenarios of resistance across the DAICc [2 points higher than the next best model. landscape. Several models, including the simplest For all models, model fit was examined using the model representing proximity of E. camaldulensis 2 2 marginal (GLMM Rm ) and conditional R values forest to mallee, were all within 2 AICc points of the 2 (GLMM Rc ) described by Nakagawa and Schielzeth best model (Table 2). Only the scenario of open/ 2 agricultural land solely causing ‘resistance’ to move- (2013). GLMM Rmrepresents the amount of variabil- ity in the data explained by the fixed effects compo- ment had markedly less support than the alternatives nent of the model (i.e. the landscape composition and (Table 2). 2 The second-stage of modelling revealed that both connectivity measures) and GLMM Rc is the variabil- ity explained by the whole model including the fixed connectivity and landscape composition affected the and random effects (i.e., landscape). occurrence of regent parrot nesting locations. The Although regent parrots are relatively easy to detect relative occurrence of regent parrot nesting was in the field, there has been no systematic analysis of highest where sites were located in patches with a detectability-of-nesting for this species. And, although higher connectivity (low resistance) with large mallee the region has been extensively surveyed, we recog- blocks and in landscapes with less semi-arid wood- nise that our ‘absence’ patches may represent pseudo- lands and a larger proportion of E. camaldulensis absences representative of the broader ‘availability’ of forest (Table 3; Fig. 2). The model selection process nesting locations for regent parrots. Consequently, we incorporating the best connectivity model (ranking treat all analysis as use-availability models, rather than semi-arid woodland as the highest resistance and traditional presence-absence models, where the results agricultural/open land as moderate resistance) and the represent the occurrence of the organism ‘‘relative’’ to extent of the three key vegetation types (E. camaldul- the availability of sites across the region (Aarts et al. ensis forest, semi-arid woodland and mallee vegeta- 2012). Logistic regression is a robust method for tion) revealed strong support (DAICc [ 2) for a model developing use-availability models when large num- which included three variables: connectivity, extent of bers of availability points cannot be obtained (Aarts E. camaldulensis forest and extent of semi-arid et al. 2012), as is often the case with endangered woodland. This model had excellent explanatory 2 species and large landscape scale projects. power (marginal Rm ¼ 0:72), all three variables were considered important predictors of nesting location (coefficients ± 2 SE did not overlap 0, Table 3). Results

Fifteen of the 31 landscapes contained active regent Discussion parrot nesting areas. We found strong evidence that both the connectivity between E. camaldulensis Our research highlights how landscape heterogeneity forests and mallee vegetation and the composition of affects patterns of species’ distribution and habitat use vegetation types in the landscape affected the occur- (Tscharntke et al. 2012). While an important explan- rence of regent parrot breeding locations. atory variable of regent parrot occurrence was the total Connectivity between E. camaldulensis forest and extent of core breeding habitat (E. camaldulensis mallee clearly affected the occurrence of regent parrot forest), we found that the occurrence of the parrot was nesting, with the fixed effects component of all affected also by the extent of vegetation types not used connectivity models (representing the level of con- for breeding or foraging—a strong negative effect of nectivity) having good–very good explanatory power semi-arid woodlands. Furthermore, nesting was

123 Landscape Ecol (2014) 29:1249–1259 1255

Table 2 Univariate models explaining the effect of connectivity between E. camaldulensis forest and patches of mallee[100 ha on the occurrence of regent parrot nesting locations across 92 sites in 31 landscapes Resistance scenario Coefficient (SE) AIC DAIC 2 2 c c GLMM RðmÞ GLMM RðcÞ

Scenario 6 -6.2422 (2.0644) 88.27 0.62 0.79 Scenario 5 -6.1176 (2.0296) 88.33 0.06 0.61 0.78 Scenario 3 -5.8918 (1.9619) 89.20 0.94 0.57 0.78 Scenario 1 (proximity) -5.5974 (1.8693) 89.78 1.52 0.54 0.77 Scenario 2 -5.6558 (1.9017) 90.24 1.97 0.54 0.78 Scenario 4 -4.9873 (1.6788) 92.25 3.99 0.46 0.76 Shown are the standardized coefficients and standard errors for each of the six different scenarios of vegetation resistance described in the same table

Table 3 The six models which had the strongest levels of support (cumulative weights \0.95) to explain the effects of landscape connectivity and composition on the occurrence of regent parrot nesting Connectivity between Extent of semi-arid Extent of E. Extent of mallee AIC DAIC Cumulative w 2 2 c c i RðmÞ RðcÞ E. camaldulensis woodland camaldulensis forest and mallee forest

-5.079 (2.0720) -3.06 (1.5111) 1.933 (0.9583) 83.9 0 0.509 0.72 0.81 -5.294 (2.3484) -3.045 (1.5208) 1.941 (0.9630) -0.2176 (1.0854) 86.2 2.25 0.674 0.72 0.81 -4.828 (2.0322) -2.144 (1.2973) 87 3.06 0.784 0.63 0.78 -6.88 (2.1691) 1.359 (0.9350) 88.1 4.16 0.848 0.68 0.82 -6.242 (2.0644) 88.3 4.35 0.906 0.62 0.79 -5.087 (2.3212) -2.125 (1.3089) -0.2725 (1.1415) 89.2 5.24 0.943 0.63 0.78

Standardized coefficients and standard errors are shown. There is good support for the best model (DAICc [2), which has excellent explanatory power (marginal R2 = 0.72) highlighting the importance of both connectivity and composition of vegetation in the landscape. Connectivity was included in all of the models with strong levels of support concentrated in those landscapes with greater connec- demonstrate that regent parrots require access to tivity between E. camaldulensis forest and mallee (key mallee from their nesting locations, but the relative foraging habitat). While increasing the amount of influence of mosaic elements on connectivity versus preferred habitat is a cornerstone goal of many simply the proximity of mallee to breeding habitat was conservation and restoration programs (Fahrig 2003; uncertain. There was clear evidence that models that Radford et al. 2005; Crossman and Bryan 2006), for defined only agricultural/open vegetation as a barrier species such as the regent parrots, conservation to movement for regent parrots perform poorly. This managers need to consider the composition and highlights that fragmentation studies that focus solely structure of entire landscapes. on human defined barriers to connectivity may draw The level of connectivity between E. camaldulensis misleading conclusions and expert opinion which forests and mallee had the strongest influence on assumes greater resistance of agricultural land should regent parrot occurrence, underscoring the interacting be undertaken with caution. effects of fragmentation and different landscape The strong positive association between the regent elements on landscape suitability. Disrupted connec- parrot and E. camaldulensis forest and negative tivity between landscape elements used regularly (e.g. relationship with increasing resistance (decreasing daily) will impact negatively on local population connectivity) with mallee is consistent with observa- persistence—an under-appreciated effect of landscape tions that the parrot requires E. camaldulensis forests fragmentation compared to the more widely recogni- for breeding and mallee for feeding and roosting sed impacts on species dispersal and metapopulation (Burbidge 1985). Also, regent parrots appear to make dynamics (Fahrig 2003; Henle et al. 2004). Our results regular foraging forays away from their nest into 123 1256 Landscape Ecol (2014) 29:1249–1259

Fig. 2 The influence of 1.00 1.00 a resistance between E. a b camaldulensis forest and 0.80 0.80 mallee (scenario 6) (connectivity), b the extent 0.60 0.60 of semi-arid woodland (ha), and c the extent of E. 0.40 0.40 camaldulensis forests (ha) on the relative probability of 0.20 0.20 occurrence of regent parrot probability of occurrence nesting. Predicted probability of occurrence 0.00 0.00 probabilities of occurrence 0 0.25 0.5 0.75 1 0 5000 10000 15000 20000 (solid line) ±SE (dotted 1.00 lines) are shown from the c best model 0.80

0.60

0.40

0.20 probability of occurrence

0.00 0 1500 3000 4500 6000 patches of mallee vegetation, possibly multiple trips compared to other vegetation in the region (Gates per day during breeding (Higgins 1999). These daily 1996). In mallee, occurrence of tree hollows is movements contrast with other species that require strongly related to tree age and time since last fire, different landscape elements for different life stages with branches able to support hollows large enough for (e.g. amphibians and invertebrates; Forman 1995), or regent parrots only found in very old trees (Clarke use different elements in different seasons (Heard et al. 2010; Haslem et al. 2012), which are increasingly et al. 2004; Both et al. 2010). While the occurrence of rare (Clarke et al. 2010). Some semi-arid woodlands— the regent parrot was positively related to E. camal- particularly those dominated by E. largiflorens— dulensis forest, it was also negatively related to the support hollows and hollow-dependant fauna, but in extent of semi-arid woodlands. That is, large stands of lower densities than in E. camaldulensis forest (Gates semi-arid woodlands appeared to disrupt the comple- 1996; Bennett et al. 1989). Regent parrots are only mentary relationship between the vegetation used for rarely observed using semi-arid woodlands for forag- breeding and foraging. This negative effect has not ing and to the best of our knowledge have never been been identified previously for this species, and found to breed in these vegetation types and our results researchers and conservation managers should be support this observation, with no nesting occurrence aware of the possibility that some landscape elements located in these vegetation types. within heterogeneous mosaics may disrupt or diminish The use of mallee by regent parrots for foraging is landscape suitability for some species by disrupting likely driven by the diversity of food resources that the complementarity between multiple, preferred this vegetation type supports. The local topographical elements. variability and varied soil types created by dune-swale Understanding the mechanisms underlying regent systems within mallee support a greater number of parrot responses to landscape composition is chal- woody species than in semi-arid woodlands or E. lenging. The use of E. camaldulensis forests likely camaldulensis forests (Land Conservation Council relates to the dependence of regent parrots on large 1987). Consequently, mallee is likely to support a tree hollows for nesting (Higgins 1999), with hollow greater diversity and abundance of food resources that availability being relatively high in these forests are regularly available for regent parrots. Furthermore, 123 Landscape Ecol (2014) 29:1249–1259 1257 most semi-arid woodlands in the region have been reduce the extent of E. camaldulensis forests even degraded through grazing, and, since many understo- further (Mac Nally et al. 2011). Reduced flooding rey in these communities are susceptible to frequency may result also in the encroachment of E. intensive grazing pressure (Keith 2004), semi-arid largiflorens woodlands into areas previously covered by woodlands are likely to have a reduced availability of E. camaldulensis forests. food resources for regent parrots and other fauna. Our study used a range of data sources, and our results highlight the merits of amalgamating large data Conservation and management implications sources to study rare and threatened species. Never- theless, future studies should aim to build more Landscape restoration through revegetation is a com- detailed knowledge of site occupancy and use stan- monly employed conservation measure and is a dardized survey protocols to allow analysis of abun- potentially important management action for regent dances of species. Here, we only consider native parrots (Baker-Gabb and Hurley 2011). Yet, our study vegetation covers. Landscapes (including our study demonstrates that, for some species, revegetation region) are increasingly characterised by a diverse mix programs will need to focus on the restoration of of natural and anthropogenic land covers. Agricultural multiple habitat types and consider the juxtaposition landscapes are particularly diverse and dynamic, of these types in the landscape. For regent parrots, regularly experiencing major and rapid changes in revegetation should occur in landscapes that can land cover (Watson et al. 2014). It is likely that potentially support a greater extent of E. camaldul- different land covers represent different levels of ensis forest with mallee highly connected and in close suitability for use and permeability for the regent proximity, and without extensive areas of semi-arid parrot. For example, almond crops have expanded woodland, particularly where these woodlands frag- tenfold in the last 10 years in our study region (to ment the core habitat types. Landscape revegetation [20,000 ha), and regent parrots have been observed projects aimed at restoring semi-arid woodlands, regularly feeding and moving through these crops although critical for many other species in the region, (Luck et al. 2014). Additionally, we focused only on are unlikely to result in conservation gains for regent connectivity with large mallee blocks ([100 ha). parrots. In general, landscape restoration projects that Future research should investigate the relative impor- consider the complementarity and connectivity among tance of different sized mallee blocks and how this is multiple landscape elements may have greater bene- affected by their surrounding land covers (e.g. crop- fits to species conservation than those which focus on land v almond orchards). Furthermore, researchers one-type of habitat or ignore interactions among should aim to understand the influence of changes to habitats. all types of vegetation and land-uses (i.e. including Throughout the world, climate shifts are predicted to agricultural land-covers) on landscape connectivity, alter vegetation associations (Cramer et al. 2001). For and the extent of available resources for species that require multiple vegetation types in close occupying heterogeneous landscapes. proximity for survival, these climate shifts could have severe negative impacts. For example, in our research Acknowledgments We thank Kevin Smith, Rick Webster, area, E. camaldulensis forest grows on foodplains that David Leslie, Chris Belcher, Alex Holmes, Peter Robertson and Victor Hurley whose field research and reports provided data for have experienced regular inundation by flood waters, the project. We acknowledge the regent parrot recovery team for while semi-arid E. largiflorens woodlands grow on support of the project. This research was funded through the New floodplains that are less regularly inundated (Keith South Wales Office of Environment and Heritage (OEH) and an 2004). Regulation of rivers and more frequent and Australian Research Council linkage Grant (LP0883952). We particularly thank Peter Ewin and Damon Oliver from the NSW severe droughts, driven in part by climate change (van OEH for supporting this research. Dijk et al. 2013), has resulted in less frequent floods and extensive degradation of E. camaldulensis forests (Mac References Nally et al. 2011). The changing climate in southern Australia could have a twofold negative effect on regent Aarts G, Fieberg J, Matthiopoulos J (2012) Comparative inter- parrot populations, as the prolonged dry-periods are pretation of count, presence–absence and point methods for likely to result in greater water stress to floodplains and species distribution models. Methods Ecol Evol 3(1):177–187 123 1258 Landscape Ecol (2014) 29:1249–1259

Allouche O, Kalyuzhny M, Moreno-Rueda G, Pizarro M, Fischer J, Lindenmayer DB (2006) Beyond fragmentation: the Kadmon R (2012) Area—heterogeneity tradeoff and the continuum model for fauna research and conservation in diversity of ecological communities. Proc Natl Acad Sci human-modified landscapes. Oikos 112(2):473–480 109(43):17495–17500 Forman RTT (1995) Land mosaics. The ecology of landscapes Baker-Gabb D, Hurley V (2011) National recovery plan for the and regions. Cambridge University Press, Cambridge regent parrot (eastern subspecies) Polytelis anthopeplus Forman RTT, Godron M (1986) Landscape ecology. Wiley, monarchoides. Department of Sustainability and Environ- New York ment, Melbourne Garnett ST, Crowley GM (2000) Regent parrot (eastern). The Bates D, Maechler M, Bolker B (2012) lme4: linear mixed- action plan for Australian . Environment Australia, effects models using S4 classes. R package version Canberra, pp 314–316 0.999999-0. http://CRAN.R-project.org/package=lme4. Acces- Gates JA (1996) The use of hollows by birds and mammals in sed 3 April 2013 dead river red gum and black box trees at Disher Creek, Bennett AF, Lumsden LF, Menkhorst PW (1989) Mammals of National Park, South Australia. S Aust the mallee region of south-eastern Australia. In: Noble J, Ornithol 32:65–75 Bradstock R (eds) Mediterranean landscapes in Australia, Haila Y (2002) A conceptiual genealogy of fragmentation mallee ecosystems and their management. CSIRO, Can- research: from island biogeography to landscape ecology. berra, pp 191–220 Ecol Appl 12(2):321–334 Bennett AF, Radford JQ, Haslem A (2006) Properties of land Harrisson K, Pavlova A, Amos JN et al (2012) Fine-scale effects mosaics: implications for nature conservation in agricul- of habitat loss and fragmentation despite large-scale gene tural environments. Biol Conserv 133(2):250–264 flow for some regionally declining woodland bird species. Both C, Van Turnhout CAM, Bijlsma RG, Siepel H, Van Landscape Ecol 27(6):813–827 Strien AJ, Foppen RPB (2010) Avian population conse- Haslem A, Callister KE, Avitabile SC et al (2010) A framework quences of climate change are most severe for long-dis- for mapping vegetation over broad spatial extents: A tance migrants in seasonal habitats. Proc R Soc B technique to aid land management across jurisdictional 277(1685):1259–1266 boundaries. Landsc Urban Plan 97(4):296–305 Burbidge A (1985) The regent parrot: a report on the breeding Haslem A, Avitabile SC, Taylor RS et al (2012) Time-since-fire distribution and habitat requirements along the Murray and inter-fire interval influence hollow availability for River in south-eastern Australia. Australian National Parks fauna in a fire-prone system. Biol Conserv 152:212–221 and Wildlife Service, Canberra Heard GW, Black D, Robertson P (2004) Habitat use by the Burnham KP, Anderson DR (2002) Model selection and mul- inland carpet python (Morelia spilota metcalfei: Pythoni- timodel inference. A practical information-theoretic dae): Seasonal relationships with habitat structure and prey approach, 2nd edn. Springer, New York distribution in a rural landscape. Austral Ecol 29(4): Clarke MF, Avitabile SC, Brown L et al (2010) Ageing mallee 446–460 eucalypt vegetation after fire: insights for successional Henle K, Davies KF, Kleyer M, Margules C, Settele J (2004) trajectories in semi-arid mallee ecosystems. Aust J Bot Predictors of species sensitivity to fragmentation. Biodi- 58(5):363–372 vers Conserv 13(1):207–251 Cramer W, Bondeau A, Woodward FI et al (2001) Global Higgins PJ (1999) Polytelis anthopeplus regent parrot. Hand- response of terrestrial ecosystem structure and function to book of Australian, New Zealand and Antarctic birds: CO2 and climate change: results from six dynamic global parrots to dollarbird, vol 4. Oxford University Press, vegetation models. Glob Change Biol 7(4):357–373 Melbourne, pp 295–306 Crossman ND, Bryan BA (2006) Systematic landscape resto- Joyal LA, McCollough M, Hunter ML (2001) Landscape ecol- ration using integer programming. Biol Conserv ogy approaches to wetland species conservation: a case 128(3):369–383 study of two turtle species in Southern Maine. Conserv Biol Dallimer M, Skinner AMJ, Davies ZG, Armsworth PR, Gaston 15(6):1755–1762 KJ (2012) Multiple habitat associations: the role of offsite Keith DA (2004) Ocean shores to desert dunes: the native habitat in determining onsite avian density and species vegetation of New South Wales and the ACT/David Keith. richness. Ecography 35(2):134–145 Dept. of Environment and Conservation NSW, Hurstville Dunning JB, Danielson BJ, Pulliam HR (1992a) Ecological Land Conservation Council (1987) Vegetation. Report on the processes that affect populations in complex landscapes. Mallee area review. Land Conservation Council, Mel- Oikos 65(1):169–175 bourne, pp 104–141 Dunning JB, Danielson BJ, Pulliam HR (1992b) Ecological Law BS, Dickman CR (1998) The use of habitat mosaics by processes that affect populations in complex landscapes. terrestrial vertebrate fauna: implications for conservation Oikos 65:169–175 and management. Biodivers Conserv 7:323–333 Environmental Systems Resource Institute (2010) ArcMap 10.0. Lindenmayer D, Hobbs RJ, Montague-Drake R et al (2008) A ESRI Inc., Redlands checklist for ecological management of landscapes for Fahrig L (2003) Effects of habitat fragmentation on biodiversity. conservation. Ecol Lett 11:78–91 Annu Rev Ecol Evol Syst 34:487–515 Luck GW, Spooner PG, Watson DM, Watson SJ, Saunders ME Fahrig L, Baudry J, Brotons L et al (2011) Functional landscape (2014) Interactions between almond plantations and native heterogeneity and biodiversity in agricultural ecosystems: lessons learned from north-western Victoria. landscapes. Ecol Lett 14(2):101–112 Ecol Manag Restor 15(1):4–15

123 Landscape Ecol (2014) 29:1249–1259 1259

Mac Nally R, Cunningham SC, Baker PJ, Horner GJ, Thomson Tews J, Brose U, Grimm V et al (2004) Animal species diversity JR (2011) Dynamics of Murray–Darling floodplain forests driven by habitat heterogeneity/diversity: the importance under multiple stressors: the past, present, and future of an of keystone structures. J Biogeogr 31(1):79–92 Australian icon. Water Resour Res 47(12):W00G05 Tscharntke T, Tylianakis JM, Rand TA et al (2012) Landscape McRae BH, Shah VB (2011) Circuitscape user guide (Online). moderation of biodiversity patterns and processes—eight The University of California, Santa Barbara. http://www. hypotheses. Biol Rev 87(3):661–685 circuitscape.org. Accessed 3 April 2013 van Dijk AIJM, Beck HE, Crosbie RS et al (2013) The mil- McRae BH, Dickson BG, Keitt TH, Shah VB (2008) Using lennium drought in southeast Australia (2001–2009): nat- circuit theory to model connectivity in ecology, evolution ural and human causes and implications for water and conservation. Ecology 89(10):2712–2724 resources, ecosystems, economy, and society. Water Re- Nakagawa S, Schielzeth H (2013) A general and simple method sour Res 49(2):1040-1057 for obtaining R2 from generalized linear mixed-effects Watson SJ, Luck GW, Spooner PG, Watson DM (2014) Land- models. Methods Ecol Evol 4(2):133–142 use change: incorporating the frequency, sequence, time Quinn GP, Keough MJ (2007) Experimental design and data span, and magnitude of changes into ecological research. analysis for biologists. Cambridge University Press, Front Ecol Environ 12(4):241–249 Cambridge Radford JQ, Bennett AF, Cheers GJ (2005) Landscape-level thresholds of habitat cover for woodland-dependent birds. Biol Conserv 124(3):317–337

123