*Correspondence: Joern Fischer, Centre for Resource and Environmental Studies, The Australian National University, Canberra ACT 0200, Australia. E-mail: email@example.com
Landscape modification and habitat fragmentation are key drivers of global species loss. Their effects may be understood by focusing on: (1) individual species and the processes threatening them, and (2) human-perceived landscape patterns and their correlation with species and assemblages. Individual species may decline as a result of interacting exogenous and endogenous threats, including habitat loss, habitat degradation, habitat isolation, changes in the biology, behaviour, and interactions of species, as well as additional, stochastic threats. Human-perceived landscape patterns that are frequently correlated with species assemblages include the amount and structure of native vegetation, the prevalence of anthropogenic edges, the degree of landscape connectivity, and the structure and heterogeneity of modified areas. Extinction cascades are particularly likely to occur in landscapes with low native vegetation cover, low landscape connectivity, degraded native vegetation and intensive land use in modified areas, especially if keystone species or entire functional groups of species are lost. This review (1) demonstrates that species-oriented and pattern-oriented approaches to understanding the ecology of modified landscapes are highly complementary, (2) clarifies the links between a wide range of interconnected themes, and (3) provides clear and consistent terminology. Tangible research and management priorities are outlined that are likely to benefit the conservation of native species in modified landscapes around the world.
A range of reviews have investigated fragmentation-related topics to date. Most of these have dealt with specific sub-components of the enormous body of literature on landscape modification and habitat fragmentation, such as the amount of native vegetation cover in relation to birds and mammals (Andrén, 1994), the relative effects of habitat loss and habitat sub-division (Fahrig, 2003), experimental approaches to studying ‘fragmentation’ (Debinski & Holt, 2000; McGarigal & Cushman, 2002), the history of fragmentation research (Haila, 2002), metapopulation dynamics (Hanski 1998), or edge effects (Ries et al., 2004; Harper et al., 2005). To date, only the review by Saunders et al. (1991) — which is now 15 years old — has covered a breadth of inter-related topics that are relevant to conservation biologists working in modified landscapes. The separate treatment of different sub-components of landscape change has led to a range of parallel research paradigms. Much research currently approaches modified landscapes from one particular perspective, such as metapopulation dynamics or vegetation thresholds. Such highly focused research has produced useful insights, but arguably, it can also be problematic because it can lead to the oversight of valuable insights from other perspectives (e.g. Moilanen & Hanski, 2001).
The aim of this paper is to provide a holistic view of the ecology of modified landscapes by synthesising recent developments across a range of different research themes. By necessity, this breadth means that any given sub-component of landscape change is covered only in a cursory manner, emphasizing recent key insights over areas of internal disagreement. We believe that this conscious focus on breadth is needed to help researchers from a range of backgrounds to appreciate work from different backgrounds that is equally relevant to understanding the ecology of modified landscapes. We acknowledge that our choice of material included in this review may be contentious. Most importantly, we hope this review will stimulate more holistic thinking about the ecology and management of modified landscapes.
First, we introduce a conceptual framework for understanding the effects of landscape modification on species and assemblages. Second, threatening processes associated with landscape modification, and how they affect individual species and immediate species interactions, are discussed. Third, the relationships between human-perceived patterns in modified landscapes and assemblages of species are investigated, and cascading effects arising from landscape modification are considered. Finally, key knowledge gaps are identified, and a short list of tangible management recommendations for conservation management in modified landscapes is provided.
The main emphasis of this review is on animals, with plants largely referred to in the context of their interactions with animals. Despite this bias, we acknowledge that landscape modification also may change plant assemblages (Hobbs & Yates, 2003; Mayfield & Daily, 2005). The themes covered in this paper are outlined in Fig. 1. Frequently used terms that have contributed to confusion in past work are defined explicitly in Table 1.
Table 1. Selected key terms to conceptualise the ecology of modified landscapes
Connectedness of ecological processes at multiple spatial scales (see Souléet al., 2004, for details)
Following McIntyre & Hobbs (1999), a landscape characterised by a strong contrast between vegetation patches and their surrounding matrix (native vegetation cover typically c. 10–60%); often seen in formerly forested areas
The range of environments suitable for a particular species
Connectedness of habitat for a particular species; the opposite of habitat isolation
Degree of isolation between habitat patches used by a particular species; opposite of habitat connectivity
Loss of habitat for a particular species
Sub-division of habitat for a particular species
A human-defined area ranging in size from c. 3 km2 to c. 300 km2
A human perception of the connectedness of native vegetation cover in a landscape
A human perspective of environmental gradients and land-cover types in a landscape
The dominant background patch type in a landscape; in modified landscapes usually not native vegetation
Following McIntyre & Hobbs (1999), a landscape characterised by sharp boundaries between a minimal amount of remnant native vegetation (< 10%) and surrounding modified land; often seen in areas with intensive agriculture
Following McIntyre & Hobbs (1999), a landscape characterised by gradual boundaries between native vegetation and surrounding modified land (native vegetation cover typically c. 60–90%); often seen in areas with extensive livestock grazing
A CONCEPTUAL FRAMEWORK OF LANDSCAPE MODIFICATION
Typical patterns of landscape modification
In this paper, we are concerned with landscape modification that has occurred within the last few centuries and that has negatively affected native species diversity. Broadly similar patterns of this type of landscape modification have been documented around the world (e.g. Sharpe et al., 1987; Bennett, 1990; Saunders et al., 1993; Webb, 1997). Typically, as landscape modification increases, more native vegetation is lost, land-use intensity in modified areas increases, and remnants of native vegetation are increasingly influenced by processes originating in modified areas (McIntyre & Hobbs, 1999; Fig. 2). Common spatial patterns in human-modified landscapes include the perforation, dissection, sub-division, shrinkage, and attrition of native vegetation (Forman, 1995; Fahrig, 2003). Generally, native vegetation is cleared first in areas with high primary productivity (e.g. Norton et al., 1995; Landsberg, 1999), although this trend is somewhat modified by jurisdictional and cultural aspects of land-use history (Lunt & Spooner, 2005).
A continuum of research approaches
Fragmentation-related research has suffered from a lack of a clear conceptual basis and reliance on overly restrictive conceptual models of modified landscapes (reviewed by Bunnell, 1999; Haila, 2002). In addition, many themes are widely agreed to be important, but have usually been treated separately in existing reviews. To overcome these problems, an integrated rather than a divisive conceptual framework is briefly introduced here, which recognises the complementary contributions of many different types of research.
Two extremes can be identified along a continuum of approaches to understanding the effects of landscape modification on species and assemblages. The first extreme is ‘species-oriented’. Species-oriented approaches are often centred around individual species, which are believed to respond individualistically to their environment (Gleason, 1939; Austin & Smith, 1989; Manning et al., 2004). Species-oriented approaches recognise that each species responds individualistically to a range of processes related to its requirements for food, shelter, space and suitable climatic conditions, as well as interspecific processes like competition, predation and mutualisms (Fischer & Lindenmayer, 2006; Fig. 3). A key limitation of species-oriented approaches is that it is impossible to study every single species in any given landscape.
At the other end of the continuum are ‘pattern-oriented’ approaches. Here, the focus is typically on human-perceived landscape patterns and their correlation with measures of species occurrence, including aggregate measures such as species richness. Pattern-oriented approaches originate from island biogeography (MacArthur & Wilson, 1967) and are the traditional stronghold of ‘fragmentation-related’ research (Haila, 2002). Two widely used pattern-oriented conceptual landscape models are the patch–matrix–corridor model (Forman, 1995) and to a lesser extent the variegation model (McIntyre & Barrett, 1992; Ingham & Samways, 1996). Although various pattern-oriented approaches differ in their assumptions, all are based on human-defined land cover (often native vegetation), and seek to establish correlations with species or groups of species to infer potential ecological causalities (Fig. 3). The main limitation of pattern-oriented approaches is that they frequently aggregate across individual species and ecological processes. In some cases, this aggregation may lead to an under-appreciation of the complexity of ecological processes and differences between individual species.
Given the unique practical and theoretical strengths and weaknesses of species-oriented and pattern-oriented approaches, the two approaches are highly complementary for understanding and managing modified ecosystems. The two approaches are considered separately below, and research priorities and management recommendations are derived that recognise the value of both approaches.
A useful way to understand potentially negative effects of landscape modification on native taxa is to consider the range of processes that may threaten a given individual species. Threatening processes associated with landscape modification may be broadly classified as exogenous (originating independently of the species’ biology) or endogenous (originating as part of the species’ biology), although this distinction may be blurred in some instances.
Exogenous threatening processes
Habitat may be broadly defined as the range of environments suitable for a given species. That is, it is a species-specific concept. Habitat loss is the dominant threat to species around the world (Sala et al., 2000). Landscape modification for agriculture and urbanisation typically causes habitat loss for many species (Kerr & Deguise, 2004; Luck et al., 2004). Because native vegetation is important for many species, numerous authors have equated ‘habitat’ with ‘native vegetation’ (e.g. Andrén, 1994). Although this classification may be appropriate in some situations (Terborgh et al., 2001), in many situations it can be misleading. This is because a binary classification of land into habitat (native vegetation) and non-habitat (other land cover) ignores habitat suitability gradients and differences between species with respect to what constitutes suitable habitat for them (Fig. 3; Andrén et al., 1997); importantly, many native species can be conserved in well-managed production landscapes (Daily, 2001; Lindenmayer & Franklin, 2002). On this basis, we suggest that the term ‘habitat’ and associated terms like ‘habitat fragmentation’ be used only in a single-species context (Table 1). The broader use of the term habitat (i.e. equating it with native vegetation) can result in the under-appreciation of differences between the unique habitat requirements of different species, and the under-appreciation of the potential habitat value of modified environments for some species. For a given species, habitat loss rarely occurs in isolation from other threats, but tends to coincide with habitat degradation, habitat sub-division and a range of additional threatening processes (Liu et al., 2001; Fig. 4).
Habitat degradation is the gradual deterioration of habitat quality. In degraded habitat, a species may decline, occur at a lower density, or may be unable to breed (Temple & Cary, 1988; Felton et al., 2003; Hazell et al., 2004). Degraded habitat may constitute an ‘ecological trap’ to which individuals of a species are attracted but in which they cannot reproduce (Battin, 2004). Habitat degradation can be difficult to detect because: (1) some types of degradation take a long time to manifest (e.g. recruitment failure of cavity trees; Saunders et al., 2003), and (2) some species with slow life cycles may continue to be present in an area even if they are unable to breed (e.g. cockatoos reliant on nesting hollows; Saunders, 1979). Factors related to habitat degradation vary widely between species, and may include pressure from grazing (Spooner et al., 2002), logging (Recher & Serventy, 1991), or changed thermal regimes (Jäggi & Baur, 1999). Detailed autoecological studies and monitoring are often required to detect and counteract habitat degradation (Borghesio & Giannetti, 2005).
Habitat sub-division is the breaking apart of continuous habitat into multiple patches; it is synonymous with what some authors have termed ‘fragmentation’ (Fahrig, 2003). Smaller habitat patches can lead to population declines (Bender et al., 1998), for example because resources in smaller patches may be more limited (Zanette et al., 2000). In addition, habitat sub-division increases the isolation of remaining habitat areas. Habitat isolation can negatively affect day-to-day movements of a given species (e.g. between nesting and foraging resources; Saunders, 1980; Luck & Daily, 2003). Habitat isolation also may negatively affect the dispersal of juveniles (Cooper & Walters, 2002). Metapopulations, i.e. ‘set[s] of local populations which interact via individuals moving between local populations’ sometimes develop as a result of habitat isolation (Hanski & Gilpin, 1991). Notably, patchy populations are true metapopulations only if movement between sub-populations is neither very uncommon nor very common (Hanski & Simberloff, 1997). Finally, habitat isolation may negatively affect large-scale movements of species such as seasonal migration or range shifts in response to climate change (Souléet al., 2004). The extent to which landscape modification results in habitat isolation depends on the interaction between a given species’ dispersal behaviour, mode and scale of movement, what constitutes suitable habitat for it, and how a given landscape has been modified. Habitat connectivity is the opposite of habitat isolation, and is contrasted against other connectivity concepts in the section entitled Connectivity below.
Endogenous threatening processes
In addition to direct negative impacts on a species’ habitat, declining species in modified landscapes often experience disruptions or changes to their biology, behaviour and interactions with other species. These changes are often triggered by exogenous threats, but may constitute threatening processes in their own right. Landscape modification can lead to altered breeding patterns and social systems. For example, birds may have shorter breeding seasons, lay fewer eggs, and rear fewer nestlings (Hinsley et al., 1999; Zanette et al., 2000), or their mating systems may change (Ims et al., 1993). Many other types of behavioural and biological changes have been observed for animals in modified landscapes, including disruptions to dispersal (Brooker & Brooker, 2002), changed movement patterns over greater distances (Recher et al., 1987; Norris & Stutchbury, 2001), altered home ranges (Pope et al., 2004), higher incidences of fluctuating body asymmetry (Sarre, 1996), changed vocalisation patterns (Slabbekoorn & Peet, 2003; Lindenmayer et al., 2004), and disrupted group behaviours (Gardner, 2004).
Changes to species interactions may affect competition, predation, parasitism and mutualisms. Increased competition can occur, for example, for insectivorous woodland birds in many Australian farming landscapes where the aggressive noisy miner (Manorina melanocephala) has increased in abundance (Grey et al., 1997). Increased predation and parasitism have both been frequently reported in modified landscapes, especially for birds (Robinson et al., 1995; Lahti, 2001; Zanette et al., 2005), but more recently also for complex insect–plant food webs (Valladares et al., 2006). Increased pressure from competition and predation can be particularly severe when introduced species are involved. For example, competition by exotic snails has severely decimated Hawaii's native snail fauna (Hadfield et al., 1980).
Landscape modification also may disrupt mutualisms. For example, Cordeiro & Howe (2003) demonstrated the disruption of the mutualism between the endemic tree Leptonychia usambarensis and fruit-dispersing birds in a modified landscape in Tanzania. Similarly, Kearns et al. (1998) argued that landscape modification may disrupt pollination throughout the world, as recently emphasized by Ricketts et al. (2004) for Costa Rican coffee farms.
Disruptions to species interactions have particularly severe effects when strongly interacting species are involved, which play a disproportionate role in maintaining ecosystem function (Terborgh et al., 2001; Souléet al., 2005). Such species are sometimes also termed keystone species (Paine, 1969; Power et al., 1996), and their importance is discussed in more detail below in the section entitled Extinction cascades.
Stochastic threatening processes
The exogenous and endogenous threatening processes discussed above are deterministic. Deterministic threatening processes are those which predictably lead to declines (Gilpin & Soulé, 1986; Fig. 4). In addition to deterministic processes, stochastic processes may threaten species in modified landscapes. Exogenous stochastic threats are related to environmental variability, such as fluctuations in climate or natural catastrophes like hurricanes or wildfires (Simberloff, 1988). Endogenous stochastic threats occur as part of a species’ life cycle, and include demographic stochasticity (e.g. year-to-year variability in reproductive success) and genetic stochasticity (e.g. genetic drift). Endogenous stochastic threats are more pronounced in small populations (Roughgarden, 1975; Keller & Waller, 2002).
Interactions among threatening processes and extinction proneness
Species declining as a result of landscape modification are typically affected by both deterministic and stochastic threats. Exogenous threats often lead to the initial decline of a species. The resulting smaller populations, in turn, are more susceptible to endogenous threats that reinforce the decline of the species (Clark et al., 1990; Fig. 4).
Many factors have been suggested to be related to the extinction proneness of species in modified landscapes, including habitat or niche specialisation, home range size, mobility, extent of geographic distribution, population density or rarity, edge sensitivity, body size and dietary specialisation (Johns & Skorupa, 1987; Brashares, 2003; Koh et al., 2004; Cardillo et al., 2005; Kotiaho et al., 2005). Although these factors are often considered as equally plausible in studies on extinction proneness, it is important to note that some of them are more directly related to extinction proneness in a causal sense than others. For example, body size is a proxy for a range of other ecological attributes, including area requirements, mobility and dietary specialisation (Laurance, 1991), whereas habitat specialisation is directly linked to the threatening process of habitat loss. Table 2 outlines some of the most direct links between known threatening processes and likely factors contributing to (or ameliorating) extinction proneness. Notably, some of these links are complex, and potentially contentious. For example, although high mobility can help species to move between habitat patches (Table 2), it may also lead to an increased number of individuals dispersing into unsuitable habitat, thereby threatening population persistence (Gibbs 1998; Casagrandi & Gatto 1999). Further investigation of the links between threatening processes and extinction proneness in the future may lead to an increasingly robust, process-based understanding of species extinctions. A likely emergent pattern is that a given species’ ability to withstand human landscape modification is related to the extent to which landscape modification causes habitat loss and isolation, and the disruption of biological and interspecific processes for that individual species. Small population size (natural or human-induced) will further exacerbate a species’ risk of extinction due to stochastic events (Fig. 4; Table 2).
Table 2. Proposed relationship between key threatening processes associated with landscape modification and biological attributes of species which ameliorate extinction proneness
Ameliorating biological attribute
Habitat loss and habitat degradation
Low habitat specialisation
Specialised species are more likely to lose their habitats as a result of landscape change
Disturbance-tolerant species are more likely to find suitable habitat in modified landscapes
Ability to live in the matrix
Species that can live in the matrix experience no habitat loss as a result of landscape modification
Habitat isolation and sub-division
Ability to move through the matrix
Species that can move through the matrix are less likely to suffer the negative consequences of habitat isolation
Strong dispersers may be more likely to maintain viable metapopulations (but note this is contentious — see text)
Disrupted species interactions
Limited dependence on particular prey or mutualist species
Species that can switch prey or mutualists are more likely to withstand landscape change
Species that are strong competitors are less likely to be outcompeted by species whose habitat expands as a result of landscape change
Low biological and behavioural complexity
Species with a complex biology (e.g. social or breeding systems) are more likely to have their biological processes disrupted as a result of landscape change than species with simpler biological systems
High density populations contain many individuals even in a small area, and hence are more resilient to stochastic threats
Assessing the relationship between human-perceived landscape patterns and individual species or species assemblages has been a popular, albeit controversial, research area in ecology and conservation biology (Haila, 2002). Some key attributes of landscape pattern are discussed below.
Amount and condition of native vegetation
For many species of plants and animals, habitat conditions are ideal in large areas of unmodified native vegetation. Hence, the loss of native vegetation at landscape and regional scales has been linked to the loss of native species around the world (e.g. Andrén, 1994; Kerr & Deguise, 2004). Similarly, the loss of native vegetation at the local scale tends to reduce native species richness. Other things being equal, small patches of native vegetation support fewer native species than large patches (e.g. Bellamy et al., 1996; Rosenblatt et al., 1999). Such species–area relationships have long been known (Arrhenius, 1921). Several mechanisms are likely to underlie species–area relationships (McGuinness, 1984). Larger patches may have a higher ratio of colonisations to extinctions (MacArthur & Wilson, 1967), are more likely to contain undisturbed areas which are required by some species (Harris, 1984), are more likely to capture a range of environmental conditions which constitute habitat for different sets of species (Harner & Harper, 1976; Fox, 1983), and are more likely to capture patchily distributed species by chance (Connor & McCoy, 1979).
Despite strong evidence of the importance of large areas of native vegetation, it is important to recognise that: (1) small areas of native vegetation can be important complements to large areas (Fischer & Lindenmayer, 2002; Tscharntke et al., 2002); (2) land management outside patches of native vegetation can have both positive and negative context effects on patches (Wethered & Lawes, 2005); and (3) different types of native vegetation will support different sets of species (Austin & Smith, 1989; Sabo et al., 2005). For these reasons, the exclusive focus on large patches of native vegetation is often overly restrictive and represents an outdated conceptual model of landscape modification (Haila, 2002; Manning et al., 2004).
Even where extensive areas of native vegetation remain, deterioration of vegetation structure and deterioration of the physical environment can negatively affect many native species. For example, domestic livestock grazing can simplify vegetation structure and exacerbate the replacement of native species by introduced species (Hobbs, 2001), as well as reduce regeneration rates of native woody vegetation (Spooner et al., 2002). In dry environments, these processes may eventually result in desertification (Milton et al., 1994). Many other land management practices such as logging or firewood collection can be linked to the deterioration of vegetation structure and the physical environment (Driscoll et al., 2000; Chettri et al., 2002; Lindenmayer & Franklin, 2002). Typically, the loss and degradation of native vegetation co-occur in modified landscapes (Liu et al., 2001; Klink & Machado, 2005).
Edge effects are changes in physical and biological conditions at an ecosystem boundary or within adjacent ecosystems. Abiotic edge effects refer to changes in physical variables such as radiation, moisture, temperature, humidity, wind speed and soil nutrients (Chen et al., 1990; Matlack, 1993; Weathers et al., 2001). Biotic edge effects are changes in biological variables such as species composition of plants and animals, or patterns of competition, predation and parasitism (Malcolm, 1994; Robinson et al., 1995; Lahti, 2001; Valladares et al., 2006). The penetration depth of edge effects can vary widely from tens of metres for variables like soil moisture (Laurance et al., 1997) to several kilometres in the case of recruitment failure of Dipterocarpaceae in Gunung Palung National Park in western Borneo (Curran et al., 1999).
Despite substantial regional variation in edge effects (e.g. for nest predation; Batary & Baldi, 2004), several factors are likely to enhance the presence and magnitude of edge effects. These include high structural contrast at the edge, high wind speeds and temperatures, and the presence of pioneer, exotic and invasive taxa that may benefit from edge environments (Harper et al., 2005). In addition, a large amount of variability in ecological patterns around edges may be effectively explained by distinguishing between four fundamental underlying mechanisms: ecological flows, access to spatially separated resources, resource mapping and species interactions (Ries et al., 2004). Although edge effects are considered to have negative effects on native ecosystems, not all edges are necessarily detrimental for all native species (Yahner, 1988), especially where edges are gradual or of low structural contrast (Tubelis et al., 2004).
It is widely agreed that connectivity is important for biological conservation (Taylor et al., 1993). However, the term ‘connectivity’ is often used loosely, and different authors use the term in different ways. For example, Tischendorf & Fahrig (2000) considered landscape connectivity to be an attribute of landscapes that resulted from the interaction of land cover with individual species’ movement rates. In contrast, Moilanen & Hanski (2001) took a metapopulation perspective and suggested that connectivity was better understood as an attribute of individual patches. Other authors have distinguished between structural connectivity and functional connectivity. Using this distinction, structural connectivity is related to landscape pattern and results from the density and complexity of corridors, the distance between patches, and the structure of the matrix (Uezu et al., 2005). In contrast, functional connectivity is often defined by the extent to which an individual species of interest can move through a landscape (Uezu et al., 2005). The many different, and often implicit, definitions of connectivity have contributed to considerable confusion and debate; for example, regarding the question if corridors provide connectivity (Simberloff et al., 1992; Beier & Noss, 1998; Proche@ et al., 2005).
To overcome existing confusion, we differentiate between three types of connectivity (Fig. 5).
1 Habitat connectivity is the connectedness between patches of suitable habitat for a given individual species. It may be defined at the patch scale (e.g. Moilanen & Hanski 2001) or at the landscape scale (e.g. Hanski & Ovaskainen 2000; Tischendorf & Fahrig 2000). The term is chosen to include the word ‘habitat’ to emphasize its species-specific nature.
2Landscape connectivity is a human perspective of the connectedness of native vegetation cover in a given landscape. It may be expressed using various buffer- or distance-based metrics that can be calculated from maps of human-defined land cover (e.g. Gustafson 1998; Moilanen & Nieminen 2002). The term is chosen to include the word ‘landscape’ to emphasize its anthropocentric nature — the concept of a landscape is a human construct (Table 1).
3Ecological connectivity is the connectedness of ecological processes across multiple scales, including trophic relationships, disturbance processes and hydroecological flows. The measurement of ecological connectivity is not straightforward and depends on which aspect of ecological connectivity is to be estimated. Despite this difficulty, ecological connectivity is an important concept that is not adequately captured by existing definitions of connectivity. The term is based on a discussion by Souléet al. (2004).
The three connectivity concepts are related but not synonymous (Fig. 5). Landscape connectivity may translate into habitat connectivity for some but not all species. For example, corridors and stepping stones (i.e. small vegetation patches scattered through a landscape) always contribute to landscape connectivity, but may not be used by all native species – that is, they do not contribute to habitat connectivity for those species (Forman, 1995; Beier & Noss, 1998). Similarly, the relationship between landscape connectivity and ecological connectivity tends to be positive, but not all ecological processes are effectively facilitated through all types of landscape connectivity. For example, seed-dispersing birds used corridors in a study in South Carolina (USA). Here, enhanced landscape connectivity increased one aspect of ecological connectivity, that is, the process of seed dispersal (Levey et al., 2005a). However, neither corridors nor stepping stones may effectively maintain some aspects of ecological connectivity such as natural hydrological flows or the natural spread of fire throughout a landscape (Souléet al., 2004).
In practice, landscape connectivity is the most easily manageable aspect of connectivity because it requires no detailed understanding of individual species’ habitat requirements or ecological processes. Landscape connectivity may be enhanced through corridors, stepping stones, and the maintenance of a ‘soft’ matrix which is structurally similar to native vegetation (sensu Franklin, 1993). Despite some ecological risks such as potentially facilitating the spread of introduced species (Simberloff et al., 1992; Proche@ et al., 2005), increased landscape connectivity is usually more likely to have desirable effects on native species and ecological processes than undesirable effects (Noss & Beier, 2000; Haila, 2002; Levey et al., 2005b).
Some landscapes are so heterogeneous that it becomes difficult to delineate patches and a matrix in a meaningful way. Heterogeneous landscapes are characterised by a diversity of environmental gradients and land-cover types (August, 1983). Although situations where heterogeneity reduces species richness may exist, generally, heterogeneous modified landscapes support more species than otherwise similar but less heterogeneous modified landscapes (Tscharntke et al., 2005). For example, landscape heterogeneity is recognised as beneficial for native species in European farming landscapes (Benton et al., 2003; Hole et al., 2005), Central American farming landscapes (Luck & Daily, 2003; Mayfield & Daily, 2005), and forestry landscapes around the world (Lindenmayer & Franklin, 2002). Heterogeneity arising from natural disturbance processes often provides a useful indication of the landscape or vegetation patterns most likely to benefit native species (Lindenmayer & Franklin, 2002).
Landscape modification typically results in the loss of native vegetation and changes to its spatial distribution, altered disturbance regimes and deterioration of vegetation structure (see above). These processes can interact and cause cascading ecosystem changes and regime shifts in ecosystem functioning. Regime shifts occur when inter-relationships between key variables in an ecosystem change fundamentally — they can be thought of as transitions where an ecosystem ‘flips’ from one state to another (Scheffer et al., 2001; Folke et al., 2004; Groffman et al., 2006). Extinction cascades occur where the extinction of one species triggers the loss of one or more other species, which in turn leads to further species extinctions (e.g. Terborgh et al., 2001).
Amount and configuration of native vegetation cover
Cascading effects on native species can result from the simultaneous reduction of the total amount of native vegetation and landscape connectivity. Several studies have suggested that connectivity loss is particularly severe below a 30% threshold of native vegetation cover, leading to increased losses of species reliant on native vegetation (Andrén, 1994; Fahrig, 2003; Radford et al., 2005). Notably, not all species depend on native vegetation, and many other aspects of landscape modification also affect the distribution of species. Hence, the ‘30% rule’ does not apply for all species or all ecosystems (Parker & Mac Nally, 2002; Lindenmayer et al., 2005). Perhaps the most useful, albeit general, conclusion from the literature on vegetation thresholds is that extinction cascades are particularly likely to occur at low levels of native vegetation cover.
Disturbance regimes and structural complexity
Both natural and human disturbances can alter the vegetation structure in modified landscapes. For example, fire is a major disturbance agent that has been significantly altered in modified landscapes around the world (Agee, 1993; Turner et al., 2003). Anthropogenic disturbances, like livestock grazing or logging, also can significantly alter vegetation structure (see above).
The loss of some structural elements can have particularly severe negative effects on native species, and is especially likely to trigger extinction cascades. Tews et al. (2004) defined elements of structural complexity that are particularly important to many species as ‘keystone structures’. Examples of keystone structures include scattered trees in dry savannas, which alter the abiotic environment and provide habitat for a large number of plants and animals (Vetaas, 1992; Dean et al., 1999; Manning et al., 2006).
A change in species composition per se can trigger extinction cascades. The loss of individual species is particularly likely to trigger extinction cascades when ‘keystone’ or ‘strongly interacting’ species are involved because they exert a disproportionate effect on ecosystem function relative to their abundance (Paine, 1969; Power et al., 1996; Souléet al., 2005). Strong ecological interactions may be apparent via habitat enrichment, predation, competition or mutualisms (Souléet al., 2005). For example, Terborgh et al. (2001) showed that the absence of predators from forest fragments in Venezuela had led to dramatic increases in herbivores. This, in turn, had a range of cascading effects throughout the food chain, including reduced levels of tree recruitment.
Ecosystems are more likely to absorb external shocks and maintain their function if multiple species fulfil similar ecological roles (Elmqvist et al., 2003; Tscharntke et al., 2005). Such ‘functional redundancy’ among species increases the chances of some species being able to compensate for a given species’ ecological role if it becomes rare or goes extinct (Walker, 1992, 1995). Functional redundancy across multiple scales is considered an important aspect of ecosystem resilience (Elmqvist et al., 2003; Folke et al., 2004).
SUMMARY: KEY INSIGHTS FROM EXISTING WORK
Complementary insights on the ecology of modified landscapes have been gained from species-oriented and pattern-oriented work. Species-oriented work has the advantage of being based on well-established ecological causalities, but its main limitation is that it is impossible to study every individual species in any given landscape. In contrast, pattern-oriented work provides broadly applicable general insights, but may oversimplify complex causal relationships and subtle differences between individual species. Key insights from existing work are listed below.
• Habitat is a species-specific concept, and should not be equated with native vegetation.
• Threats for a given species arise from negative changes to its specific habitat, and disruptions to its biology, behaviour and interactions with other species.
• Extinction proneness can be studied in direct causal relation to the above threats.
• A large amount of unmodified native vegetation tends to benefit native species, but there appears to be no universally applicable minimum ‘threshold’ amount of native vegetation.
• Edge effects are diverse in their consequences and penetration depth, but recent reviews suggest they may be less idiosyncratic than previously thought.
• Existing debates on connectivity may be partly solved by distinguishing between landscape connectivity, habitat connectivity and ecological connectivity.
• Conditions in the matrix and landscape heterogeneity are fundamentally important in modified landscapes, and deserve equal attention in research and management as patches of native vegetation.
• Regime shifts and extinction cascades are particularly likely to arise when native vegetation cover is severely reduced, when vegetation structure throughout the landscape is greatly simplified, and when entire functional groups or keystone species are lost.
Table 3. List of pattern-oriented and species-oriented management strategies to mitigate potential negative effects of landscape modification on native species (summarised from Fischer et al., 2006)
Maintain and/or restore large and structurally complex patches of native vegetation
Provide core habitat for species which depend on native vegetation
Maintain and/or restore a matrix that is structurally similar to native vegetation
Provide habitat for many species throughout the landscape
Provide habitat connectivity for many species throughout the entire landscape
Reduce structural contrast between modified and unmodified areas, thereby reducing edge effects throughout the landscape
Maintain and/or restore buffers around sensitive areas
Protect sensitive areas like aquatic systems
Reduce edge effects at specific locations
Maintain and/or restore corridors and stepping stones
Provide habitat connectivity for species that cannot move through the matrix
Provide additional habitat for species that use native vegetation but are not area-sensitive
Maintain and/or restore landscape heterogeneity and capture environmental gradients
Provide a diversity of habitats that are useful to a range of different species
Distribute different land-use intensities across natural gradients in climate, topography, and primary productivity
Maintain key species interactions and functional diversity
Protect important ecosystem processes
Protect characteristic ecosystem structure and feedbacks
Maintain or apply appropriate disturbance regimes
Encourage characteristic vegetation structure
Create characteristic spatial and temporal variability in vegetation patterns
Maintain species of particular concern
Ensure the continued survival of rare or threatened species
Control aggressive, over-abundant and invasive species
Reduce competition and predation by undesirable species which could negatively affect desirable species
Maintain characteristic species composition
Minimise ecosystem-specific threatening processes
Identify problems that may affect biodiversity but are not directly related to landscape modification
Establish protocols to eliminate these problems
FUTURE RESEARCH PRIORITIES
The conceptual framework outlined above recognises the complementary contributions of pattern-oriented and species-oriented research approaches in work on landscape modification and habitat fragmentation (Figs 1 & 3). Already, a combination of work on patterns and individual species has led to particularly useful insights on ecosystem functioning, as demonstrated by examples on metapopulation dynamics (Hanski 1994; Wiegand et al., 2003), species distribution patterns (Bennett 1987) and amphibian declines (Pounds et al., 2006). The explicit recognition of complementarity between species-oriented and pattern-oriented work may contribute to a further closing of the perceived gap between pattern and process in the ecological sciences (Hoekstra et al., 1991; Hobbs, 1997; Fischer et al., 2004). The proposed framework (Figs 1 & 3) may improve future work for several reasons. First, it overcomes vagueness in terminology, especially the confusion of species-specific concepts like habitat loss or habitat connectivity with pattern-based concepts like vegetation clearing or landscape connectivity (Table 1). Second, it provides a conceptual basis to simultaneously appreciate subtle species-specific differences while maintaining the search for general patterns (Fig. 3). Third, it provides greater conceptual clarity, thereby facilitating improved synthesis of past work and improved hypothesis generation for future work (e.g. Table 2). Fourth, it covers a wide breadth of inter-related issues, thereby encouraging holistic thinking about the ecology of modified landscapes.
Particularly interesting research opportunities exist in the following areas.
5 Future research could be strengthened by investigating larger spatial and temporal scales than have been typical in the past (Fazey et al., 2005), for example through the use of well-designed natural experiments (sensu Diamond, 1986).
6 Given potential trade-offs and synergies between biodiversity conservation and other objectives of human land use, ecologists should further strengthen links with other academic disciplines such as economics, agriculture and ethics (Daily & Ehrlich, 1999) as well as conservation policy and management (Mattison & Norris, 2005; Souléet al., 2005).
We are grateful for constructive criticism of our work on the ecology of modified landscapes by D. Saunders, R. Noss, Y. Haila, A. Bennett, Adam and Annika Felton, N. Munro, R. Montague-Drake and K. Youngentob, as well as for critical comments on this paper by I. Hanski and an anonymous referee. We greatly appreciate financial support by the Kendall Foundation, the Australian Research Council, and Land & Water Australia.
Joern Fischer is a postdoctoral fellow at the Centre for Resource and Environmental Studies (CRES) at The Australian National University. His main research interest is biodiversity conservation in human-modified landscapes.
David Lindenmayer is a professor at the Centre for Resource and Environmental Studies (CRES) at The Australian National University. He has published extensively in wildlife ecology, forest ecology and conservation biology.