The discipline of landscape ecology has developed over the past three decades. It integrates studies of human impacts on biodiversity with studies of natural population dynamics and variation in diversity across different spatial and temporal scales. Reviews and text books describing the discipline and methods are numerous (e.g., Brandt and Agger 1984; Naveh and Lieberman 1984; Forman and Godron 1986; Skånes 1997; Gergel and Turner 2002; Collinge 2009). The expansion of landscape ecology has been enabled by technological advances in areas such as: geographic information systems, remote sensing, global positioning systems, internet, and
computer technology. Together these have made it possible to access, manipulate and model increasingly high-resolution data for large geographical regions to determine the landscape-scale impact of humans on gene flow, distribution and abundance of species, communities and ecosystems.
Among the many concepts and models within the discipline of landscape ecology is Island Biogeography Theory (MacArthur and Wilson 1967), which relates the size of islands and their isolation from other land areas to species colonization and extinction rates. This model has been generalised to explain the effects of landscape
Chapter 1 – Introduction fragmentation on meta-population dynamics and biodiversity (Levins 1969; Saunders et al. 1991, 1999). In landscape ecology, ‘habitat fragmentation’ is the term often used to describe the process anthropogenic landscape modification such as vegetation clearance, which results in the development of non-contiguous remnants of native habitat patches. The term 'habitat' is used to describe the environment suitable for use by a particular species (Lindenmayer and Fischer 2006). The term 'matrix' is used to describe the land cover type with the greatest area within the landscape, which in most fragmentation studies is usually represented by the most disturbed land cover type such as agricultural land or urban areas (Lindenmayer and Fischer 2006). In Island biogeography theory, a patch, like an island, is defined as an area of native land cover, surrounded by a matrix of land or water. Within many studies the landscape is classified and mapped into two types of land cover: habitat and matrix. For such situations, island biogeography theory suggests that patch size, habitat heterogeneity within the patch, and patch isolation may be strong influences on local species richness (MacArthur and Wilson 1967; Saunders et al. 1991, 1999). Hence habitat fragmentation and configuration patterns caused by vegetation clearance or other anthropogenic disturbance may influence and change individual species distributions (Noss and Harris 1986; Noss 1987).
The limitations of Island Biogeography Theory for ecological predicting biodiversity responses are now well documented (Gascon and Lovejoy 1998; Lindenmayer and Fischer 2006; Laurance 2008; Le Roux et al. 2015). For example the edges of habitat areas juxtaposed to cleared land are often subject to a myriad of processes and influences collectively referred to as edge effects (Gascon and Lovejoy 1998). These include the greater probability of weed invasion from agricultural or urban land, increased browsing pressure, predation from introduced animals (e.g. cats), nutrient enrichment from fertiliser drift, pollution, exposure to more extreme micro-climatic variation etc (Saunders et al. 1991). The variability of patch habitat quality, including structure, and the potential variation in species use of, or ability to move through, the matrix may impact more on species mortality and colonization rates than patch size and isolation in some circumstances (Lindenmayer and Fischer 2006; Laurance 2008; Le Roux et al. 2015). In particular the qualities of the surrounding landscapes,
Chapter 1 – Introduction including the abundance of particular habitat types and heterogeneity may all
influence the probability of a species occurrence or abundance at the site.
The term 'landscape context' (LC) is defined for the purposes of this thesis as the vegetation or land area surrounding a site, including both habitat and matrix in the surrounding area (sensu Noss and Harris 1986). The realisation that LC may influence species distribution has resulted in an explosion of studies in landscape ecology (Henle et al. 2004), but mostly in fragmented or relictual landscapes (sensu McIntyre and Hobbs 1999) dominated by cleared land. These have investigated the effects of both reduced habitat area and fragmentation associated with land clearance
(McGarigal and Cushman 2002; Fahrig 2003; Collinge 2009).
Habitat fragmentation is associated with several processes including reduction in total habitat, decreasing patch size, increasing isolation of patches and an increase in the ratio of the patch that is subject to edge effects (Saunders et al. 1991; McIntyre and Hobbs 1999; Saunders et al. 1999). Hobbs and McIntyre (1999) developed a model which describes commonly observed steps and processes in converting natural ecosystems (including habitat) into areas of anthropogenic land-use such as agricultural or urban land. It distinguishes thresholds in landscape conversion commencing from the ‘intact’ landscape in which more than 90% of the area is occupied by largely unmodified native vegetation. At the other end of the scale are ‘fragmented’ landscapes, distinguished as those in which there is less than 60% native vegetation cover, at least some of which is still unmodified. In contrast the few
remaining native vegetation patches in ‘relictual’ landscapes are usually highly modified. Importantly Hobbs and McIntyre (1999) distinguish an intermediate ‘variegated’ landscape stage in the evolving conversion of landscapes. In this stage native vegetation occupy more than 60% of the landscape but vary in their quality as a consequence of modification processes associated with surrounding land use change. Such modifying processes may include silviculture and fire regimes, both known to impact on biodiversity. In this thesis, the term landscape fragmentation is also applied to disturbance processes that result in the conversion of mature forest to regrowth forest.
Chapter 1 – Introduction Given the strong association of fragmentation processes with multiple processes including habitat loss, isolation, habitat change, and edge effects, it is inevitably challenging for scientists to distinguish between the which of these processes is responsible for observed biological responses. A large number of metrics have been devised to measure the various properties of habitat patches and landscapes that may impact on biodiversity, and the software package FRAGSTATS produced to assist with generating these (McGarigal and Marks 1995; McGarigal et al. 2002). However many landscape characteristics, and the metrics describing them are strongly auto correlated potentially confounding the results of many studies (McGarigal and Cushman 2002; Smith et al. 2009; Thornton et al. 2011; Mairota et al. 2015).
Distinguishing between the separate effects of the various processes that accompany landscape modification and fragmentation may be critical to mitigating against these impacts (Lindenmayer and Fischer 2006). Fahrig (2003) reviewed 100 fragmentation studies and found that most were patch-scale rather than landscape scale studies and most designs were unable to distinguish patch size effects from the other effects of fragmentation. Because many studies fail to distinguish the effects of habitat loss from other processes of fragmentation, and because the term fragmentation is often applied to all the associated processes, Lindenmayer and Fischer 2006 have suggested the use of the term 'habitat sub-division' to distinguish the process of separating a single large habitat area into several smaller areas.
The landscape fragmentation literature, is generally applied to the study of impacts on biodiversity associated with anthropogenic landscape modifications, usually
landscapes which are fragmented by clearance or conversion to plantations
(Lindenmayer and Fischer 2006). However, it also encompasses studies in regions of native vegetation where anthropogenic modification processes have led to the loss or sub-division of a particular habitat types such as late-stage or oldgrowth forest types, as a consequence of timber harvesting, or anthropogenic alterations to fire regimes (Lindenmayer and Fischer 2006). Modifications to fire and grazing regimes are usually associated with or precede intensification of land use and land clearance and impact on species dynamics and populations (Hobbs 1987; Syphard et al. 2007). Spatial and temporal variation in the natural environment, including climate and
Chapter 1 – Introduction (Lindenmayer and Fischer 2006). This patterning has had a profound influence on the evolution of species and is likely to contribute significantly to explaining natural species distribution and richness patterns (e.g. Haig et al. 2000; Dullinger et al. 2011; Frey et al. 2012).
Distinguishing the effects of natural environmental variation including natural fragmentation patterns from those of anthropogenic habitat fragmentation on species responses provides another challenge for empirical studies that should also not be overlooked (Lindenmayer and Fischer 2006). Landscape fragmentation is not random but often targets particular habitats and environments leading to a confounding between natural environmental variation and landscape metrics.
In the landscape fragmentation literature there has been an increasing tendency to view landscapes as a mosaic of different types of patches based on the approach of Forman (1995) or as a continuum of habitat suitability (Manning et al. 2004; Fischer et al. 2004; Lindenmayer and Fischer 2006; McGarigal et al. 2009). Patches are often defined more specifically as a relatively homogenous, non-linear area, of habitat (Lindenmayer and Fischer 2006). In such studies different habitat types may be more narrowly defined in terms of a specific characteristic, for example, dominance by a particular life form, age-class or species that make it suitable for a particular focal species or species group. The patch is often distinguished from corridors. Corridors being linear extents of habitat that are connect two or more patches (Lindenmayer and Fischer 2006). These may assist some species move between patches in the landscape but due to their narrowness, corridors are often unsuited to more continuous
occupation by the focal species and do not provide core habitat.
The landscape fragmentation literature is immense and still growing (Lindenmayer and Fischer 2006). Several reviews and syntheses of the effects of landscape
fragmentation have been published (e.g. Saunders et al. 1991; Mazerolle and Villard 1999; Debinski and Holt 2000; McGarigal and Cushman 2002; Fahrig 2003; Hobbs and Yates 2003; Aguilar et al. 2006; Bennett et al. 2006; Lindenmayer and Fischer 2006; Collinge 2009; Swift and Hannon 2010; Thornton et al. 2011; Humphrey et al. 2015). However there has been a bias towards vertebrate animals, particularly birds, more studies targeting threatened species compared with more common taxa, and few
Chapter 1 – Introduction studies being multidisciplinary (Debinski and Holt 2000; McGarigal and Cushman 2002; Fahrig 2003; Collinge 2009). This bias is not restricted to the landscape fragmentation literature but has been reported more generally within the published biological literature (Fazey et al. 2005).
Debinski and Holt (2000) reviewed 20 studies and concluded that although there were inconsistencies between results, most supported the theory that “movement and species richness are positively affected by corridors and connectivity, respectively” (Debinski and Holt 2000, p 342). Likewise, Bennett et al. (2006) reviewed empirical fauna studies that investigated the nature conservation implications of landscape fragmentation for within agricultural landscapes. They concluded that the properties of agricultural land mosaics (extent, composition and configuration) all strongly influence the occurrence of fauna, but they observed that the responses varied greatly. An example of a result from a landscape mosaic study in Victoria, Australia,
demonstrated that of the landscape attributes examined tree cover accounted for 55% of the variation woodland bird species richness among landscapes (Radford et al. 2005). Another Australian bird study, not included in the review by Bennett et al. (2006), provided evidence that 50 percent of the bird species studied within riparian habitats had abundance responses related only to variation in LC while 80%
responded to the combined effects of local site condition and LC (Martin et al. 2006). A review by Swift and Hannon (2010) of 17 fauna and one fungi study similarly demonstrated LC effects on species responses. However, this review also
demonstrated that most species tend to show a non-linear response to landscape fragmentation, with most responses occurring only after a critical threshold in habitat loss had occurred. The investigation of critical thresholds at which species respond to both habitat loss or fragmentation is therefore another important aspect of the
fragmentation literature (e.g. Villard and Metzger 2014).
Swift and Hannon (2010) also noted that a temporal lag often occurred between landscape change and species responses. Factors potentially influencing the response of species to habitat sub-division include the total amount of habitat available and the quality or 'resistance' of the matrix for species movement and survival (Fahrig 2001).
Chapter 1 – Introduction Such non-linear responses and temporal lags add to the difficulty in detecting LC effects on biodiversity.
A review of results for 954 fauna species from 122 focal patch studies was undertaken by Thornton et al. (2011) from which a diverse range of taxa were also found to respond to LC, patch-size and within patch heterogeneity. Among fauna groups they observed that mammals were particularly sensitive to their LC. They also noted that the probability of detecting landscape responses was influenced by study methods, choice of response variable, sample size and choice of landscape metric (Thornton et al. 2011). They noted that few studies tested for spatial autocorrelation within their data sets, or correlations among the predictor variables. They noted that this oversight could be leading to erroneous conclusions about the nature of species relationships with their environment, a concern raised by other researchers (Betts et al. 2006; Cushman et al. 2008; Smith et al. 2009).
The development metrics, and the problems associated with the scale of landscape responses have also been the focus of several studies (Neel et al. 2004; Cushman and McGarigal 2008; Cushman et al. 2008). Although Thornton et al. (2011) found no evidence among the studies they reviewed of a sensitivity in the response of fauna to multiple buffer scales, they nevertheless recommended the use of multiple buffers for the generation of LC variables citing the results demonstrating sensitivity to landscape radius from a simulation study by Moilanen and Nieminen (2002). Other studies have also suggested species vary in the scale at which they are sensitive to landscape effects (Chust et al. 2004). For example, Steffan-Dewenter et al. (2002) found that the abundances of solitary wild bee species were associated with differences in the
proportion of native vegetation present in the surrounding areas measured up to distances of 750 m radius; but honey bees were associated only with landscape differences measured at larger radial landscape distances. Differences among species response to scale of landscape patterns increase the difficulty of detecting LC
influence.
Many of the results from fragmentation studies undertaken in southeastern Australia prior to 2005 have been reported within the synthesis provided by Lindenmayer and Fischer (2006) on the subject of landscape fragmentation and habitat change. Among
Chapter 1 – Introduction these are some of the results coming out of the Tumut long term experimental project studying eucalypt forest remnants within a matrix of pine plantations, at Wog Wog, New South Wales (Margules 1992; Margules 1996). This area was fragmented experimentally by the conversion of eucalypt forest to Pinus radiata plantations. Lindenmayer et al. (1999) reported no association found between abundance of individual mammal species and LC or remnant patch size. In contrast, Lindenmayer et al. (2000c) found mammal assemblages were impoverished in remnants compared with areas of continuous eucalypt forest and that species richness increased with remnant patch size. Even small remnants were able to be occupied by a greater number of vertebrates species than anticipated, while some native species were located with the radiata pine stands, although their presence in the matrix may have been associated with proximity to remnant eucalypt stands (Lindenmayer and Fischer 2006).
Of the vascular plant studies undertaken at the Tumut long term experimental site, the first finding reported was that common plants constitute more of the flora in small remnant native vegetation patches (0.25 ha) compared with either large patches (3 ha) or intact forest, a difference attributed to the greater environmental change in the smallest remnants (Morgan and Farmilo 2012). Small remnants (0.25 ha) also contain a higher species densities when measured at small spatial scales (1 m2) and medium spatial scales (16 m2) compared with intact forests but not at the largest spatial scale of sampling (144 m2). However, species densities did not differ between larger patches and intact forest for any sample size (Farmilo et al. 2014). The differences observed in small patches were attributed to the greater influence of pine plantation to microclimate, and soil moisture characteristics (Farmilo et al. 2013; Farmilo et al. 2014).
Fragmentation studies were established earlier in Amazonia than Australia and have revealed much more about the landscape ecology of vascular plant species for tropical forest ecosystems. Results there provide evidence that forest fragmentation by
clearance influences tree species composition and forest structure at the edge of tropical forest remnants, causing mortality in large canopy and emergent trees and associated losses in above-ground biomass (Laurance 1991; Ferreira and Laurance
Chapter 1 – Introduction 1997; Laurance et al. 1998b; Laurance et al. 1998a; Mesquita et al. 1999; Laurance et al. 2000; Laurance et al. 2003; Laurance et al. 2006).
Although long term experimental fragmentation studies and simulation studies enable greater capacity to distinguish between the various processes of landscape
fragmentation, a growing number of empirical studies have demonstrated that plant responses often lag well behind landscape change (Tilman et al. 1994; Vellend et al. 2006; Cousins 2009; Koyanagi et al. 2012; Rigueira et al. 2013; Bagaria et al. 2015). The lag time, which is referred to within Island Biogeography Theory as the
relaxation time, contributes to what is termed an 'extinction debt' or 'colonization credit' within the biodiversity of patches that have not reached equilibrium with their current landscapes. Evidence that some plant species patterns are better explained by historical landscape patterns than current patterns have been provided by studies in both European grasslands (Lindborg and Eriksson 2004; Reitalu et al. 2009; Cousins 2009; Piqueray et al. 2011; Koyanagi et al. 2012; Bagaria et al. 2015) and temperate forest and woodlands in Europe and northern America (Östlund et al. 1997; Gerhardt and Foster 2002; Verheyen et al. 2003; Graae et al. 2004; Vellend et al. 2006;
Verheyen et al. 2006; Kimberley et al. 2015). These studies suggest that variation in life history, reproductive and dispersal traits contribute to species vulnerability to extinction and therefore also the rate at which plant species distribution, abundance and associated plant species richness patterns respond to landscape change. The observation that many life-history traits co-vary across species has led to the concept of the fast-slow continuum hypothesis (Franco and Silvertown 1996). Species with slow metapopulation dynamics, defined as those with both low rates of colonization and extinction, are likely to be slow to respond to such landscape changes due to characteristics such as long life spans or the capacity to reproduce vegetatively (Vellend et al. 2006). Some slow forest herbs have been identified as having particularly slow migration rates with a high risk of extinction unless habitat connectivity is maintained (Matlack and Monde 2004).
Despite the complications of a lag in plant responses to landscape change, more than half of the European plant studies incorporated within a review of woodland
fragmentation found evidence of positive associations in richness or plant occurrence with woodland patch size, proximity to other patches, ecological continuity, and patch
Chapter 1 – Introduction quality (Humphrey et al. 2015). One study in Italy found a linear increase in perennial herb richness up to a patch size of 35-40 ha (Digiovinazzo et al. 2010), whereas a study of forest plants in Belgium found an association with forest age but no
association with current patch size (Honnay et al. 1999). When the distributions of 59 individual species were analysed in Belgium woodlands 34 were found to be
associated with three or more landscape metrics including patch size (Jacquemyn et al. 2003). Vellend et al. (2006) observed that Belgium woodlands, which have
become fragmented only since 1775, had higher than expected richness levels of 'slow species' based on models of richness to patch area developed for long fragmented landscapes in England.
Aguilar et al. (2006) reviewed the effects of landscape fragmentation on reproductive success in plants (fruit or seed production) for 89 animal pollinated plant species from 53 studies and found that on average fragmented landscapes were associated with