MOVEMENT BEHAVIOR, PATCH OCCUPANCY, SUSTAINABLE PATCH NETWORKS AND CONSERVATION PLANNING FOR AN ENDEMIC UNDERSTORY BIRD By TRACI DARNELL CASTELL"N A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
Copyright 2006 by Traci Darnell CastellÃ³n
iii ACKNOWLEDGMENTS I sincerely thank my advisor, Kathryn Sieving, for her unwavering support and assistance, and my graduate committee, Lyn Branch, Michael Binford, Graeme Cumming, Doug Levey, and Emelio Bruna. I am also extremely grateful to my field assistants Hector JaÃ±ez, John Davis, Alvaro Wurstten, Emma Elgueta, Juan Carlos Correra, computer assistant Nia Haynes, and the many land owners in ChiloÃ© and Osorno who graciously provided access to their farms. This work would not have been possible without their contributions. In addition, I have appreciat ed and benefited from the support of my friends and colleagues Daniel Smith, Mike Milleson, Matt Reetz, Tom Contreras, Nat Seavy, Marcella Machicote, Gr eg Jones, and Ivan DÃaz. I gratefully acknowledge Mary Willson, Juan Armesto, and Cecilia Smith. Finally, and above all, I thank my parents; Carolyn Blethen and Char les Darnell, my husband Charles CastellÃ³n, and my friends, for helping me remember what is important. Partial funding was provided by the Disney Conservation Fund. In-kind support was provided by FundaciÃ³n Senda Darwin, the University of Florida Map and Image Library, Geoplan Center, the Geography Department, the Land Use and Environmental Change Institute, and the Department of Wildlife Ecology and Conservation.
iv TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES.............................................................................................................vi LIST OF FIGURES..........................................................................................................vii ABSTRACT.....................................................................................................................vi ii CHAPTER 1 INTRODUCTION........................................................................................................1 Theoretical Background................................................................................................2 Study System................................................................................................................5 Research Overview.......................................................................................................8 2 AN EXPERIMENTAL TEST OF MATRIX PERMEABILITY AND CORRIDOR USE.......................................................................................................11 Introduction.................................................................................................................11 Methods......................................................................................................................13 Study System.......................................................................................................13 Experimental Design...........................................................................................15 Selection of Release Sites....................................................................................16 Capture and Handling..........................................................................................17 Radio Telemetry..................................................................................................18 Measurement of Landscape Metrics....................................................................21 Data Analysis.......................................................................................................22 Results........................................................................................................................ .23 Discussion...................................................................................................................27 Corridor versus Matrix Movement......................................................................27 Generality of Results...........................................................................................28 Conservation Implications...................................................................................29 3 LANDSCAPE HISTORY AND FRAGMENTATION EFFECTS ON PATCH OCCUPANCY............................................................................................................36 Introduction.................................................................................................................36 Methods......................................................................................................................38 Study System.......................................................................................................38
v Land Cover Analysis...........................................................................................40 Patch Occupancy Surveys...................................................................................42 Habitat Quality Characterization.........................................................................43 Predictive Models................................................................................................45 Results........................................................................................................................ .47 Land Cover Analysis...........................................................................................47 Chucao Distribution Patterns...............................................................................47 Predictive Models................................................................................................48 Discussion...................................................................................................................49 Landscape Analysis.............................................................................................49 Chucao Distribution Patterns...............................................................................50 Model Validation.................................................................................................52 4 SUSTAINABLE PATCH-NETWORK CRITERIA..................................................63 Introduction.................................................................................................................63 Methods......................................................................................................................65 Study Species.......................................................................................................65 Landscape Criteria for Sustainable Populations..................................................65 Sustainable Population Size................................................................................66 Habitat Area Requirements.................................................................................67 Matrix Composition and Connectivity................................................................70 Sustainability Criteria Summary.........................................................................72 Graph Theory Format..........................................................................................73 Analysis of Test Landscapes...............................................................................74 Results........................................................................................................................ .75 Discussion...................................................................................................................76 Restoring Landscape Connectivity......................................................................80 Conclusions.........................................................................................................84 5 CONCLUSIONS AND RECOMMENDATIONS.....................................................91 Dissertation Summary................................................................................................92 Conservation Implications and Recommendations....................................................97 LIST OF REFERENCES.................................................................................................101 BIOGRAPHICAL SKETCH...........................................................................................114
vi LIST OF TABLES Table page 2-1. Summary of radio-telemetry re sults for translocated subjects.................................31 3-1. Metrics describing spatial patterns of forest cover in the ChiloÃ© and Osorno study areas in 1961 and 1993...................................................................................56 3-2. Mean (Â± S.E.) values for me trics describing focal patches......................................58 3-3. Classification-tree m odel confusion matrices..........................................................59
vii LIST OF FIGURES Figure page 2-1. Schematic representation of the three landscape treatments....................................33 2-2. Hazard function reflecting chance of movement by translocated Chucaos.............34 2-3. Map (1:2,000) of wooded habitat patches................................................................35 3-1. Satellite image (grayscal e) of the study region........................................................59 3-2. Mean patch area (ha) plotted agai nst patch density (p atches per ha).......................60 3-3. Classification-tree for predicting patc h occupancy by Chucao Tapaculos in the ChiloÃ© landscape.......................................................................................................61 3-4. Classification-tree for predicting patc h occupancy by Chucao Tapaculos in the Osorno landscape......................................................................................................62 4-1. Example landscape graph.........................................................................................86 4-2. ChiloÃ© study area landscape graph...........................................................................87 4-3. Osorno study area landscape graph..........................................................................88 4-4. Puerto Montt study area landscape graph................................................................89 4-5. Suggested design of large-scale corridors................................................................90
viii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MOVEMENT BEHAVIOR, PATCH OCCUPANCY, SUSTAINABLE PATCH NETWORKS AND CONSERVATION PLANNING FOR AN ENDEMIC UNDERSTORY BIRD By Traci Darnell CastellÃ³n May 2006 Chair: Kathryn Sieving Major Department: Wildlife Ecology and Conservation This dissertation research examined effect s of habitat loss and fragmentation on the Chucao Tapaculo ( Scelorchilus rubecula ), an understory bird e ndemic to South American temperate rainforest. I experimentally test ed the relative permeability of three landscape elements to movement by the Chucao, a ssessed cover changes since 1961 in two landscapes that differed in levels and duration of fragmentation, and developed classification-tree models to predict patch o ccupancy in each. Based on these analyses, I developed a set of simple criteria to disti nguish habitat configurat ions with reasonably high expectancy of supporting sustainable Chucao populations. Then, I applied the criteria in three test landscapes to assess the potential for long-term persistence under a range of landscape conditions, and tested the potential conservation benefit of hypothetical restoration of habitat conn ections among isolated patches.
ix Results indicated that Chucao movement was significantly constrained by open habitat but occurred equally well through w ooded corridors and shrub-dominated matrix. Thus, corridor protection and management of vegetation in the matrix (to encourage animal movement) may be equally viable al ternatives for maintaining connectivity. Patch occupancy patterns provide further in sights into populationtrajectories that may result from ongoing fragmentati on in other parts of the biom e, and the sustainability criteria were used to identif y a range of landscape configur ations in which persistence was highly unlikely without conservation ac tion, but where management to enhance landscape connectivity could significantly increase l ong-term persistence.
1 CHAPTER 1 INTRODUCTION In landscapes where formerly contiguous habitat is fragmented, conservation strategies rely on sufficient levels of move ment among habitat patches to rescue small populations from eminent extinction, or to permit recolonization once extinction has occurred (Brown & Kodric-Brown 1977; Hans ki 1998). Thus, maintenance of landscape Â“connectivityÂ” (availability of movement r outes among landscape elements) has become a major focus of conservation planni ng (Forman & Godron 1986; Mann & Plummer 1993; Rosenberg et al. 1997). Landscape conne ctions may include corridors of natural vegetation that connect larger vegetated areas (Forman & Godron 1986), or patches of habitat types through which organisms are physically and behaviorally capable of dispersing. However, it is not immediately clear what actio ns are needed to maintain connectivity, since factors that influence anim al movement are high ly species-specific, and data on movement are lacking for most species. This dissertation examined effects of fo rest fragmentation on movement of the Chucao Tapaculo ( Scelorchilus rubecula ), an endemic understory bird of conservation concern in the South American temperate rain forest biome. Probability of inter-patch movement was assessed directl y, through translocati on experiments, and indirectly, by analyzing observed patterns of patch occupa ncy at the landscape-scale in two alternative landscapes. Predictive patch-o ccupancy models and sustaina bility criteria were also developed and tested in alte rnative landscapes. This intr oductory chapte r provides the
2 theoretical basis for the disse rtation, an introduction to the study system, and an overview of the research program. Theoretical Background Broad underlying hypotheses for this work include the following: 1. occupancy of habitat patches is influen ced by immigration rates (Brown & KodricBrown 1977; Hanski 1998; Ti schendorf et al. 2003) 2. these rates differ among patches dependi ng on the permeability /connectivity of intervening landscape elements (Wiens 1994; Haddad 1999a; Roland et al. 2000; Tischendorf & Fahrig 2000; RodrÃguez et al. 2001) 3. permeability/connectivity is influenced by individualsÂ’ behavioral resistance to entering and moving through particular la ndscape elements (Stamps et al. 1987; Lima & Dill 1990; Sieving et al. 1996; Rail et al. 1997; St. Clair et al. 1998; Grubb & Doherty 1999; Haddad 1999b; RodrÃguez et al. 2001; BÃ©lisle & Desrochers 2002) 4. resistance varies among landscape elemen ts depending on vegetation structure and the distance that must be crossed (Sta mps et al. 1987; Machtans et al. 1996; Desrochers & Hannon 1997; Rail et al. 1997; St. Clair et al. 1998; Grubb & Doherty 1999; Roland et al. 2000; Ricketts 2001; RodrÃguez et al. 2001; BÃ©lisle & Desrochers 2002; Bakker & Van Vuren 2004) To date, conservation in fr agmented landscapes has focused largely on protection or restoration of vegetated corridors, wh ich are thought to provide passageways for movement among otherwise isolated patches (Forman & Godron 1986; Mann & Plummer 1993; Fahrig & Merriam 1994; Rosenbe rg et al. 1997). The corridor concept stems largely from principles of Isla nd Biogeography Theory (IBT), pioneered by MacArthur and Wilson (1967), which proposed a dynamic equilibrium whereby numbers of species on islands were determined by rates of immigration and extinction. Though originally developed in reference to oceanic islands, principles of IBT have also been applied to Â“habitat islandsÂ” in fragmented terrestrial systems. Using IBT as a foundation, a series of prin ciples were derived for the design of nature reserves, with the goal of maximi zing numbers of species conserved (Diamond 1975). Among these was the suggestion that is olated patches (or reserves) would support
3 higher numbers of species if they were c onnected by movement co rridors, based on the assumption that immigration is important fo r maintenance of species richness. The corridor concept has been hotly debated in the conservation literature (Hobbs 1992; Simberloff et al. 1992; Beier & Noss 1998; Haddad et al. 2000; Noss & Beier 2000; Proche et al. 2005), in part because corridor e ffects have been difficult to study using large-scale controlled and repli cated experimental designs. Although numerous studies have shown that corridors function as movement routes for some species (e.g., Beier 1995; Haas 1995; Sutcliffe & Thomas 1996; Aars & Ims 1999; Brooker et al. 1999; Sievi ng et al. 2000; Berggren et al. 2002; Haddad et al. 2003; Levey et al. 2005a), other studies show am biguous results (Gustafsson & Hansson 1997; Rosenberg et al. 1997; NiemelÃ¤ 2001). Furthe r, potential negative consequences of corridors have been suggested. Of particul ar concern are indirect effects on species competition and predator-prey interactions, enhanced spread of invasive species, and the opportunity costs associated with corridor establishment, which may preclude other conservation approaches (Hobbs 1992; Rosenbe rg et al. 1997; Simberloff et al. 1992; NiemelÃ¤ 2001; Tewksbury et al. 2002; Orrock & Damschen 2005; Proche et al. 2005). Despite the intuitive appeal of the corrido r concept, presence of a corridor between patches may not be necessary to maintain c onnectivity if movement through the matrix is high (Tischendorf & Fahrig 2000). Several auth ors have pointed out that, unlike islands surrounded by water, the matrix surrounding terrestrial habitat islands may not be entirely inhospitable to dispersing organisms (For man & Godron 1986; Wiens 1994; Roland et al. 2000; Ricketts 2001). In reality, the matrix likely functions as a Â“selective filter,Â” allowing movement of some species but i nhibiting others (K ozakiewicz 1993).
4 Furthermore, the matrix itself is heteroge neous, and various habitat components within the matrix will be differentially permeable in a species-specific manner. Therefore, the critical question regarding corridor efficacy in this regard, as defined by Harris and Scheck (1991), is whether a system of hab itat fragments will function better for long-term conservation with corridor c onnections than without. An swering this question will require documentation of behavioral responses to both corridors and habitat types that dominate the landscape matrix. Given the di fficulty and considerable expense of creating and restoring corridor networks, more work is needed to determine the circumstances under which corridors are appropriate for mee ting conservation needs. In many cases, it may be more feasible to improve connectiv ity through management of vegetation in the matrix to encourage animal movement, rather than reconne cting patches with corridors (Franklin 1993; Bowne et al. 1999; Vandermeer & Carvajal 2001). For some forest birds, open matrix may function as a barrier (Sieving et al. 1996; Desrochers & Hannon 1997; Rail et al. 1997; St. Clair et al. 1998; Grubb & Doherty 1999; RodrÃguez et al. 2001; BÃ©lisle & Desr ochers 2002) due to increased risk of predation (Lima & Dill 1990; Suhonen 1993; R odrÃguez et al. 2001), limitation on the distance at which individuals perceive other forest patches across the matrix (Lima & Zollner 1996), or lack of necessary resources outside forested areas. Nonetheless, if conditions permit movement without unacceptable risk or energetic costs, it is possible that many forest species will disperse th rough partially vegetated or otherwise suboptimal matrix habita t types. Since the conditions that permit movement by forest species are currently unknown, informati on on movement behavior would greatly advance our capacity to manage landscapes for conservation (Turchin 1998).
5 Study System South American temperate rainforest is a global hotspot for ecosystem destruction that threatens endemic species (Glade 1988; Collar et al. 1992; Balmford & Long 1994). At present, 45% of the original forest cover has been lost, and pressures driving deforestation are stronger than ever. Study areas used for this research included three fragmented agricultural landscap es in south-central Chile. One study site was located on northern ChiloÃ© Island, near Ancud (41o55Â’ S, 73o35Â’ W), one was located on the mainland near Puerto Montt (41o28Â’ S, 73o00Â’ W), and the other was located on the mainland near Osorno (40o35Â’ S, 73o05Â’ W). On ChiloÃ©, large-scale forest fragmentation occurred relatively recently (within the last 50 Â– 100 y ears; Willson & Armesto 1996), and the landscape is now dominated by smallscale subsistence farming. ChiloÃ© appears relatively well connected in that most forest fragments are linked by corridors or isolated by 100 m of non-forest habitat. The Osorno st udy area, in contrast , has both a greater extent and longer history (100150 years) of human-induced disturbance (Smith-Ramirez unpublished data), and was dominated by larg e-scale intensive agriculture throughout much of the last century. The Puerto M ontt study area is intermediate between the previous two. In South American temperate rainforest, bi rd species of the family Rhinocryptidae (commonly known as tapaculos) are among the most sensitive to forest fragmentation (Willson et al. 1994). These include the C hucao Tapaculo, Black-throated Huet-huet ( Pteroptochos tarnii ), Ochre-flanked Tapaculo ( Eugralla paradoxa ), and Magellanic Tapaculo ( Scytalopus m. magellanicus ). All are understory insectivores that forage primarily on the forest floor , and are rarely observed outsi de forest habitat. Their sensitivity to fragmentation is most like ly due to poor flying ability, coupled with
6 behavioral resistance to entering open habita t (Sieving et al. 1996). This sensitivity makes tapaculos good candidates as focal spec ies to represent the habitat connectivity component of landscape design because landsca pes that provide functional connectivity for tapaculos would probably meet the m ovement requirements of most forest vertebrates. Previous research provided strong evidence that landscape connec tivity is enhanced for tapaculos by availability of wooded corridor s. Using tape-recorded calls to elicit territorial responses, Sieving et al. (2000) s howed that tapaculos could be induced to enter and move 100-200 m in narrow (< 25 m) wooded corridors embedded within a matrix of open habitats such as pasture. Understory vegetation density was the dominant predictor of corridor us e by all four species, and corridors with small streams and steep banks were favored by Chucao Tapaculos. For all tapaculos, barriers to movement appear largely due to behavioral responses to differences in habitat structure that influence birdsÂ’ propensity to travel through va rious matrix types. For many bird species, escape from attacking predators requires a qu ick dash into vegetative cover, which may provide a physical or visual impediment to pr edators. Numerous studies have shown that bird species reliant on escape cover for predat or avoidance are relu ctant to venture far from vegetation (Lima & Dill 1990; Todd & Cowie 1990; RodrÃguez et al. 2001). Species such as tapaculos that are poor flyers and cannot use sudden flight or other aerial escape tactics may be particul arly reliant on vegetative cove r to avoid predators. For tapaculos, aerial predators are a serious thre at in open habitats where attacks have been observed (Sieving et al. 2000) and this is probably an important factor influencing movement decisions.
7 The Chucao Tapaculo was identified as the best subject for intensive research because they are locally abundant, more easil y captured than other tapaculos, and not currently threatened wi th extinction. The Chucao is a mid-sized tapaculo (40 g) that rarely flies more than a few meters when fl ushed, and usually nests in cavities 1-4 m off the ground in trees, stumps, logs, or earthen banks. Like other tapaculos, Chucaos are strongly associated with understory vegetation (Johnson 1965, 1967; Ried et al. 2004), and are reluctant to enter open ha bitat (Sieving et al. 1996). Several lines of research on Chucaos point to dispersal limitation as a leading cause of regional declines. Chucao reproductive success is relativel y high in forest fragments, with 63 % of pairs fledging at least one young per clutch and of producing two or three clutches per season (De Santo et al. 2002). Nestling growth is similar in fragments and continuous forest, suggesting that fragme ntation does not limit food resources (e.g., Thessing 2000), and there are no documented ed ge-related effects on nest fate, nestling growth, or early juvenile survival (De Santo et al. 2002). Nest site limitation is indicated by reuse of sites in forest fragments. Howeve r, Chucaos nest in a wide variety of cavities or semi-enclosed niches, and sometimes build domed or open-cup nests on the ground or in small trees, so it is unlikely they are si gnificantly constrained by nest site limitation (Willson 2004). In patches isolated by open habitat, probability of Â“apparentÂ” natal emigration (i.e., disappearance of banded juveniles from the na tal patch) was lower than in connected patches. In isolated patches, 21 % of indi viduals banded as nestli ngs were sighted in their natal patches the following year, compar ed to 3 % in connected patches. This pattern would be expected if a higher proportion of juven iles were constrained from
8 leaving natal patches surrounded by open matri x, compared to those in natal patches connected by corridors. Further, pairing succes s declines in isolated patches, indicated by a preponderance of unmated males (18 %) in isolated patches, compared to 1 % in connected patches (Willson 2004); a finding common among fragmentation sensitive species (e.g., Villard et al. 1993). This pr esumably occurs because juvenile females (normally the most dispersive sex in bird s; Greenwood 1982) leave their natal patches (perhaps to avoid inbreeding), while patch isolation reduces the flow of immigrating females back into the patches. This research indicates that that fragmentation-sensitivity for the species is caused not by reproductive failure but, rather, by l ack of access to breeding sites located in impermeable portions of the landscape mosaic . However, strong inference requires a direct test of the proposed mechanism that m ovement constraint reduces immigration into isolated patches. This dissert ation research critic ally tested hypothese s related to Chucao movement through various matrix compone nts and provided evidence for populationlevel effects of constrained movement. The research targeted specific holes in knowledge regarding ChucaosÂ’ behavioral re sponses to commonly occurring components of actual landscapes where the conservation efforts will be focused, and provides urgently needed information for conservation planning. Research Overview The goal of this dissertation was to id entify factors enhanc ing/impeding landscape connectivity for the Chucao Tapaculo and to assess potential for long term persistence in landscapes that differ in habitat area and c onfigurations. Phase I was a translocation experiment designed to directly test hypothe ses regarding habitat permeability. Phase II included landscape-scale surveys of patch o ccupancy in two landscapes that differed in
9 level and history of forest fragmentation, a nd development of predictive patch occupancy models. In Phase III, I developed a set of landscape criteria for identifying patch configurations with reas onably high potential for s upporting sustainable Chucao populations; then I applied these criteria in test landscapes to a ssess their conservation status and the potential benef its of conservation action to en hance landscape connectivity. Strengths of the research de sign include the following: (1 ) it addressed movement in wooded corridors and common matr ix habitat types; (2) behavioral responses to varied landscape structure were tested directly through translocation experiments; (3) the subject was a fragmentation sensitive vertebrate species; and (4) tests were conducted at spatialscales appropriate to the subjectÂ’s vagi lity and relevant cons ervation questions; and (5) generality of predictive patch models wa s tested in alternativ e landscapes differing characteristics and histories. Phase I, presented in chapter 2, was a tran slocation experiment to directly test permeability of wooded corridors and two comm on matrix habitat-types (shrub fields and pastures). The a priori expectation was that vegetate d-corridors and dense secondary vegetation in the matrix would encourage C hucao movement. Thus, I predicted that translocated individuals would disperse more quickly from release patches that adjoined forest corridors, or were embedded in vege tated matrix, compared to patches surrounded by pasture. The objective of Phase II, presented in ch apter 3, was to analyze patch occupancy patterns in two landscapes that differed in level and duration of fragmentation and to develop models for predicting fragmentation e ffects in landscapes at different stages of the deforestation process. Although previous studies indicated that Chucaos are reluctant
10 to enter certain habita t types (Sieving et al. 1996), it was not clear how significant these potential barriers to movement might be for long-term persistence at a regional scale. Within the complexes of forest patches in the two landscapes, absen ce of Chucaos from a given patch was considered indicative of patch characteristics that limit long-term population viability, and characteristics of the surrounding matrix that impede immigration/recolonization. In Phase III, presented in chapter 4, I developed a set of criteria to identify patch configurations wherein Chucao populations would have a rela tively high expectancy of long term sustainability. I then applied the criteria in 100 km2 study areas from three test landscapes at differing levels of fragmentati on. Based on this analysis, I was able to differentiate among landscape c onfigurations in which pers istence was likely without conservation intervention, conditions where ex tinction was probable regardless of efforts to enhance connectivity, and conditions wh ere restoration of connectivity could significantly increase the potenti al for long term regional persistence. Chapter 5 provides an integrated summary of results from the th ree research phases. This research provides urgently needed data for conservation deci sion-making in fragmented South American temperate rainforest, and addresses the impor tant issue of differe ntial permeability of matrix habitats, which has been largel y ignored in fragmentation research.
11 CHAPTER 2 AN EXPERIMENTAL TEST OF MATRIX PERMEABILITY AND CORRIDOR USE Introduction To offset effects of habitat fragmenta tion, maintenance of landscape connectivity has become a major focus of conservation pl anning, and provision of movement corridors is currently a favored approach (e .g., Desrochers & Hannon 1997; Haddad 1999a; Berggren et al. 2002). However, research has largely ignored the degree to which animals move through the matrix of non-prefer red habitat (Beier & Noss 1998; Ricketts 2001; Hudgens & Haddad 2003). For many specie s, the matrix constitutes unsuitable and potentially Â“hostileÂ” habitat (A rendt 2004), but it is rarely a complete barrier to movement. In some cases, movement through the matrix may be sufficient for immigration to offset extinction of local (sub-)populations (Witt & Huntly 2001; Hudgens & Haddad 2003), and distinct habita t types within the matrix (defined by vegetative and other structural features) may be differentially permeable to a variety of species (Ricketts 2001; Ries & Debinski 2001; RodrÃguez et al. 2001). Understanding how habitat structure in the matrix influe nces permeability to animal movement is essential to managing complex landscapes for conservation (Turchin 1998; Ricketts 2001; Vandermeer & Carvajal 2001). Experimental methods are the most efficient means for identifying causal mechanisms, but most experimental studies addressing animal movement have used invertebrate subjects, often in highly artificial experimental landscapes (e.g., Berggren et al. 2002; Hein et al. 2003). To date, experime ntal methods applied to vertebrate subjects
12 include use of tape-recorded songs to provide a stimulus for forest birds to enter wooded corridors or cross open habitat gaps (e.g., Siev ing et al. 1996; St. Clai r et al. 1998; BÃ©lisle & Desrochers 2002), and translocation of small mammals and birds to assess gap crossing decisions (Bright 1998; Bowman & Fahrig 2002) and movement paths in the context of homing behavior (Bright 1998; BÃ©lisle & St. Cl air 2001; Bakker & Van Vuren 2004). Translocation has also been used to assess migration rates among patches with and without corridors, relative to a single high-contrast matrix type (Bowne et al. 1999), and to assess homing time in landscapes with differing percentages of open matrix (BÃ©lisle et al. 2001; Gobeil & Villard 2002). From these studies we know that a variety of forest vertebrates are clearly averse to entering open habitat, that wooded corridors likely facilitate movement for some species , and that homing behavior may be impeded in landscapes dominated by open matrix. Although these studies have c ontributed greatly to deve lopment of the dispersal barriers and landscape connectiv ity concepts, they fail to distinguish the relative importance of corridors versus alternative matrix types and many rely on assumptions that hinder applicability to natural dispersal. Playback experiments, for example, can yield detailed observations on specific move ment choices, but they may have little relevance to dispersal because movement in re sponse to perceived territorial intruders or predators may differ from disper sal. Similarly, homing behavior of subjects translocated short distances misrepresent dispersal in that the landscape near a subjectÂ’s home range is familiar, whereas dispersing individuals must respond to novel landscape mosaics and unfamiliar risk environments (Yoder et al. 2004 ). In contrast, long-range homing studies may generate excellent approximations of dispersal but often lack the level of
13 observational detail to detect specific movement choices (e .g., the decision to cross an area of matrix or detour thr ough a corridor or alternative ha bitat type). Thus, direct comparison of movement by dispersing indivi duals through corridors versus alternative matrix habitats is still needed (Nicho lls & Margules 1991; In glis & Underwood 1992; Simberloff et al. 1992). I used radio-telemetry to monitor moveme nts of translocated Chucao Tapaculos ( Scelorchilus rubecula ) to compare the permeability of wooded corridors relative to two common matrix habitat types (open pastur e and shrubby secondary vegetation). The experimental design avoided p itfalls associated with pass ive observation, including lack of control over landscape composition and lack of detail regarding specific movement choices. Yet empirical realism remained high because subjects were wild-caught individuals released into very small pa tches (inadequate for breeding), stimulating movements typical of birds searching for ne w territories (Ims 1995). Furthermore, I translocated subjects far enough (5 Â– 9 km) from their capture s ites to prevent homing and eliminate bias due to pr evious knowledge of the lands cape (Yoder et al. 2004). To my knowledge, this is the first study to experi mentally test permeability of matrix types differing in vegetation structure relative to m ovement corridors using a vertebrate subject. Methods Study System South American temperate rainforest o ccupies a narrow zone between 35Â°S and 55Â°S in Chile and western Argentina, and is considered a global hotspot for endemic species loss (Balmford & Long 1994; Davis et al. 1997; Stattersfield 1998). Of forest birds, endemic understory insectivores in the family Rhinocryptidae (tapaculos) are among the most sensitive to fragmentation in the biome (Willson et al. 1994). Tapaculos
14 are primarily terrestrial species that are poor flyers and stro ngly associated with dense forest-understory (Reid et al. 2004). Although reluct ant to enter open ha bitat, they will use narrow wooded strips for movement am ong forest patches (Sieving et al. 1996, 2000). This sensitivity to movement habita t makes the group potentially valuable as focal species for planning the connectivity component of landscape design, because a landscape that provides func tional connectivity for the group would probably meet the movement requirements of many forest species. I identified the Chucao Tapaculo as the be st subject for experimentation because it is locally abundant and is intermediate in size and vagility among the four tapaculos (Sieving et al. 2000). Chucaos are year-round re sidents that are strongly territorial and usually retain the same breeding territory ye ar after year (De Sa nto et al. 2002; Willson, unpublished data). They are very poor fl yers and move by walking, hopping, and flying short distances (no more than a few meters), usually within or near dense vegetative cover. The study was conducted in a fragmented agricultural landscape on northern ChiloÃ© Island, Chile (41o55Â’ S, 73o35Â’ W). Pastures and abandoned agricultural fields dominate the landscape, with wooded ha bitat covering approximately 35% of the total study area (approx. 300 km2). At present, many forest patc hes remain interconnected by linear vegetated strips including fencerows, ripari an draws, and ravines. Dominant trees include broad-leaved evergr eens and a few conifers ( Nothofagus nitida [Phil.] Krasser, Drimys winteri J. R. et G. Forster , Eucryphia cordifolia Cav., and Podocarpus nubigena Lindl.), with understory compos ed principally of bamboo ( Chusquea valdiviensis Phil.) and saplings (Donoso 1993). Secondary ve getation in many abandoned agricultural
15 fields is dominated by Baccharis magellanica (Lam.) Pers., a persistent shrubby invader of poorly drained soils, with thick mats of Sphagnum spp. moss covering the ground. Experimental Design Subjects were captured in large forest tr acts, fitted with radio transmitters and released individually (each subject tested onc e) into small wooded patches at the centers of experimental landscapes (descriptions follo w). Release patches were unoccupied by con-specifics and large enough to meet the subjectsÂ’ imme diate requirements for food and shelter, but too small (< 0.30 ha) to serve as adequate breeding territories ( 1 ha; De Santo et al. 2002). This appr oach provided a standardized stimulus for rapid movement from release patches, allowi ng direct observation of movement paths. I interpreted delayed movement of a subject from the re lease patch as a meas ure of behavioral resistance to entering and moving through the matrix element(s) presented. The experimental treatments consisted of re lease patches that were either entirely surrounded by open habitat, entirely surrounde d by dense shrubs, or linked to other patches by wooded corridors embedded in open ma trix (Fig. 2-1), but within 150 m of at least one suitable habitat patch. The fourth block of the experimental design, forest patches with wooded corridors embedded in a shrub matrix, was not included because this configuration did not exist in the study landscape. The Â“openÂ” and Â“shrubÂ” matrix types consisted, respectively, of pastures and shrubby vegetation dominated by 1 to 2-mtall B. magellanica . Wooded corridors adjoining release patches in the corridor treatment were either continuous or had breaks in the vegetation of 2 m and were otherwise surrounded by open pasture. Each corridor was approximately 10 m wide, but margins were not perfectly linear and some corridors narrowed or expanded within a range of 2-
16 15 m at various points. Lengths of corridors (to the nearest adjoining patch) varied from 60-500 m. In general, corridor vegetation cons isted of 2 to 3-m-tall native trees, with occasional sections of lower stature vege tation, and understory dominated by saplings and bamboo. Corridors of these dimensions we re expected to function only as conduits for movement rather than providing habitat fo r long-term survival or breeding (Sieving et al. 2000). Selection of Release Sites Replicate release sites used in each tr eatment were chosen from the existing landscape through assessment of aerial photogra phs and site visits. I standardized replicates as much as possible with regard to release patch area, habitat quality, and landscape context. Each release patch wa s surrounded by only one matrix type, with similar ranges among treatments for potential ly important landscape-context factors (distance to the nearest patch large enough to support a br eeding territory [ 1 ha], distance to the nearest patch of any size, a nd the percentage of w ooded habitat within a 100-m buffer). I used a total of 25 sites (8-9 replicates per treatment), with a maximum of two trials conducted at each site. Repeated use of sites was necessary due to scarcity of locations with appropriate characteristics. However, si tes were never reused until I was certain the previous subject was no longe r present. Although release patches were free of conspecifics, all but the most isolated patches 3 ha in the surrounding landscape were occupied (Chapter 3). Despite my best efforts to minimize differe nces among release patches, the corridor sites available in the study landscape tended to be less disturbed than openand shrubmatrix sites. Therefore, some release patc hes in the corridor treatment were slightly larger and of better quality than in other tr eatments. Conversely, release patches in the
17 open matrix (pasture) treatment were of lowe st quality due to effects of livestock. Because I expected the stimulus for movement to be strongest when habitat quality was low (e.g., Buddle & Rypstra 2003), use of lo wer quality release patches (stimulating quicker movement) in the open matrix treatm ent made my analysis conservative with regard to the a priori expectation of longer movement delays for this group. Capture and Handling Chucaos were captured during the breeding seasons (2000-2002) with walk-in traps baited with earthworms. I marked each subjec t with colored plastic leg bands and a small radio transmitter (1.3-1.7 g, approximately 3-4% of adult body mass). I attached transmitters with a skin safe epoxy (ARC 5; Composite Technology, Stonham, Massachusetts), formulated to cure rapi dly under low temperature, high humidity conditions. The transmitters were bonded to th e tops of tail feathers (rectrices), at the base of the feathers near th e quills, positioned to avoid cont act with the uropygial gland. Because Chucaos do not use long-distance flight for travel, foraging, or predator avoidance, mounting transmitters onto tail feathers was appropriate. I captured subjects between sunrise and 1100 hours, and held them a maximum of 2.5 hours from the time of capture to release. Each subject was tran sported individually in an opaque container that prevented visual assessment of the surroundings. Prior to release, each subject was left undisturbed inside the container (provisioned with earthworms as food) at the release site for a period of 15 minutes to minimize the influence of stress from handling and transpor t on movement behavior. Also to minimize disturbance, the observer conducted the re lease from a concealed location by pulling a long string to remove a section of the containerlid. This allowed the subject to exit in an upward direction, preventing bias in the post -release movement path. Once the subject
18 exited the container, the observe r left the area as quickly a nd quietly as possible. To prevent temporal bias among treatments, a rel ease was conducted in one replicate of each treatment prior to conducting a subsequent release in any of the previous treatments. It was not possible to standardize for sex or age among subjects because data on plumage characteristics are not available to readily distinguish sexes (laparoscopy was deemed too invasive), or between older juvenile s and adults. This la ck of standardization made my study conservative because juveni le and female birds are typically more dispersive than adults and males (res pectively, Greenwood & Harvey 1982; Johnson & Gaines 1990). I was also unable to standardi ze among subjects that we re territory holders versus floaters at the time of capture. Nonethel ess, I expect that all subjects, regardless of their territorial status, were strongly motivat ed to disperse from the unsuitably small release patches. Radio Telemetry Each corridor adjoining a release patch was monito red continuously (during daylight hours) by a telemetry operator stati oned in a nearby concealed location or by an automated receiver and data-logger (m odels R2100 and DCC D5041; Advanced Telemetry Systems, Isanti, Minnesota). I pl aced a 25-cm omni-direc tional antenna at the center of each corridor (at the midpoint lengthwise) and se t the telemetry reception range to the corridor width by adjust ing the receiver gain. This configuration ensured that presence of the subject would only be reco rded when it passed th rough the corridor. If the subject dispersed from the release patch but was never recorded inside the telemetry reception area, it was assumed that movement occurred through the open matrix in which the release patch and corridor were embe dded. This design was an improvement over
19 many previous studies that assumed corrido r use without addressing the potential for movement through the matrix (Simberl off et al. 1992; Beier & Noss 1998). In addition to continuous monitoring of corridors, each subject was located once daily by an observer on foot with a hand-held receiver and directional antenna. To avoid disturbance to subjects, the observer a pproached only close enough to identify the occupied habitat patch, and then left the ar ea. If no movement was observed over a 3-day period, the observer approached close enough to obtain visual confirmation that the transmitter was attached and the subject wa s alive. Initial movement was defined functionally as movement from an occupi ed patch to any other landscape element (usually another wooded patch, but movements into the shrub matrix were observed). Because release patches in non-corridor treatments were each surrounded by only one matrix type (either open habitat or shrubs), I assumed that any s ubject found outside the release patch must have crossed the associat ed matrix, moving a distance equal to or greater than the minimum distance to the ne arest neighbor patch (not necessarily the patch occupied at the time of observation). However, no assumptions could be made about the actual travel path in the matrix be cause any number of routes could have been followed to reach the observed location. I used data from continuous monitoring of corridors only to document corridor use, whereas the number of days each subject rema ined in the release patch was determined based on daily telemetry surveys. This dis tinction was important because some subjects made excursions into a corridor (detected vi a continuous monitoring) but returned to the release patch prior to being observed outside the patch during the daily survey. Under these circumstances, use of data from corrido r monitoring to establish time to initial
20 movement would have been inappropriate because non-corridor treatments were not monitored with equal intensity, and similar exploratory movements into the matrix may have been undetected. Therefore, use of th ese data would have biased time to initial movement estimates, making it appear that movement occurred more quickly in the corridor treatment. Once subjects dispersed outside the experi mental areas (each with only one matrix type) I had no control over the landscape conditions they encount ered, and there were many factors I could not monitor that pot entially influenced movement decisions, including intra-specific interactions, terr itory vacancies, and habitat quality. Thus, monitoring of movements outside experiment al areas was entirely descriptive and interpretation was necessarily very conservative. Given an observed movement from one documented location to another, my qualitati ve analysis addressed only the following measures: the minimum distance across the matr ix the subject must have moved to reach the new location (i.e., the minimum distance from the occupied patch to the nearestneighbor patch, making no assumptions about th e actual travel path), the area of each visited patch, the number of times each patch was visited by the same subject, and the linear distances between observe d locations. While these measures were basic, they nonetheless provided previously unavailable an ecdotal information with potential value for conservation planning. As constraints permitted, monitoring of each subject was continued until the transmitter failed ( 30 days) or detached, or until a s ubject settled in a new patch for 2 weeks. On a few occasions monitoring was te rminated early for s ubjects that dispersed into large roadless areas where tracking wa s impractical. Also, some signals were
21 permanently lost for unknown r easons. When this occurred, I searched intensively by driving and walking within a search radius of several kilometers from the last known location. Attempts were made to locate the lost subject for 3 days, then I searched opportunistically as time permitted. Measurement of Landscape Metrics I used remote sensing and geographic information system (GIS) analysis to quantify landscape metrics. Scanned pa nchromatic orthophotogr aphs (1:20,000), taken during January 1994, were georeferenced and join ed to create a sing le digital orthophotomosaic with 2-m2 pixel resolution. I hand digitized wooded habitat cover (old growth and secondary combined) at a minimum mapping unit of 10 m2 in ArcView 3.2 (ESRI 1999). Photo-interpretation was ground truthed extensively in the field during site visits for telemetry monitoring, and significant cha nges in land-cover since acquisition of the photographic data were correct ed on the digital maps. I plotted radio telemetry location data on the digital map and estimated movement distances by measuring the minimum straight -line distances between daily locations for each subject and the maximum displacement of each subject from its release patch. In addition, I measured the orientation-angle of each subjectÂ’s move ment path (from the release patch to the last recorded location) relative to the center of its capture site. Landscape context metrics were calculated using ArcView and FRAGSTATS version 3.3 (McGarigal et al. 2002). These included release patch area, di stance to the nearest neighbor patch (of any size and 1 ha), percentage of wooded habitat within a 100-m buffer, and the area of each visited patch. The 100-m buffer radius was selected because the hypothesized relationship between landscap e context and a subjectÂ’s decision to disperse presumed that the subject was capab le of evaluating the surrounding landscape.
22 Because I had no independent data on Chucao perceptual range (Lima & Zollner 1996), I conducted a preliminary analysis of patch occu pancy data (Chapter 3) and selected the smaller of the two buffer distances (100 a nd 300 m) wherein landscape context variables were identified as significant predic tors of Chucao occupancy. Data Analysis The translocation experiment assessed the num ber of days birds remained in release patches prior to initial movement, which wa s interpreted as a m easure of subjectsÂ’ reluctance to disperse, and conversely, resistan ce of the presented landscape elements to movement. I used Cox Regression to comp are this response among treatments. Cox regression compares survival curves (surviva l = time elapsed prio r to occurrence of a terminal event) among treatment groups. In th is experiment, moveme nt of a subject from the release patch was treated as the terminal event, whereas remaining in the patch was analogous to survival. Survival analysis wa s appropriate because it permitted use of censored (i.e., incomplete) data collected on su bjects that died or lost their transmitters prior to movement. Predictor variables incl uded the landscape treatment (corridor, open matrix, and shrub-matrix treatments) and the set of landscape context variables described in the previous paragraph. I include d the interaction term (treatment distance to the nearest patch 1 ha) because a more pronounced effect was expected for the open matrix treatment, which was presumed mo st resistant to movement. Model fitting was conducted using forwar d-stepwise likelihood-ratio estimation (Harrell 2001). At each step of model building, I added the variable that produced the most significant ( p 0.05) change in the model chisquare (equal to the difference between the -2 log-likelihood of the model at the previous step and the current step).
23 Then, to arrive at the most parsimonious set of predictors, I independently removed variables already in the model and calculated the change in the chi-square. If the change was not significant ( p 0.10) I removed the specified vari able. This iterative process was continued until no more variables coul d be added or removed. I determined significance of differences among treatments w ith linear contrasts. Alternative model building approaches (backward stepwise and singl e-step entry of all va riables) were also tested to ensure that conclusions were not dependent on the model selection procedure. To assess potential loss of data indepe ndence caused by repeated use of some release sites, I conducted a prel iminary analysis of data from these sites in which site identity was entered as a categorical factor. Because site identity was not a significant predictor of days to initial movement, and si nce each individual subject was tested only once, each trial was treated as an independent st atistical unit in the final analysis. Prior to analysis, I calculated Pearson correlation coefficients for pair-wise comparisons among landscape variables to ensure th at none were strongly correlat ed (i.e., r > 0.7). Finally, to identify any potential influence of homing be havior on movement direction I assessed the orientation angles of movement paths relativ e to subjectsÂ’ capture sites with a V test (Batschelet 1981). All statistical tests were performed at a 95% confidence interval in SPSS 11.0.1 (SPSS 2001) and Oriana 2.01c (for ci rcular statistics; Kovach Computing Services 2004). Results Forty-one Chucaos were translocated, a nd 558 locations were obtained during daily telemetry surveys. Thirteen subjects were released into replicates of the corridor treatment, and 14 subjects each were released into the openand shrub-matrix treatments. Of the 41 subjects, 33 (78%) dispersed succe ssfully from release sites, including all
24 subjects in the shrub-matrix treatment and 11 subjects in the corridor treatment (Table 21). Data for the two remaining subjects in th e corridor treatment were censored (7-8 days after release). Results for the open-matrix treatment were more variable. Of the 14 subjects, 7 dispersed successfully, 4 were cen sored (within 5-7 days), and 3 remained in the release patch for the duration of the 30-day monitoring period. Landscape treatment was the only significant (Wald2 = 7.55, p = 0.02) factor predicting time to movement in the fo rward-stepwise Cox regression (model fit; 2 2 = 8.30, p = 0.02), and treatment was cons istently identified as the most important factor in the backward (Wald2 = 8.02, p = 0.02) and full model (Wald2 = 8.47, p = 0.01) analyses. Thus, for simplicity, only results for the forw ard stepwise method are discussed. Contrast results showed that mean (Â± S.D.) time to initial movement was significantly longer for subjects in the open matrix treatment (10.29 days Â± 11.23, Fig. 2-2) th an in the corridor (3.38 days Â± 2.93, Wald1 = 3.75, p = 0.05) and shrub matrix (2.93 days Â± 4.51, Wald1 = 7.59, p = 0.01) treatments, whereas time to move ment was similarly short in the latter two treatments (Wald1 = 0.67, p = 0.41). The V test showed no clear bias in travel paths toward the initial capture sites (u30 = 1.40, p = 0.08). For illustrative purposes, the movement path of subject C1a is presented in Fig. 3, and results for the remaining subjects are summarized in the following paragraphs. Subject C1a, released at loca tion (loc. 1), dispersed via the corridor (loc. 2) to a 0.5-ha patch (loc. 3) the day following release. The linear distance to loc. 3 and the distance via the corridor were approximately 90 m each. The following morning the subject returned to the corridor (loc. 4) but was found later the same day in a 2-ha patch (loc. 5) that adjoined the corridor. Locations 4 and 5 were 360 m apart, but the r oute via the corridor
25 was approximately 460 m. Subsequently, the subj ect returned to the co rridor (loc. 6) but was found again the following day in the 2-ha patch (loc. 7), where it remained for 3 days. It made a 1-day excursion to a small (< 0.5-ha) linear patc h (loc. 8), crossing a 20m open-matrix gap, and then returned to the 2ha patch (loc. 9). Ne xt, it was found in the corridor (loc. 10) about 200 m from the prev ious location, which was a 230-m path via the corridor route. Finally, on the tenth day, the detached transmitter was found in the 0.5-ha patch (loc. 11) that was first visited the day after release. Based on these data it cannot be definitively concluded that the corr idor was always used as a travel route among the adjoining patches. However, the subject was never observed outside wooded habitat and it was repeatedly detected inside the corridor dur ing periods between observations in patches. Therefore, I assume that the majority of movements occurred via the corridor, although the s ubject made two documented moves 20 m across the open matrix. Eleven of the 13 subjects in the corri dor treatment dispersed from the release patches via corridor routes. One of the rema ining subjects (C4b) moved partially down the corridor, within range of the receiver, th en took a shortcut across the open matrix to reach the second patch. The minimum distance of the shortcut was approximately 50 m, whereas the corridor route was approximatel y 150 m. Subject C6b also dispersed 25 m across the open matrix instead of using th e 185-m corridor, but moved into the corridor from the second patch later the same day. The mean confirmed distance traveled (Â± S.D.) by subjects through corridors was 351 m (Â± 236), which ranged from 1-9 times the lengths of the non-corridor routes. The mean corridor-to-linear distance ratio was 2.66 (Â± 2.22) times longer for corridor routes.
26 Data on movement distances across the open and shrub matrix were more uncertain because the matrix could not be monitored continuously, and any number of paths could have been taken. Thus, I could only measure minimum distances across the matrix that I knew subjects must have crossed to reach th e observed locations. Mean distances were 56 m (Â± 27) for the open matrix and 100 m (Â± 45) for the shrub matrix, although many travel paths were probably considerably longer. Regardless of intervening ha bitat, the mean linear dist ance traveled within a 24hour period was 170 m (Â± 225), and the maxi mum distance was 1400 m. For subjects tracked > 20 days, the mean displacement distance from the release patch was 674 m (Â± 606), and the maximum displacement was 2200 m. Finally, the mean size of wooded patches that subjects were known to have visited was 20.94 ha (Â± 77.00), whereas the mean patch size for the study area as a whole was 5.85 ha (Â± 103.15). Subjects dispersing across the matrix frequently moved among small stepping stone patches, sometimes making repeated return visits. Thirty -six percent of disper sing subjects visited at least one small patch (< 1 ha), and 45% re turned at least once to a previously visited patch. In some cases, subjects dispersing thr ough both open and shrub matrix failed to orient toward suitable habitat at relatively short distances (< 150 m). For example, subject O8 failed to orient toward a 13.5-ha patch located 85 m across the open matrix, moving instead to a tiny (< 0.1-ha) patch lo cated 95 m across the open matrix in the opposite direction. The subject ultimately settled on a territory in the 13.5-ha patch, after traveling a long circuitous rout e to reach the patch it initially ignored. Likewise, in the shrub matrix, subject S4 wandered seemi ngly at random for 7 days before dying,
27 apparently of starvation. This subject failed to orient toward several wooded patches on the periphery of the shrub field that were 130 m from the release patch. Necropsy results showed no signs of injury or illness, but the digestive tract was empty at the time of death. All other subjects that died during experimental trials (Table 2-1) were found either partially consumed or with w ounds consistent with predation. Discussion Corridor versus Matrix Movement My results indicate that wooded corr idors and shrubby vegetation function similarly as movement habitat for dispersi ng Chucaos. Thus, these elements may be similarly viable for use in landscape manageme nt to enhance connectivity. As expected, results indicate that open habitat constrai ns Chucao movement, but does not entirely prevented it. Open habitat gaps 20 m were crossed routin ely, but subjects appeared reluctant to cross gaps 60 m, and few were known to cross gaps 80 m. Although the release of subjects into such unsuitable pa tches undoubtedly provided a strong stimulus for movement, the fact that some individua ls remained in the tiny release patches 30 days (rather than crossing 120-130 m of open ma trix), demonstrates the strength of their resistance to dispersing in the open. This c onstraint may reduce immigration into isolated forest patches, potentially influencing patch o ccupancy patterns at the landscape scale. Indeed, census data confirm that Chucaos are frequently absent from isolated patches, suggesting population-level effect s of movement constraint (Cha pter 3). In contrast to open habitat, subjects regularly traveled distances 300 m in narrow wooded corridors, and easily crossed distances 100 m in shrub-dominated matr ix. Regular use of small stepping-stone patches indicat es that such patches may serve an important function for movement in fragmented landscapes.
28 Generality of Results Among forest bird species, understory insec tivores have repeatedly been identified as highly sensitive to fragmentation (e.g., Lovejoy et al. 1986; Sieving & Karr 1997; Recher & Serventy 1991). In South American temperate ra inforest, this group includes the endemic tapaculos, which are among the most sensitive to fragmentation in the biome (Willson et al. 1994). Four species occur in the region including the Chucao Tapaculo, Ochre-flanked Tapaculo ( Eugralla paradoxa ), Magellanic Tapaculo ( Scytalopus magellanicus ), and Black-throated Huet-huet ( Pteroptochos tarnii ). A previous experimental study (using song-playback as a stimulus) showed that these species are reluctant to enter open habitat (pasture), but that edge permeability generally increased with increasing density of vegetation in the matrix (Sieving et al. 1996). Also, 40% of respondents could be drawn into narrow ( 10 m) wooded corridors with the song playback method (Sieving et al. 2000). Only the smallest species, the Magellanic Tapaculo (ca. 11 g), readily entered sparsely vegetated matrix. This species occupi es the widest variety of habitat types, has the largest geographic range (Sibley & Monroe 1990; Ridgely &Tudor 1994), and is arguably the strongest flyer (T.D.C., personal observations) among the four species. The Ochre-flanked and Chucao Tapaculos have intermediate body mass (ca. 25 g and 40 g, respectively) and very small geographic ranges , while the largest species, the Huet-huet (ca. 150 g), has a range that is intermediate in size (Sib ley & Monroe 1990; Ridgely & Tudor 1994). While the Chucao and Huet-hue t responded similarly to habitat boundaries, the Ochre-flanked Tapaculo, the rarest and most patchily distributed of the species, appeared most reluctant to enter th e matrix (Sieving et al. 1996).
29 Similar reluctance among tapaculos (ex cept the more dispersive Magellanic Tapaculo) to enter open or sparsely vegetate d matrix indicates that results obtained for the Chucao, which is intermediate in size, vagility, and habita t specificity, may be generally applicable for pl anning conservation action to support the entire group. Although maximum travel distances likely di ffer among species, these differences may be predictable based on differences in body ma ss and territory sizes (Sutherland et al. 2000; Bowman 2003). However, planners s hould be cognizant of autecological differences among species (e.g., bamboo special ization of the Ochre-flanked Tapaculo [see Sieving et al. 2000], and ne st-site specificity and larg er home-range sizes for the Huet-huet [DeSanto et al.2002]) to ensure that habitat networks meet the requirements of the entire species suite. My results may also be generally applicable fo r understory and shrub requiring birds elsewhere (e.g., Siev ing & Karr 1997) and, because tapaculos are essentially terrestrial, their responses may be indicative of many non-volant species (e.g., Bakker &.Van Vuren 2004). Conservation Implications For species such as tapaculos, which are poor flyers and cannot use sudden flight or other aerial escape tactics for predator avoida nce, behavioral resist ance to entering open areas may be due to lack of escape cover (L ima 1993; RodrÃguez et al. 2001; Sieving et al. 2004). Thus, it may be possible to en courage movement through the matrix by managing vegetation to increase cover and, in some cases, fully restoring forested habitats or corridors may not be required to restore connectivity. Chucao movement was facilitated by shrubby vegetation in the ma trix, and anecdotal observations (movement through both low-stature secondary forest a nd shrub fields dominated by invasive B.
30 magellanica ) indicate that cover provided by th e vegetation, rather than species composition, was the relevant factor. Availability of such alternative management strategies (i.e., management of matrix vegetation structure rather than corridor prot ection/restoration) is useful because it will allow planners to optimize conservation effort s in response to local constraints (Arendt 2004). For example, it may be advantageous to protect or restor e wooded corridors in regions where land-area is at a premium fo r economic uses (because corridors require minimal area), whereas natural regeneration of secondary vegeta tion (requiring little economic investment) may be adequate in re gions where land use is less intensive. Natural regeneration of permeable vegetation in the matrix, as an alternative to corridor restoration, may be especially beneficial where local ecologica l constraints inhibit forest regeneration, for example, in ChiloÃ©, wher e waterlogged soils are invaded by persistent hydrophytic assemblages (i.e., Sphagnum sp. and B. magellanica ; Van Breemen 1995) that are, nonetheless, permeable to Chucao movement (this study). My results clearly demonstrate that treati ng all non-forest hab itats as homogeneous and impermeable could lead to omission of potentially useful alternatives for conservation planning. However, provision of travel habitat, in the fo rm of corridors or permeable matrix, should not be viewed as a vi able alternative to protecting large tracts of primary forest needed for breeding and l ong-term survival (Rosenberg et al. 1997). Census data show that Chucaos and other endemic tapaculos rarely forage or defend territories in wooded corridors < 25 m wide (Sieving et al. 2 000) and are virtually never observed in fields dominated by B. magellanica (T.D.C., personal observations). Predatory risk also appears greater in sma ll patches and corridors (where six subjects
31 were lost to predation) than in more exte nsive forested areas (see also Willson et al. 2001). Further, the fact that some subjects ha d difficulty orienting toward suitable forest patches while dispersing through sh rubs (i.e., indicative of a restricted perceptual range in this habitat type; Zolln er & Lima 1997), one apparently dyi ng of starvation, indicates that excessively large areas of shrub land could re present a form of Â“dispersal trapÂ” (see Schlaepfer et al. 2002). Nonetheless, my resu lts are encouraging in that they support two viable alternatives for maintaining functional landscape-connections that may allow managers to optimize conservation effo rts at local and regional scales. This chapter was previously printed in the Conservation Biology journal (CastellÃ³n and Sieving 2006).
31Table 2-1. Summary of radio-te lemetry results for translocated subjects (each located once daily) a nd landscape characteristic s of experimental sites where s ubjects were released*. Site/subject Days in patch Days total Disposition Patch area (ha) N-N 1 ha (m) N-N (m) % Wooded C1a 1 9 lost transmitter 0.19 340 20 7.53 C1b 3 6 death " " " " C2 8 8 lost transmitter 0.30 50 50 13.58 C3 2 30 completed 0.12 380 50 6.92 C4a 2 23 death 0.20 590 60 12.14 C4b 1 30 completed " " " " C5a 1 30 completed 0.16 280 20 12.15 C5b 7 25 completed " " " " C6a 8 30 completed 0.21 180 30 6.79 C6b 2 30 completed " " " " C7a 1 3 lost transmitter 0.13 300 50 11.18 C7b 1 30 completed " " " " C8 7 7 lost transmitter 0.12 280 130 3.95 O1a 6 6 death 0.09 130 70 3.98 O1b 30 30 completed " " " " O2a 30 30 completed 0.14 120 120 3.15 O2b 30 30 completed " " " " O3a 1 30 completed 0.07 90 80 2.94 O3b 5 5 lost transmitter " " " " O4a 15 25 lost transmitter 0.23 280 60 5.71 O4b 5 30 completed " " " " O5 6 6 death 0.20 90 40 3.48 O6a 2 17 death 0.05 30 30 12.65 O6b 1 30 completed " " " " O7a 2 30 completed 0.04 60 30 3.44 O7b 7 7 death " " " " O8 3 30 completed 0.03 101 100 1.11 S1 18 24 lost transmitter 0.10 280 50 12.24 S2a 2 30 completed 0.30 220 100 4.23 S2b 5 9 lost transmitter " " " "
32Table 2-1. Continued Site/subject Days in patch Days total Disposition Patch area (ha) N-N 1 ha (m) N-N (m) % Wooded S3a 1 6 roadless 0.25 140 60 4.44 S3b 2 30 completed " " " " S4 4 10 death 0.09 130 90 1.80 S5a 1 16 unknown 0.02 60 60 9.15 S5b 1 23 completed " " " " S6 2 6 lost transmitter 0.04 120 10 3.10 S7a 1 6 roadless 0.05 190 90 1.09 S7b 1 2 lost transmitter " " " " S8 1 2 roadless 0.09 60 60 2.05 S9a 1 5 lost transmitter 0.05 120 120 0.88 S9b 1 30 completed " " " " * The site/subject code indicat es the treatment group (C, corridor; O, open; S, shrubs) and the replicate-site number. Codes ending in a or b indicate sequentia l releases conducted in the speci fied site. Days in patch is the number of days the subject remained in the release patch prior to initial move ment, death, or transmitter loss. Days tota l is the duration of monitoring and dispo sition indicates the conditions under which monito ring was terminated. These included comp letion of monitoring, death, transmitter lo ss, movement into roadless areas, or signal loss for unknown reasons. Release site char acteristics included th e release patch area , distance from the releas e patch to the nearest-neighbor (N-N) patch 1 ha, distance to the N-N patch of any size, and the percentage of wooded habitat with in a 100-m buffer centered on the release patch.
33Open matrix Shrub matrix Forest corridor= = = Figure 2-1. Schematic representation of the three landscape treatments (release patches surrounded by open matrix, adjoining a wooded corridor, or surrounded by a matrix of dense shrubs), along with aerial photographs (1:10,000) of corresponding translocation sites for each treatment (one replicate each shown), and photographs of the sites take n in the field. Te st subjects were released (individually) into a small wooded patch at the center of each replicate.
34 Cumulative hazardDays after release20 10 0 4 3 2 1 0 Shrub matrix Corridor Open matrix Figure 2-2. Hazard function showing the condi tional probability that movement will occur during each (1-day) time interval, gi ven that it has not occurred prior to the specified time.
35 Release site Indicates location sequence in time Bird location Antenna location for continuous corridor-monitoring Transmitter foundN 1 2, 4, 635, 7, 9 8 11 10 Wooded habitat Figure 2-3. Map (1:2,000) of wooded habita t patches embedded in a matrix of open pasture showing sequential (numbered) locations of a translocated subject (C1a) recorded over a 10-day period
36 CHAPTER 3 LANDSCAPE HISTORY AND FRAGMENTATION EFFECTS ON PATCH OCCUPANCY Introduction Predictive geographical modeling has gained importance in recent decades for examining impacts of land-use change and de veloping conservation strategies (Guisan & Zimmermann 2000). The simplest approach for modeling landscape effects on wildlife populations relies on analysis of species di stribution patterns in extant fragmented landscapes (Hanski 1994; ter Braak et al. 1998). Under this Â“incidenceÂ” based approach, differences in patch occupancy by particular taxa are assumed to reflect the combined effects of landscape factors on population pro cesses that result in their absence from areas with unsuitable patch or landscape context characteristic s (Hanski 1994). Incidence based models are less data-intensive than al ternative approaches (e.g., dynamic simulation modeling), but they are stat ic in nature and automatically assume equilibrium conditions by statistically rela ting species distributions to the present environment (Korzukhin et al. 1996; Guisan & Zimmermann 2000). The static character of incidence based models is problematic b ecause occurrence of a species at a certain moment in time does not necessarily imply pe rsistence, due to potential time lags in equilibration of population processes followi ng landscape alteration (Â“extinction debtÂ”; Tilman et al. 1994). Therefore, incidence ba sed models may be va lid for interpolating patch occupancies (Â“filling inÂ” occupancies for non-sampled patches) at the time and place where data were collected. However, th ey may fail when used as a forecasting tool
37 to predict future distribution pa tterns, or when extrapolated to alternative landscapes with differing levels or histories of fragmentation (Rykiel 1996). Despite their limitations, incidence based models are often the only feasible means for informing conservation planning, sin ce comprehensive data for modeling dynamic responses to environmental change are available for so few species (Guisan & Zimmermann 2000). Nonetheless, models th at fail to account for differences in fragmentation levels or extinction time lags may significantly overe stimate persistence, with disastrous consequences for conservati on planning. Thus, it is critically important that planners recognize the hazards of extrapolating these models beyond their appropriate domains of applicab ility. If incidence-based mode ls must be extrapolated to address more general conservation questions, they should be validated using independent data from alternative landscapes (Rykiel 1996). In practice, however , such assessment is rarely undertaken. In this study, I developed and tested patch occupancy models for an endemic understory bird, the Chucao Tapaculo ( Scelorchilus rubecula ), in two South American temperate rainforest landscapes that differed in levels and duration of forest loss. I assessed cover changes since 1961 in the tw o landscapes and surveyed patches for Chucao occupancy. I then developed incide nce-based predictive m odels independently for each landscape, and tested the models reci procally in the alternative study area. By testing the models under altern ative landscapes, I assessed th e range of conditions over which they could properly be applied (Rykiel 1996) and identified predictor variables whose effects were general at the regional scal e, and those with effects that varied across landscapes (Taylor 1991). By analyzing landscapes with cont rasting fragmentation levels
38 and histories, my research provided insights into potential trajectories of change in portions of the biome currently undergoing frag mentation. Hence, this research yielded information relevant to maintaining biodivers ity into the future, while reducing risks of unanticipated extinction time lags or intensified fragmentation effects that could accompany habitat loss. Methods Study System South American temperate rainforest is glob ally outstanding for its exceptional level of endemism, and for being one of the most e ndangered ecosystems on Earth. The biome is identified as a global biodiversity hotspot (Mittermeier et al. 1998; Myers et al. 2000), a Centre of Plant Diversity (Davis et al. 1997), an Endemic Bird Area (S tattersfield et al. 1998) and a Global 200 Ecoregion (Olson & Di nerstein 1998). Among endemic forest birds, understory insectivores in the fam ily Rhinocryptidae (tapaculos) are among the most sensitive to fragmentation (Willson et al. 1994). I identified the Chucao Tapaculo as the best subject for intens ive research because it is loca lly abundant and intermediate in size (40 g) and vagility within th e group (Sieving et al. 2000), and because supplemental data were available from previ ous studies on population densities, territory sizes, reproductive success, and movement (S ieving et al. 2000; De Santo et al. 2002; Willson 2004; Chapter 2). The Chucao is a year-round resident that is territorial and, like other tapaculos in the biome, is strongly associated with unders tory vegetation (Reid et al. 2004). Although reproductive success is relatively high in fragmented landscapes (De Santo et al. 2002), inter-patch movement is constrained by ope n habitat, perhaps due to inadequate vegetative cover for avoiding predatory attack (Sieving et al. 1996; Chapter 2). Chucaos
39 are extremely poor flyers and move by wa lking, hopping, or flying short distances ( a few meters), usually within or near dens e vegetation. Though relu ctant to enter open habitat, they will use wooded corridors and dense low-stature secondary vegetation for movement (Sieving et al. 1996, 2000; Chapter 2). This sensitivity to movement habitat makes the group potentially valuable as fo cal species for planning the connectivity component of landscape design, because a landscape that provides functional connectivity for the group would probably m eet the movement requirements of many forest species. I studied patch occupancy by the Chucao in two agricultural landscapes in southcentral Chile, one located on north ern ChiloÃ© Island, near Ancud (41o55Â’ S, 73o35Â’ W; referred to hereafter as ChiloÃ©; Fig. 3-1), and the other located on the mainland near Osorno (40o35Â’ S, 73o05Â’ W; hereafter Osorno). The two landscapes lie within the Valdivian temperate rainforest zone (37o45Â’ S to 43o20Â’ S), comprised principally of evergreen broad-leaved trees and a few coni fers, with dense unde rstory composed of bamboo ( Chusquea spp . ) and shade-tolerant saplings (V eblen et al. 1983). Prior to settlement, this zone was larg ely covered by unbroken forest. Large-scale forest fragmentation began re latively recently in the ChiloÃ© study area (within the last 50 Â– 100 years; Willson & Armesto 1996) and the landscape is now dominated by small-scale subsistence farmi ng. Forest habitat a ppears relatively well connected in ChiloÃ©, in that most patc hes are linked by corridors or isolated by 100 m of non-forest matrix. Osorno, in contrast, is more fragmented and has a longer history of human-induced disturbance (100 Â– 150 years; Donoso & Lara 1995). This area is located
40 in the Chilean Central Valley, with the highest human populati on density and most intensive commercial land-use in th e biome (Armesto et al. 1998). Land Cover Analysis Remote sensing and Geographic Information System (GIS) analysis were used to evaluate land cover changes in the two study areas and to assess matrix composition surrounding subsets of focal patches. Remote sensing data consisted of panchromatic aerial photographs from 1961 and 1993 (the longest period of photographic data available), each taken during mid-to-late summer. All data were available in orthophotograph format (1:20000) except the 19 61 data for ChiloÃ©, which consisted of non-rectified partially overlapping photographs (1:70000). Digital images of each study area were produced as follows using E NVI 3.2 (RSI 1999). All photographs were scanned and the orthophotographs were ge oreferenced using Universal Transverse Mercator (UTM) grids printed on the photograp hs as references. The non-rectified 1961 ChiloÃ© photographs were registered to the 1993 orthophotograph, sp atially corrected using a rubber-sheet algorithm (root mean square errors < 0.5 m), resampled using a nearest-neighbor algorit hm, and arranged to produce a digital photographic mosaic. Comparison of the ChiloÃ© and Osorno la ndscapes, and change analysis from 1961 to 1993 in each landscape, was based on land cover classification within a 100-km2 subset of each study-area. Land cover t ypes were hand-digitized with a minimum mapping-unit of 0.1 ha using ArcView 3.2 (ESR I 1999) on-screen-digi tizing functions. For the 1993 coverages, clas sification included open, shrubby, and wooded habitats. These classes were selected because prior research indicated that these cover types differed in permeability to Chucao movement , with movement significantly constrained by open matrix (Chapter 2). For the 1961 c overages, only two hab itat classes, wooded
41 habitat and non-wooded matrix, were digitized due to lower spatial resolution of the ChiloÃ© photographs, which made it difficult to distinguish shrubby from open matrix types. Wooded patches were defined as c ontinuous areas of w oodland separated from surrounding patches by gaps 10 m, or connected to other patches via wooded corridors or bottlenecks that were at least an orde r of magnitude narrower than the adjoining patches. The resulting cover maps were used to quantify landscape metrics and estimate extents of deforestation over the referenced period (measured directly from the coverages or using FRAGSTATS 3.3; McGarigal et al. 2002). Landscape-scale metrics included the mean patch area in each study landscape, patch density, patch areato-density ratio (a measure of habitat contiguity), and the per centage of area covered by each habitat class (Table 3-1). Patch and landscape-context metr ics were also quantified for a subset of focal patches in each landscape (focal pa tches were those surveyed for Chucao occupancy, see below). Patch-scale metrics in cluded focal patch area, percentage change in patch area from 1961 to 1993, and presen ce or absence of corridor linkages. Landscape-context metrics (qua ntified within 100-m and 300-m buffers surrounding each focal patch) included the pr oportional coverage of each habitat type, mean area and density of wooded patches, and two additi onal metrics potentially linked to landscape connectivity. These were Euclidean distance to the nearest patch 5 ha (assumed large enough to serve as a potential source of immigrants), and HanskiÂ’s (1994) connectivity index (Si), which accounts for number, area and di stance to each potential source patch within movement range (1-km movement range assumed; Table 3-2).
42 No quantitative accuracy assessment was undertaken for the 1993 land-cover classification, but interpretations of photographs were verified extensively in the field. The ChiloÃ© landscape had changed relatively little since 1993, but several patches in the Osorno landscape were cleared since that time. The effect of differences between the photographic data and current conditions in Osorno was minimized by selecting focal patches that had changed little since the phot ographs were taken. However, comparison between ChiloÃ© and Osorno reflects the stat us of the landscapes in 1993, rather than current conditions. Ground truth data were unavailable for the 1961 photographs. Patch Occupancy Surveys A subset of patches selected from an area of approximately 300 km2 in each landscape was surveyed for Chucao occupa ncy. To randomize patch selection, each study area was subdivided into grids of 4-km2 blocks. Then, isolated patches were censused within blocks selected using a ra ndom number table. In ChiloÃ©, 100 focal patches were censused during the 2000-2001 bree ding season (November to February). Then, following the same protocol, 62 patche s were censused in Osorno during early 2004. Fewer patches were visited in Osorno b ecause the landscape was more fragmented there, increasing travel time s between patches and the numbe r of landowners that had to be contacted. Further, most patches in Osorno were unoccupied and, therefore, requiring more intensive surveys to verify absence. During the initial visit to each patch, censuses were conducted at a maximum density of one census point per 10 ha of patch area. Census points were located 50 m from the forest edge (unless patches were prohibitively small) and were separated from each other by 100 m. Because the objective was only to document Chucao occupancy, once a Chucao was recorded in an individua l patch, no further censuses were conducted.
43 However, if a Chucao was not encountered du ring the initial visit, a more intensive census was conducted on a subsequent day ( 5 d later), at a sampling density of one point per hectare. At each census point, an 8-minute passive census was conducted (after Willson et al. 1994) in which any Chucao heard or seen within the patch bounda ry was recorded. Following the passive census, song playback was used to increase detectability by eliciting vocal respon ses (Jimenez 2000; Sieving et al. 2000). Playbacks were conducted by sequentially broadcasting tape-recorded C hucao territorialand contact-calls. A maximum of eight calls were played, each followed by a 1-min period of silence. Because sampling occurred during the breeding season, responsiveness to taped calls was high (e.g., Sieving et al. 1996, 2000). All censu ses were conducted between dawn and 1100 hrs (EST) on mornings wit hout strong wind or rain. Habitat Quality Characterization Habitat quality is a potentially important pred ictor of patch occupancy. Therefore, to incorporate this variable in the model buildi ng process, I visually assessed and classified habitat quality for each censused patch. Focal patches consisted principally of mixed age wooded stands that were either selectively logged or entire ly secondary. Such patches were assumed adequate for Chucaos becau se reproductive success is similar in continuous forests and fragments, and there is no evidence of si gnificant edge-related effects on nest fate (De Santo et al. 2002) . Because habitat quality for Chucaos is determined largely by understory vegetation density (Reid et al. 2004), defined by stem density and the degree of contact among bran ches, I visually characterized understory density using three cat egories; Â“sparse,Â” Â“dense,Â” and Â“very dense.Â” Very dense understory had many stems in contact with each other, at places forming impenetrable
44 tangles. Dense understory had many stems but less contact among pl ants overall, though patchy areas of very dense vegetation were so metimes present. Sparse understory was easy to walk through, had few stems in contac t, and generally lack ed dense patches. Given previous knowledge of habitat suitabil ity for Chucaos (Sieving et al. 2000; Willson et al. 2004; Reid et al. 2004), only patches with understory vegetation classified as Â“denseÂ” or Â“very denseÂ” were considered in the analysis. To test the assumption that my visual ch aracterizations were unbiased (i.e., that I did not unconsciously bias habitat qualit y classifications of patches based on my knowledge of their occupancy status), I qua ntitatively assessed understory vegetation density in a subset of occupi ed and unoccupied patches (10 each, ranging in size from 0.1 Â– 3.7 ha) in the ChiloÃ© study area. In each pa tch, 5-m-radius circular sampling stations were established at random locations with a sa mpling density of four stations per hectare. For each station, understory vegetation density was indexed at four points (located at right angles from each other on the perimeter of the circle) then averaged. Density was measured by standing a 3-m pole (2-cm diam eter) perpendicular to the ground and counting the number of leaves or stems th at touched the pole. A t-test showed no statistically significant difference in mean understory density between occupied and unoccupied patches (t18 = -1.72, P = 0.10; although dens ity was somewhat higher in occupied [107.60 Â± 27.11 S.E.] versus unoccupied [70.60 Â± 17.92] sites). Based on this result, I accepted the visual classifications as reasonably accurate and unbiased measures of habitat characteristics in the censused patches.
45 Predictive Models I used classification tree analysis to deve lop predictive patch occupancy models for Chucaos in the two study lands capes. Classification-tree modeling involves recursively partitioning (splitting) a data set into incr easingly homogeneous subsets (nodes), with each split defined by a simple rule based on the value of a single predictor variable (Breiman et al. 1984; DeÂ’ath & Fabricius 2000) . At each split, each predictor variable entered into the model is assessed independent ly and the one variable that generates the greatest improvement in homogeneity of the two resulting daughter nodes is selected as the node splitting criterion. Then the value for this variable is identified that minimizes heterogeneity in the daughter nodes when used as a Â“cut pointÂ” or threshold value for segregating the data. Classification-tree analysis has been shown to produce better prediction of plant and animal distributions than other popular modeling approaches, such as generalized linear models and generalized additive models (Franklin 1998; Vayssieres et al. 2000). The method is appropriate for complex ecologi cal data sets that include imbalance, nonlinear relationships, and high-order interactions, which are dealt with by partitioning the observations and then analyzing each group separately. Further, model building is not hindered by colinearity because each split is based on the value of a single predictor variable. Predictors may also be used repeat edly at different points in the tree, thus the method is inherently context specific. Analyses were performed using DTREG (S herrod 2003). Data sets for ChiloÃ© and Osorno were analyzed independently to build two separate predictive models. Gini goodness of fit measures were used to determine optimum splits. To avoid model overfitting, a minimum node size of 10 observations was required to perform a split and trees
46 were constrained to the numb er of nodes producing the mini mum relative validation error (cross validation error co st relative to a 1-node tree), which was calculated using v-fold cross validation. V-fold cross validation is performed by splitting the data set into v subsamples (for my analysis, v = 3), then pr oducing v trees, each time leaving one of the sub-samples out of the training set and using it as a test sample for validation. Crossvalidation error costs are then computed fo r each node of each tree, and these costs are used to statistically determine the optimum tr ee size. In addition, I interactively removed splits with minimal value for discerning betw een occupied and unoccupied patches (see results for further explanation). In both the ChiloÃ© and Osorno models, focal patch area was specified as the first variable used to split the data set, wh ich facilitated model comparison. This was appropriate because patch area was consistently identified as a significant predictor of occupancy during exploratory analysis. S ubsequent selection of variables for node splitting was automated. Potential predictors used in the model building process included two qualitative variables, the category describi ng understory vegetation density (dense or very dense) and corridor presence, al ong with 12 quantitative patch and landscapecontext variables (Table 3-2). Model accuracy was assessed based on the percentages of the training data and cross validation sets that were correctly classi fied. To assess model generality, I then applied each model recipr ocally in the altern ative landscape to determine the predictive accuracy of the mode ls when extrapolated beyond the specific landscape where data were collected.
47 Results Land Cover Analysis By 1961, deforestation in Osorno was advan ced and habitat con tiguity (patch area to density ratio; Table 3-1, Fig. 3-2) was already quite low. Although habitat loss continued in Osorno from 1961 to 1993, contigu ity decreased relatively little. Much more forest remained in ChiloÃ©. Al though wooded habitat cover and contiguity decreased dramatically after 1991, ChiloÃ© still retained approximately 44.92% wooded cover in 1993, compared to approximately 17.12% in Osorno. Within the non-forest matrix, the ChiloÃ© study area was composed of 35.56% open hab itat and 19.52% shrubby secondary vegetation (principally Baccharis magellanica , a persistent shrubby invader of poorly drained soils). In contrast, Osorno was dominated by open habitat (80.16%), with only 2.73% sparse or shrubby vegetation. Chucao Distribution Patterns In ChiloÃ©, Chucaos occupied 67% of focal patches. The smallest occupied patch was 0.64 ha, and the largest unoccupied patch was 5.29 ha. In Osorno, however, only 21% of focal patches were occupied. The sm allest occupied patch was 0.19 ha, and the largest unoccupied patch was 21.78 ha. In bot h landscapes, very small patches (< 1 ha) were usually occupied only if they were near larger patches, surrounded by dense secondary vegetation, or connected by corridors to other wooded habitat. In general, occupied patches in both lands capes were also larger and less likely to have undergone a major reduction in area since 1961. Further, o ccupied patches were nearer other patches 5 ha, were better connected (i.e., Si was higher), and had larger percentages of wooded habitat but smaller percentages of open habitat in the surrounding matrix (Table 3-2). In
48 ChiloÃ©, occupied patches also had higher pe rcentages of dense shr ubs in the surrounding matrix, but this trend was not observed in Osorno where shrub matrix was uncommon. Predictive Models The relative validation error for the ChiloÃ© model was minimized at a tree size of four terminal nodes, but I removed one of thes e nodes that added little information to the model (distinguishing patch occupancy by a di fference of < 1% wooded habitat cover), producing a tree with three terminal nodes (Fi g. 3-3). This model accurately classified 82.00% of the training data, and provided 77.00% predictive accuracy for cross validation sets. Most classi fication errors were errors of commission (empty patches misclassified as occupied; Table 3-3) regarding small patches surrounded by shrubdominated matrix, or relatively large but isolated patches lo cated in areas with a longer fragmentation history than ot her parts of the study area (e .g., coastal areas). Relative validation error for the Osorno tree was minimi zed at four terminal nodes (Fig. 3-4). This model correctly classifi ed 88.71% of the training set, and predictions were 78.33% accurate for cross validation sets . In this case, most miscla ssifications were errors of omission (occupied patches classified as unoccupied), principally among small patches located along riparian corridors. Although the two models were relatively c onsistent regarding variable selection, they performed poorly when applied reciprocal ly to censused patches from the alternative landscape (as a test of genera lity). They differed in that the ChiloÃ© model predicted occupancy for all patches > 3.80 ha, whereas th e Osorno model placed this cut point at 10.34 ha (Figs. 3 and 4). Patches with ade quate habitat quality that were below the referenced sizes were predicted occupied only if the immediatel y surrounding matrix consisted of < 72.61% (Osorno) to 77.50% (Chi loÃ©) open habitat. The Osorno model was
49 only 54% accurate when applied to census data for ChiloÃ© patches, (98% of the misclassifications were errors of omission pertai ning to patches 10.34 ha), and the ChiloÃ© model was only 60% accurate when applied to Osorno data (97% of the errors were errors of commi ssion regarding patches 10.34 ha). Discussion Models developed for the two landscapes were relatively consistent regarding variables chosen as important predictors of occupancy, a nd the predictive accuracy of each model was high in the landscape where training-data were collected (Figs. 3 & 4, Table 3-3). Further, the two models were remarkably consistent regarding the influence of open matrix, both predicting occupancy of small patches only if open habitat composed 75% of the surrounding matrix, on averag e. However, the models differed substantially from each other in the magnit udes of effects related to patch size and, therefore, performed poorly wh en applied reciprocally to the alternative landscape. Landscape Analysis Analysis of historic land cover data documented substant ially different levels of habitat loss and fragmentati on in the two study areas by 1961 (t he earliest data for which aerial photographs were availabl e; Table 3-1, Fig. 3-2). This observation supported my assumption that longer periods of relative is olation had elapsed in many Osorno patches by the time my surveys were conducted, allo wing more time for population decline to extinction. In Osorno, large-scale deforestat ion began in the mid-1800Â’s, so that only 21% forest cover remained in 1961, and habitat contiguity was alrea dy low. Forest loss and fragmentation continued thereafter a nd, by 1993, only 17% wooded habitat remained, but contiguity decreased little because the in dex approaches zero asymptotically with decreasing patch size. In contrast, deforest ation in ChiloÃ© began in the early 1900Â’s, and
50 by 1961 approximately half the original cover remained. Contiguity dropped considerably in ChiloÃ© during the years th at followed. Nonetheless, in 1993, the landscape still supported approximately 45% wooded habitat cover, and much of the deforested matrix was vegetated by dense sh rubby habitat, which is relatively permeable to Chucao movement (Chapter 2). The simultaneous availability of sites that differed in time since initiation of the fragmentation process allowed me to apply a space-for-time substitution approach (Warton & Wardle 2003; Purtauf et al. 2004). Although comparisons among landscapes are problematic because habitat quality often vari es clinally with climate, I assumed that pre-fragmentation habitat quality was compar able in the two regions because Chucao densities were similar in ChiloÃ© and mainla nd parks (ChiloÃ© National Park and Pumalin National Park; Willson et al. 2004). Nonetheless, my analysis is bounded by the caveat that pre-fragmentation densities may have differed, potentially influencing current occupancy patterns. Despite this uncertain ty, the Osorno landscap e represents the only area in the biome with a substa ntially longer fragmentation hi story than ChiloÃ©, and thus provides the best possible reference for comparison. Chucao Distribution Patterns Results of landscape analysis and a radi o telemetry study of Chucao movement behavior (Chapter 2) indicate that the ChiloÃ© landscape is still relatively well connected with regard to Chucao movement abilities (i.e., few patches isolated by 80 Â– 100 m, shown to seriously impede movement). Furt her, with an average patch size of 6.75 ha (Table 3-1), most patches were still large enough to support severa l breeding territories (approximately 1 ha each; De Santo et al. 2002). Thus, given that most patches are occupied and well connected, population structur e in ChiloÃ© appears to represent a single
51 patchily distributed population (Ovaskainen & Hanski 2004). Many patches < 10 ha that are currently occupied in ChiloÃ© may function as sink habita t, in that the populations are too small to persist without fr equent immigration from larg er source populations (Pulliam 1988). Nonetheless, occupancy of these pa tches may serve to augment the regional population size and may produce reproductive surp lus in some years, further increasing the flow of dispersing individuals through th e landscape (Dias 1996). In contrast, Osorno habitat conditions appear larg ely inadequate for long term sustainability of Chucao populations, with only 17% wooded habitat re maining, a mean patch size of 1.44 ha, and intervening matrix dominated by impermeable open habitat (Table 3-1). Most patches < 10 ha in Osorno are unoccupied, and population structure is in dicative of a (possibly nonequilibrium) metapopulation, indicating that immediate conservation action may be required to prevent further population declines or extinction (Harri son & Taylor 1996). Given previous knowledge that Chucaos repr oduce relatively well in forest patches (De Santo et al. 2002), their absence onl y from relatively small patches ( 10 ha) surrounded by open habitat (shown to impede movement; Chapter 2) in both study areas indicates that extinction is lik ely only for very small, func tionally isolated populations that are vulnerable to demographic and environmental stochasticity. An additional point important for conservation planning is that an average of approximately 25% permeable habitat in the matrix surroundi ng a focal patch was a good predictor of occupancy in both landscapes. Thus, 25% permeable-habitat cover in th e matrix may be an adequate goal for conservation planning to provide connectiv ity within patch networks (at the spatial scales tested). This conclusion is su pported by a radio-telemetry study documenting successful movement by Chucaos through wooded corridors and dense secondary
52 vegetation in a fragmented landscape, ev en though the permeable habitats often represented a relatively small percentage of the landscape area (Chapter 2). Model Validation Accuracy of each model was high in the land scape where data were collected (Figs. 3-3 & 3-4, Table 3-3), but the models performe d poorly when applied reciprocally due to differences in predicted occupancies of pa tches in the 3.80-ha to 10.34-ha size range. The ChiloÃ© model predicted occupancy of all bu t the smallest and most isolated patches, whereas the Osorno model was more conserva tive, predicting occupancy only for large patches ( 10.34 ha) and smaller high-quality patche s with adequate pe rmeability in the surrounding matrix. Because both models stat istically related occupancy to current environmental conditions (i.e., they were static equilibrium-based models), their differences potentially reflected the contras ting stages of habitat loss and fragmentation exemplified by the two landscapes, which was more advanced in Osorno. The preponderance of errors of commissi on when applying the ChiloÃ© model to the Osorno landscape indicated that patches used to build the ChiloÃ© model may not reflect equilibrium conditions, potentially due to an extinction debt in ChiloÃ©. Chucao demographic data are suggestive of an extin ction debt because reproductive success in forest patches is relatively high. Even though population density (Willson et al. 2004) and pairing success (Willson 2004) in fragments is lower than in continuous forest, breeding pairs still have a relatively high ch ance (63%) of fledging at least one young per clutch, and of producing two or three clutches per season (D e Santo et al. 2002). Given this level of in situ reproductive output, small populations could potentially persist in effectively isolated patches for many ge nerations following fragmentation, although long-
53 term persistence of populations with 10 breeding pairs is unlik ely due to environmental and demographic stochasticity. If an extinction debt existed in ChiloÃ©, I expected a stronger pa tch-size effect (i.e., larger unoccupied patches) in Osorno (havi ng a longer fragmentati on history), which my data confirmed. However, patches in Osorno were also more isolated than in ChiloÃ© (larger inter-patch distances and a higher proportion of Â“impermeableÂ” open matrix). This greater isolation may have further redu ced immigration rates, thereby reducing time to extinction in Osorno patches (Stacey et al. 1992; Ovaskainen 2002), which could also explain larger empty patches. Likewise, th e higher percentage of wooded habitat in ChiloÃ© provided more breeding habitat than was available in Osor no, possibly increasing the flow of dispersers through the Chilo Ã© landscape. Finally, the higher human population density in Osorno may intensify an entire suite of detrimental effects associated with human land uses (e.g., Fo rd et al. 2001), possibly reducing the Osorno Chucao population. Unfortunately, it was impossible to dissociat e the potentially interacting effects of duration, extent, and intensity of human land use on patch occupancy. However, I viewed the study landscapes as representi ng two extremes along a continuum of fragmentation occurring in the biome. Thus , occupancy patterns in the two contrasting landscapes provided insights into potential population traj ectories for areas where fragmentation is ongoing. Assuming mechanis ms driving landscape change are similar (human population growth and agricultural intensification), then the Osorno model, developed for a landscape at an advanced stag e of fragmentation (spa tial and temporal), could reasonably be used as a standard to forecast conditions in alternative landscapes
54 once they (hypothetically) reach similar stages . Clearly, the Osorno landscape represents a more advanced stage of fragmentation (based on the contiguity index), and the ChiloÃ© landscape appears to be following a similar trajec tory (Table 3-1, Fig. 32). It is therefore a reasonable expectation that current conditions in Osorno represent possible future conditions for ChiloÃ© (or other landscapes currently undergoing fr agmentation) if no action is taken to preserve higher levels of forest cover and contiguity. However, it should be noted that rates of change may be faster in landscapes currently undergoing fragmentation than occurred historically because technological advances have provided more efficient means of defore station now than in the past. Evidence that Chucao populations in Chilo Ã© may be following a trajectory similar to Osorno includes the following. Although the Osorno model generally failed to predict current occupancy patterns in ChiloÃ©, due to errors of omission for patches in the 3.80-ha to 10.34-ha size range, some extr emely isolated ChiloÃ© patches within this size range are, in fact, currently unoccupied. Further, demogr aphic isolation effects are evident in other such patches (e.g., reduced pairing success due to a preponderance of unmated males; Willson 2004). Thus, many of the ChiloÃ© patches predicted unoccupied by the Osorno model may be declining toward eventual extinct ion. If this is true, predictions of the Osorno model may approximate the future dist ribution of Chucao populations in ChiloÃ©, once localized extinctio ns have occurred. Alternatively, the ChiloÃ© model (based on data from the relatively well-connected ChiloÃ© landscape where deforestation is recent ), may be prone to errors of commission when applied to alternative landscapes (e.g., Os orno), or when used as a forecasting tool. Such forecasting errors might ar ise if extinction time lags exist or if further habitat loss
55 causes intensified fragmentation effects (by eliminating demographic sources). Thus, static models developed from distribution pa tterns in landscapes where habitat loss is recent or ongoing may be minimally applicable for conservation because they may be overly optimistic when used to forecast future distributions, or to pr edict distributions in more highly fragmented landscap es. Nonetheless, due to lack of time and financial resources for more extensive research, static mo dels will likely continue to serve as the best available information for conservati on decision making in many cases (Guisan & Zimmermann 2000). Given my results, I sugges t that landscapes used as standards for building incidence-based models intended for genera l conservation planning should be selected with caution. I recommend that species dist ribution patterns be obtained from landscapes where fragmentation has reached relatively advanced stages, preferably with histories of fragmentation long enough that time-delayed ex tinctions would already have occurred. Alternatively, prudent ecologically based crite ria, such as estimates of sustainable population sizes and the habitat area needed to suppor t the requisite populations, could be used to augment the empirically based predic tive models. Such criteria might serve to build additional safeguards into the planning process that would minimize risk of unanticipated extinctions. Of course, it is impossible to validate the predictive accuracy of models used in this way. Nonethele ss, such tools will provide planners with conservative estimates of potential outcomes that are credible because they are bounded by empirical observations in real landscapes where fragmentation is advanced. Though imperfect, such tools are critically needed since conservation decisions must be made, and predictive models are necessary to inform these choices (Rykiel 1996).
56 Table 3-1. Metrics describing sp atial patterns of forest cove r in the ChiloÃ© and Osorno study areas in 1961 and 1993 including the relative abundance of wooded habitat (percentage), mean patch area (Â± S.E.), patch density (patches per ha) and patch area to density ratio. The total area of analysis in each study landscape was 100 km2. Metric ChiloÃ© 1961 ChiloÃ© 1993 Osorno 1961 Osorno 1993 Percentage wooded 50.44 44.92 21.05 17.12 Mean patch area (ha) 13.25 Â± 8.49 6.75 Â± 2.80 2.57 Â± 0.45 1.44 Â± 0.15 Patch density 3.85 6.65 8.20 11.84 Patch area/density 3.44 1.02 0.31 0.12
57Table 3-2. Mean ( S.E.) values for metr ics describing focal patches censused for Chu cao occupancy (patch area, percentage cha nge in patch area from 1961 to 1993, distance to the nearest patch 5 ha, habitat connectivity index [Si; Hanski 1994]), and landscape context variables measured within 100-m and 300-m buffers around each patch (mean patch area and percentages of wooded, open and shrubby ha bitats), for occupied and unoccupied patches in Chilo and Osorno. Wooded habitat is the preferred habitat type. Open habitat is avoided and constitutes a movement barrie r at distances of more than approximately 50 m. Shrubby habitat is permeable to move ment but does not serve as living or breeding habitat. Chilo Osorno Variable (units) Occupied (n = 67) Unoccupied (n = 33) Occupied (n = 13) Unoccupied (n = 49) Patch area (ha) 14.07 3.74 1.11 0.25 13.22 4.00 4.09 0.64 Percentage change (ha) -12.59 5.58 -42.33 35.58 0.46 0.45 -0.97 0.22 Nearest patch 5 ha (m) 148.82 27.29 341.97 47.16 51.77 79.87 226.69 35.06 Connectivity index (Si) 238.97 77.10 96.13 19.48 54.56 7.69 28.29 1.80 Mean patch area (ha) 100-m buffer 0.96 0.18 0.25 0.04 1.82 0.39 1.11 0.15 300-m buffer 0.91 0.14 0.78 1.95 1.38 0.15 0.87 0.07 % wooded 100-m buffer 17.17 1.14 10.76 1.34 42.72 3.76 28.52 1.68 300-m buffer 26.76 2.23 20.84 2.38 29.97 1.40 23.00 1.08 % open 100-m buffer 47.00 3.42 65.60 5.53 56.56 3.75 71.17 1.70 300-m buffer 45.22 2.73 62.24 4.69 67.70 1.12 76.51 1.10 % shrubs 100-m buffer 35.83 3.38 23.64 5.61 0.72 0.40 0.31 0.19 300-m buffer 29.28 2.77 16.92 3.55 2.32 0.79 0.49 0.26
58Table 3-3. Classification-tree model confusion matrices for ChiloÃ© and Osorno tr aining data and the three v-fold cross validat ion sets (0 = unoccupied, 1 = occupied, % misclassified refers to th e number of incorrect predic tions within the referenced occupancy category). ChiloÃ© Osorno Data Set Actual category Predicted category % Misclassified Predicted category % Misclassified 0 1 0 1 Training 0 19 14 42.42 43 6 12.25 1 4 63 5.97 1 12 7.69 Validation 0 20 13 39.39 38 10 20.83 1 10 57 14.93 3 9 25.00
59 ChiloÃ© Osorno N 048 km Figure 3-1. Satellite image (grayscale) of the study region showing the Osorno and ChiloÃ© study areas. Darker areas indi cate wooded habitats and lighter areas indicate agricultural land uses. The enlarged areas are thematic maps of 100km2 subsets of the study areas showing wooded (gray) and deforested (white) habitats.
60 0 2 4 6 8 10 12 14 02468101214Patch densityPatch areaChiloÃ© 1961 ChiloÃ© 1993 Osorno 1961Osorno 1993 Figure 3-2. Mean patch area (ha) plotted against patc h density (patches per ha) in the two landscapes (ChiloÃ© and Osorno) measur ed from aerial photographs taken at two points in time (1961 and 1993). Toge ther, these two metrics provide an index of landscape contiguity, which d ecreased substantially in the ChiloÃ© landscape over the referenced period, a nd was more advanced (i.e., contiguity was much lower) in the Osorno landscape at an earlier point in time.
61 Node 1 Occupancy = 1 N = 100 Misclassificati on = 33.00% Node 4 Open matrix in 100-m buffer 77.50 % Occupancy = 1 N = 40 Misclassification = 32.50 % Node 5 Open matrix in 100-m buffer > 77.50 % Occupancy = 0 N = 23 Misclassificati on = 17.39 % Node 2 Patch area 3.80 ha Occupancy = 0 N = 63 Misclassification = 49.21 % Node 3 Patch area > 3.80 ha Occupancy = 1 N = 37 Misclassification = 2.70 % Figure 3-3. Classifica tion-tree for predicting patch occupancy by Chucao Tapaculos in the ChiloÃ© landscape. This tree accurately classifi ed 82.00% of the training data and 77.00% of the three v-fold cross validation sets.
62 Node 1 Occupancy = 0 N = 62 Misclassification = 20.97% Node 4 Understory = dense Occupancy = 0 N = 39 Misclassification = 2.56 % Node 2 Patch area 10.34 ha Occupancy = 0 N = 53 Misclassification = 13.21 % Node 5 Understory = very dense Occupancy = 1 N = 14 Misclassification = 57.14 % Node 3 Patch area > 10.34 ha Occupancy = 1 N = 9 Misclassification = 33.33 % Node 6 Open matrix in 300-m buffer 72.61 % Occupancy = 1 N = 9 Misclassification = 33.33 % Node 7 Open matrix in 300-m buffer > 72.61 % Occupancy = 0 N = 5 Misclassification = 0.00 % Figure 3-4. Classificationtree for predicting patch occupancy by Chucao Tapaculos in the Osorno landscape. This tree accura tely classified 88.71% of the training data and 80.00% of the three v-fold cross validation sets.
63 CHAPTER 4 SUSTAINABLE PATCH-NETWORK CRITERIA Introduction Conservation planning has traditionally re lied on large-scale systems of reserves and other protected areas for safeguardi ng biodiversity (Red ford & Richter 1999; RodrÃgues et al. 2004). However, it is often impossible to protect la rge areas of native habitats in human-dominated landscapes due to constraints on land availability (Nantel et al. 1998). Under these circumstances, successful conservation may requi re strategies for designing landscapes outside rese rves that retain native bi ota in addition to supporting productive land uses for humans (Rosenzweig 20 03). This is particularly relevant in areas where high human populationdensities coincide with hab itats that are exceptionally valuable for conservation (e.g., areas with elev ated biological diversity or endemism). South American temperate rainforest (S ATR) is globally outstanding due to an exceptionally high proportion of endemic species, and for being among the most endangered ecosystems on Earth. The biome is id entified as a global biodiversity hotspot (Mittermeier et al. 1998; Myer s et al. 2000), a Centre of Pl ant Diversity (Davis et al. 1997), an Endemic Bird Area (Stattersfiel d et al. 1998) and a Global 200 Ecoregion (Olson & Dinerstein 1998). A lthough a large proportion of the region is protected in reserves (19%; Jax & Rozzi 2004), the majority of protected areas (> 90%) lie outside the areas of highest endemism and diversity (at high elevations of the Andean range or at the extreme southern extent of the biome; Armest o et al. 1998). For example, forests of the Chilean Coastal Range and Central Valley, hist orically supporting th e highest levels of
64 diversity and endemism, are poorly represen ted in the reserve system. The Central Valley in particular has the greatest human population density in the region, and the most intensive commercial land-uses (primarily planta tion forestry and agriculture). Thus, it is clear that conservation for this portion of the biome must be accomplished within the context of a highly fragmented a nd densely populated landscape. Designing landscapes for both humans and wildlife is complex and requires quantifiable planning-targets to guide the efforts of conser vation practitioners. Although the conservation biology literatur e provides a number of gene ral but vague guidelines for conservation planning (e.g., large connected ha bitats are better than small fragments), translating these guidelines into pragmatic de sign criteria is challenging (Tear et al. 2005). To provide prescriptions for cons ervation planning in the SATR biome, I developed a set of simple criteria for disti nguishing habitat configur ations (in real or hypothetical landscapes) with reasonably high expectancy for sustaining viable populations of an endemic movement c onstrained bird, the Chucao Tapaculo ( Scelorchilus rubecula ). The criteria address Â“sustain ableÂ” population sizes, habitat area needed to support populations of this size, and empirical measures of functional connectivity among spatially disjunct patches. Finally, I used the criteria to assess the efficacy of a controversial conservation st rategy, restoration of habitat connections among functionally isolated patc hes (using habitat corridors or other interventions to enhance matrix permeability). The proposed crit eria represent a course-filter approach to landscape assessment, which assumes that appropriate distribu tions and amounts of habitat will directly relate to population viability. The criteria provide a basis for
65 identifying dangerous levels of fragmenta tion that may signal impending extinction, and for objective comparison among alternative conservation strategies. Methods Study Species Of bird species endemic to SATR, understory -insectivores in the family Rhinocryptidae (tapaculos) are among the most sensitive to fragmentation (Willson et al. 1994), which most likely results from movement cons traints caused by poor flying ability and behavioral resistance to entering open habita t (Sieving et al. 1996; Chapter 2). This sensitivity makes tapaculos poten tially useful focal species for planning the connectivity component of landscape design, to meet th e needs of other species less sensitive regarding movement habitats (Willson et al. 1994). The Chucao Tapaculo was identified as the best species for intensive research be cause it is intermediate in size (40 g) and vagility among the four tapaculos (Sieving et al. 2000), and because supplemental data were available on population densities, territory sizes, reproductive success and movement (Sieving et al. 2000; De Santo et al. 2002; Willson 2004; Chapter 2). Chucao nest success is relatively high in forest fragments (63%), and there is no documented negative edge-related effect on nest fate (De Sa nto et al. 2002). Like other tapaculos in the biome, Chucaos are strongly associat ed with understory vegetation (Johnson 1965, 1967; Ried et al. 2004). Though reluctant to enter open habitat, Chucaos will use wooded corridors and dense secondary ve getation for movement (Sieving et al. 2000; Chapter 2). Landscape Criteria for Sustainable Populations I developed a set of simple criteria for disti nguishing configurations of forest patches with reasonably high expectancy for long-term persistence of Chucao populations. The criteria are derived from several data sources. The necessary
66 population size was based on empirical estimat es of sustainable population sizes for a variety of similarly sized ve rtebrates (Verboom et al. 2001 ). The amount of habitat needed to support a population of the specified size was then calculated based on the size of Chucao breeding-territories (Willson et al. 2004; Willson unpublished data). In addition, empirical patch-occupancy models were used to identif y patch and landscapecontext factors related to persistence in in dividual patches (Chapt er 3), and movement data from a radio telemetry study (Chapter 2) were used to predict Chucao movement abilities. Finally, after using the criteria to identify sustainable patch configurations within a set of test landscapes, I estimated the carrying capacity of each landscape by calculating the number of breeding territories potentially supported within all sustainable networks combined. Mathematical Â“graphsÂ” were used as a conceptual format to represent the habitat patche s and connections addressed by the suggested criteria, following Urban and Keitt (2001). Sustainable Population Size Estimating minimum sustainable population si zes is difficult and controversial in practice. However, this estimate serves as a foundation for predicti ng the area of habitat required by a sustainable population and is, th erefore, necessary for providing specific quantitative targets for conservation planni ng (Allen et al. 2001; Noss et al. 2002). Verboom et al. (2001) estimated sustainable population sizes ranging from approximately 20 to 100 reproductive units for a variety of bi rds, and from 40 to 100 reproductive units for medium to small vertebrates in general. These estimates were based on empirically observed persistence of real popu lations occurring in near comp lete isolation, rather than theoretical models of population viability that can be qui te uncertain. In accord with
67 these estimates, I selected 100 breeding pairs as the minimum Chucao population size considered potentially sustainable. I selected he highest value in the rang e of population sizes because course-filter criteria such as mine should set targets above the absolute minimum to account for potential variability in hab itat quality and population densit ies in patches of different sizes and across large geogra phic regions (see Willson et al. 2004). The selected population size is also in line with prev ailing thought regarding minimum viable populations, with populations of 50 individuals assumed to represent the minimum size necessary for persistence given demographi c stochasticity (Richt erÂ–Dyn & Goel 1972; Caughley1994), and 200 individu als (i.e., 100 pairs) repres enting the minimum required to guard against inbreeding depression (L ande & Barrowclough 1987; Caughley 1994). Nonetheless, I do not assume that 100 breedi ng pairs represents the true minimum viable population for Chucaos. Rather, I used th e estimate as a working target, which may require adjustment as better data become av ailable (Margules & Pressey 2000; Pressey & Cowling 2001). Habitat Area Requirements Chucao breeding territories t ypically range in size from 1-1.3 ha in forest fragments to 0.4 ha in continuous forest (Willson unpublis hed data), and population densities vary from 2.51 (Â± 1.71 S.D.) individuals per hectare in fragments, to 5.69 (Â± 2.73) individuals per hectare in continuous forest (Willson et al. 2004). Larger territories in forest fragments are probably necessary due to habitat degradation (e.g., from livestock grazing), which may be more advanced in hi ghly fragmented landscapes. Given this variability, I selected a 1-ha territory size as my standard for calculating the necessary habitat area to support a sustainable populat ion. Although, actual territory sizes will
68 vary, empirical data suggest the 1-ha standa rd is a reasonable estimate for predicting territory densities in fragmented landscapes across patch sizes. I assumed 100 connected or semi-connected te rritories would be required to provide a reasonable expectati on of persistence (see the prev ious Â“sustainable population sizeÂ” section). Thus, using the 1-ha territory standard, I accepted habitat configurations with total area 100 ha, within a single large patch or distributed among smaller patches in a functionally connected ne twork, as potentially sustainabl e. However, it is not my intent to imply that networks consisting of multiple patches will have stability equal to a single continuous patch of equivalent area (see Lindenmayer & Lacey 1995). In fact, stochastic environmental events linked to very low survival for just one year could have devastating effects on populat ion densities in patches < 100 ha (Cumming & Willson unpublished data). Further, higher population densities and smaller territory sizes are documented in continuous forests (Willson unpublished data; Willson et al. 2004), suggesting that large continuous patches will support larger, potential ly more persistent populations. However, it is often impossible to protect large forest patches in human dominated landscapes, and in some highly frag mented landscapes patches of this size may no longer exist. Under the circumstances described above, pr otection and management of habitat in connected patch networks may be the only f easible approach for conservation. Although I assume populations in habitat configurations meeting my su stainability criteria will have reasonable potential for long-term pers istence, the possibility of extinction still exists, and may be somewhat higher in networ ks of smaller patches relative to single large patches. Thus, maintenance of connect ions that permit periodic movement among
69 spatially disjunct networks is important for maintaining genetic diversity and potential population rescue or recoloniza tion, because stochastic events may occasionally lead to population bottlenecks or extinction within patc h networks. Nonetheless, my assumption that persistence may be possible within c onnected networks is supported by data from previous studies documenting successful re production by Chucaos in relatively small forest patches (De Santo et al. 2002), a nd high occupancy levels in patches having adequate connectivity in the surr ounding matrix (Chapter 3). I selected 10 ha as the minimum-sized patch for inclusion in sustainable patch networks based on an empirical model using patch-occupancy data from a study area in the Chilean Central Valley, near the town of Osorno (hereafter Osorno; 40o35Â’ S, 73o05Â’ W). This particular area had the greate st extent and longest history of forest fragmentation occurring in the biome, and wa s an appropriate reference because my goal was to identify habitat confi gurations that would support pers istent populations in highly fragmented landscapes. The 10-ha patch size was selected because all patches 10.34 ha were predicted occupied by the Osorno model (s ee Chapter 3 for a detailed description of the study area and predictive mode l). By focusing only on patches 10 ha, I essentially ignore many smaller patches that may be o ccupied, providing demographic support to the regional population. However, th ese satellite populations (< 10 pairs) are assumed highly extinction prone due to environmental and demographic stochasticity. Thus, such populations could not persis t without frequent immigration from larger source populations. While omission of patches < 10 ha may underestimate numbers of breeding territories supported by a give n landscape configur ation, it is balanced by the rather optimistic assumed population-density of one breeding pair per hectare in all patches
70 10 ha. Although my experience in dicates that most patches 10 ha have at least some area of habitat suitable for Chucao reproducti on, habitat quality ma y be highly variable within these patches, poten tially reducing the actual numbe rs of territories supported. Matrix Composition and Connectivity As described previously, sub-populations in patches < 100 ha were assumed sustainable only within the cont ext of larger patch networks with a total cumulative area 100 ha. The patch network concept assume s adequate connectivity among patches such that exchange of individual s is sufficient to rescue d eclining populations prior to extinction, or to permit recol onization of individual patche s following local extinction (Brown & Kodric-Brown 1977). The criteria us ed to define connectivity a within patch networks were derived from observed movement behavior, and models that identified landscape-context characteristics associat ed with patch occupancy in two study landscapes. Radio telemetry data indicate that Chu cao movement is significantly constrained by open habitat in the matrix. However, they will cross short distances in the open to move among stepping stone patc hes during movement (usually 60 m), and they readily travel through wooded corrido rs and matrix habitat dominated by dense secondary vegetation (Chapter 2). Further, empirical patch-occupancy models developed for the Osorno landscape (referenced in the previous section) and a bette r-connected landscape with a shorter history of fragmentation (ChiloÃ©; 41o55Â’ S, 73o35Â’ W; Chapter 3), were remarkably consistent regarding the effect of permeable habitat in the surrounding matrix (Chapter 3). Specifically, the models predic ted occupancy of small patches if at least 22.50% (ChiloÃ©) to 27.39% (Osor no) of the surrounding matrix (within 100-m and 300-m buffers respectively) consisted of wooded or shrub dominated habitats that are permeable
71 to movement (note however that the percenta ge of shrub habitat was negligible in the Osorno landscape). Therefore, the criteria I selected to define adequate habitat connectivity specified that 25% cover (the average from the two models) by wooded or shrub dominated habitats in the matrix su rrounding a focal patch (within a 300-m radius buffer) would provide adequate habitat permeabil ity for Chucao movement into or out of the referenced patch. Alternatively, patche s with > 75% open hab itat in the surrounding matrix were assumed functionally isolated. The 300-m (rather than 100-m) buffer radius was selected to in corporate a larger landscape-context, and was appropriate because percentages of impermeable open matrix in the two buffer sizes were correlated. My assumption that movement would be possible through a matrix with as lit tle as 25% permeable habitat was supported by observed movements of radio-tagged Chucaos through sp arsely vegetated portions of the ChiloÃ© landscape (Chapter 2). Radio-tagged Chu caos also dispersed readily through wooded corridors 10 m wide and 500 m long. Thus, patches conn ected by corridors of these dimensions were also considered functiona lly connected, even if the percentage of permeable habitat in the matrix was < 25%. Although Chucaos would probably move distances > 500 m through corridors , corridors of this length we re unavailable for testing in the study landscape. Finally, to define the appropriate spa tial scale for patch networks, I set the attainable inter-patch movement distance at 600 m. At this di stance the 300-m buffers surrounding each patch (used to define matrix permeability, see above) would either abut or overlap. If the adjoining or overlappi ng buffers surrounding adjacent patches each met the previously defined habitat permeability criterion (having 25% permeable-habitat
72 cover), I assumed that connec tivity between the two patches was likewise adequate. This 600-m inter-patch distance was also appropriate with regard to Chucao movement ability because it approximately equaled the averag e movement distance (674 m Â± 606) of radiotagged subjects tracked 20 d (Chapter 2). The potential for exchange of individuals among patches up to 5 km apart (the maximu m documented movement distance; Willson 2004) is also recognized, but is considered unpredictable since I have no empirical criteria for defining adequate landscape conn ectivity at this scale (but see discussion). Sustainability Criteria Summary In summary, I predicted relatively high potential for persis tence in patches 100 ha (potentially supporting 100 breeding pairs), or in func tionally connected networks of smaller patches ( 10 ha each) with a combined area 100 ha. Functionally connected patches were defined as those with adequa te permeability in the surrounding matrix ( 25% cover by woodland or dense shrubs with in a 300-m radius buffer) lying within 600 m of one or more other patches with simila rly adequate matrix permeability, or patches connected by wooded corridors 500 m long and 10 m wide. These criteria provide an objective method for estimating the carrying cap acity of real or hypothetical landscapes, by predicting the numbers of breeding terr itories potentially accommodated within sustainable patches or patch networks (number of territori es = the total number of hectares of wooded habitat contained within all sustainable network patches combined, assuming 1-ha territory size). The criteria require minimal assumptions regarding details of population dynamics, thus many variables are ignored that may have important implications for population persistence, but that are extremely diffi cult to measure (Cumming & Willson unpublished data). However, an advantage of these cr iteria over measures based solely on habitat
73 cover or percolation thresholds is the in corporation of empirical species-specific measures of habitat permeability, movement distances, and patch size requirements. Such approaches potentially provide greater generality for cons ervation planning than more-complex models, which are difficult to parameterize and are often valid only for a single landscape (where parameterization data were collected; Raph ael & Marcot 1994; Haufler et al. 1996; Carroll et al. 2004). A critical assumption is that the 1-ha te rritory size measured in the ChiloÃ© study area also holds for more fragmented lands capes. If habitat de gradation in highly fragmented systems causes Chucaos to incr ease their territory sizes, my criteria may overestimate numbers of terr itories supported. Given su ch uncertainties, I do not recommend use of these criteria for quanti tative projection of population sizes without detailed landscape-specific investigation. Rather, the criteria provide a basis for objective comparison among altern ative conservation strategies , and as an approach for identifying dangerous levels of fragmentati on where extinction would be highly probable without conservation intervention. Finally, wh ile habitat networks designed to these specifications are predicted to support popul ations that may be less threatened by demographic and environmental uncertainties, th is design does not rule out the possibility of extinction. Thus, ongoing research and monitoring is strongly recommended. Graph Theory Format Mathematical graphs provide a useful context for representing the spatial relationships among landscape components (Can twell & Forman 1993; Keitt et al. 1997). The graph theory construct consists of sets of points (Â“verticesÂ”) and connecting lines (Â“edgesÂ”; Fig. 4-1). Using this format to apply my sustainability criteria, landscape graphs were constructed by assi gning a vertex to each patch 10 ha, then placing edges
74 between patches for which the connectiv ity criteria were met (i.e., patches 600 m apart, each having 25% permeable habitat cover in th e surrounding 300-m buffer, or patches connected by wooded corridors 500 m long). Sets of patc hes joined by edges form a Â“sub-graph,Â” which may constitute a sustainabl e network if the combined habitat area is 100 ha. The graph theory format is also usef ul for identifying particular patches (Â“articulation pointsÂ”) and thei r associated connecting elements that are important for conservation because their removal would dissect a network into smaller units (Fig. 4-1). Patches whose removal would cause the networ k to drop below the minimum habitat-area criterion may also be identified using this fr amework. Thus, the graph theory format is useful for identifying landscape elements of distinct importance for conservation, and for visualizing where habitat restora tion might be most helpful. Analysis of Test Landscapes To assess the potential sustainability of Chucao populations in existing landscapes, I applied the sustainability criteria develope d in the previous sect ions in three test landscapes located in sout h-central Chile. One 100-km2 study area was located in each of the two landscapes where patc h occupancy studies were pr eviously conducted (ChiloÃ© and Osorno, described in previous sections; Ch apter 3), and a third was located near the town of Puerto Montt (41o28Â’ S, 73o00Â’ W). Land cover maps were produced for each study-area by hand digitizing three cover t ypes (open pasture, shrubland and wooded habitats) from georeferenced panchromatic orthophotographs (1: 20000) taken during the summer of 1993 (the most recent photographs available). Wooded patches were defined as continuous areas of woodland sepa rated from surrounding patches by gaps 10 m, or connected to other patches via wooded corridors or bottle-ne cks that were at least an
75 order of magnitude narrower than the ad joining patches (see Chapter 3 for detailed descriptions of remote sensing and Ge ographic Information System methods). A graph vertex was assigned to each patch 10 ha, and graph edges were used to connect patches meeting the above referenced connectivity criteria. If the total habitat area was 100 ha, in a single large patch or summed over all patches within a connected network, the referenced patch or patch network was classified as potentially sustainable. I then estimated the carrying capacity (numbers of territories potentially supported within all sustainable patch networks) of each st udy area. Functionally isolated patches 600 m apart were also identified as locations for potential habitat restoration to re-connect the isolated patches, thereby crea ting or increasing the sizes of sustainable networks. To assess the potential conservation benefit of such efforts to enhance connectivity, I estimated the increase in the number of territories potenti ally supported if the identified landscape connections were restored. Thes e assessments demonstrated the method by which my criteria may be used to objectiv ely compare potential sustainability among landscapes with differing spatial characteristics , and to evaluate the potential benefits of conservation intervention. Results The 100-km2 study-area in the ChiloÃ© landscape consisted of approximately 45% wooded cover and 20% cover by permeable shr ub-dominated matrix. Within this area I identified 50 patches 10 ha, eight of which were 100 ha (large enough to independently support viable populations accordi ng to my criteria). Each of these eight patches, and all but five of the remaining 42 patches, were functiona lly connected to each other, forming a single large network with cumulative area potentially adequate to support 3480 breeding territories (Fig. 4-2). If restoration of habitat permeability were
76 undertaken to connect the thre e isolated patches lying with in movement-range of the network, the total area would be increased by 45 ha. The ChiloÃ© graph also contained five articulating patches whose removal woul d dissect one or more other patches away from the connected network. In general, th e landscape was highly interconnected, with more than one pathway from a ny one patch to any other. The Osorno landscape was composed of approximately 17% wooded cover and < 1% permeable shrub matrix. Within the 100-km2 study-area there were 19 patches 10 ha, but all were functionally isolated from each other, having < 25% permeable habitat in the surrounding matrix (Fig. 4-3). Most were also isolated by di stance, and none were large enough ( 100 ha) to independently support a sustainable population. If habitat management were undertaken to restore connectivity among all patches 600-m apart, five distinct networks would be formed, but none had sufficient area to fulfill the sustainability criteria. The Puerto Montt study-area had 29% w ooded cover and < 1% shrub-dominated matrix. The landscape had 52 individual patches 10 ha, but none were 100 ha and 23 were functionally isolated (F ig. 4-4). The remaining 29 patches formed 10 separate networks, but only one network, composed of eight connected patc hes, was large enough to support a sustainable populati on with a total area of 246 ha. However, restoration of connectivity could be used to increase the si ze of the referenced network and create two additional networks that, together, consisted of 34 individual patches with a total area of 782 ha. Discussion Through application of the sustainability criteria in test land scapes, I identified landscape conditions where pers istence was highly likely (ChiloÃ©), conditions where
77 regional extinction was predicted (Osorno) , and conditions where management to enhance landscape connectivity could signi ficantly increase carrying capacity (Puerto Montt). The estimated carrying capacity wa s high in the ChiloÃ© study area, which was relatively well connected with regard to C hucao movement abilitie s (Chapters 2 & 3), with 45% wooded cover and 20% cover by shrub-dominated matrix that is permeable to movement. Nearly all suitable-patches were functionally connected, forming a single large network that spanned the study area (Fig. 4-2), with habi tat potentially adequate to support up to 3480 breeding territories (a sust ainable population densit y of approximately 35 breeding pairs per km2). This estimate assumes a dens ity of one territ ory per hectare in all network patches. Howe ver, territory densities may be higher in this landscape, which has several large patc hes (> 100 ha) where habitat quality may be higher, promoting smaller territory sizes and higher densities (Willson et al. 2004). If restoration of habitat permeability were undertaken to conne ct the three isolated patches lying within movement range of the network, only 45 additi onal territories would be added to the network area. Thus, conservation action to restore connectivity may not currently be required in this landscape. Rather, at the current stage of fragmentation, conservation might be better served by efforts to protect remaining large tracts of primary forest and preserve existing landscape connect ions before they are lost. At the opposite extreme, the Osorno study area is located in the most fragmented portion of the SATR biome, with only 17% wooded cover remaining and negligible shrub-dominated or low stature secondary fore st cover. None of the 19 patches existing in this landscape was large enough to indepe ndently support a sustainable population, and all were functionally isolated. Thus, while Chucao populations are currently extant in
78 some Osorno patches (Chapter 3), my anal ysis suggests these popul ations may be on a path toward extinction due to nonviable me tapopulation sizes, even if all remaining habitat patches are protected. Further, because habitat availability in Osorno is extremely low, patch networks potentially restored th rough management of landscape connections would still be too small to support viable populations with 100 breeding pairs (FIg. 43). Therefore, assuming the sustainability criteria are valid, my results suggest that management to enhance connectivity may be ineffectual in the Osorno landscape without complementary effort to increase overall avai lability of forest ha bitat, by increasing the sizes of existing patches or restori ng forest where none currently exists. In Puerto Montt, the level of forest loss and fragmentation was intermediate between the two extremes observed in Ch iloÃ© and Osorno, with approximately 29% wooded cover remaining. Without conservati on intervention, this landscape supported a single sustainable patch networ k with total area sufficient to support up to 246 breeding pairs (approximately 2.5 pairs per km2). However, this lands cape was composed entirely of relatively small patches (< 100 ha), so actua l densities may be lower in patches that are significantly disturbed. Furthe r, the single sustainable netw ork in this landscape would likely be isolated from other sustainabl e populations without c onservation action to enhance connectivity at the regional scale. Th e most interesting result for this landscape, however, was the dramatic effect of hypothetical connectivity restorat ion. By selectively restoring habitat connections w ith 26 functionally isolated patches, it would be possible to effectively triple the estimated carrying capacity of the study area, producing three sustainable networks with sufficient area to s upport up to 782 breeding pairs (8 territories per km2; Fig. 4-4).
79 These results illustrate how application of the suggested sustainability criteria may be useful for differentiating landscapes wher e persistence is currently unthreatened, and those with dangerous levels of fragmentati on where extinction is likely. The approach also provides an objective method for eval uating hypothetical cons ervation strategies prior to committing scarce conservation dollars to these endeavors, such as identifying situations where habitat restor ation and enhanced connectivity is most critical to species conservation. Further, the gr aph-theory visualization appro ach provides a useful tool for identifying specific landscape elements that are most important for maintaining the integrity of sustainable patch networks, a nd areas where targeted restoration might achieve the best results for conservation. Further research is needed to identify the potential threshold le vel of fragmentation at which point persistence of Chucao populati ons is threatened. However, I suspect the threshold lies somewhere between 20% and 30% habitat cover, consistent with previous suggestions that fragmentation effects may be strongest when amount of suitable habitat falls within a range of 10% to 30% (AndrÃ©n 1994; Fahrig 1998). I base my prediction on the precipitous drop in densities of sustaina ble territories predicted for the three test landscapes. As described prev iously, the sustainable number of territories was high in ChiloÃ© (with 45% wooded cover) and extincti on appeared likely in Osorno (with 17% cover), while the Puerto Montt study area (w ith 29% cover) appeared on the verge of becoming functionally disconnected. Without conservation intervention, the Puerto Montt study area had only a single eight-patch sustainable network, and if either of two articulating patches (or their associated connecting elements ) were removed, the network would be split and the number of sustainabl e populations would drop to zero. Further,
80 my analysis indicated that re storation of connectivity could effectively triple the number of territories supported, pointing to disrupt ed connectivity as the factor potentially limiting persistence rather than inadequate habitat area per se . Restoring Landscape Connectivity My analysis of one possible conserva tion strategy, restor ation of habitat connections among functionally isolated patches (using habitat corridors or other interventions to enhance matrix permeability) suggests the approach may be highly advantageous in some circumstances, but potentially superfluous or dangerously inadequate in others. These results illustrate the importan ce of recognizing circumstances under which this particular approach will pr oduce the desired outcome. When the total habitat area is high, a lternative strategies such as protection of large patches and proactive conservation of ex isting landscape connections, ra ther than restoration of connections that have been lost, may be a better strategy. Alternatively, in landscapes with extremely low levels of habitat cover, increased landscape c onnectivity is unlikely to ensure persistence without concurrent effo rts to increase the total habitat area. The challenge to practitioners, then, is identifying the circumstances under which enhanced connectivity will achieve the desired conservation goals. In most cases, voluntary or low-cost efforts to enhance connectivity (e.g., protecting fence-line and riparian vegetation as movement paths) may be helpful. This is especially true when resources and data for focused planning are l acking. However, large investments in land management to increase connectivity carries opp ortunity costs, which may preclude other conservation measures (Simberloff et al. 1992; Rosenberg et al. 1997; NiemelÃ¤ 2001). This underscores the importance of carefully evaluating the potential outcomes of any such conservation intervention, wi thin the context of cost benefit analysis that considers
81 alternative approaches. It is also of highest importance that managers consider the adequacy of proposed interventions, sin ce enhanced connectivity cannot ensure persistence of a targeted species if the total ar ea of habitat is inade quate to support viable populations. When it is determined that intervention to restore connectivity is the appropriate course of action, results of pr evious studies suggest a few pot entially viable alternatives for land-use management and strategic re-veg etation (Chapter 2). These include (1) protection or restorati on of wooded corridors ( 500 m long and 10 m wide) to connect isolated patches, (2) management to incr ease density of second ary vegetation in the matrix to provide cover that encourages an imal movement, and (3) provision of patches less than approximately 60 m apart within th e deforested matrix as movement stepping stones. Further, although my research di d not address large-sc ale linkages directly, extrapolating from current understanding suppor ts some speculative insights regarding types of linkage that may provide relativ ely high functional success for long distance movement among disjunct reserves or patch networks. The logic regarding such linkages is as follows. The longest observed movement by a radio tagged subject within a 24-hr pe riod through discontinuous habitat was 1400 m (Chapter 2). Thus, it is likely that a Chucao could also move at least this distance directly due to absence of long corridors in the study la ndscape. However, to enhance probability of successful movement through long corridors , I suggest a minimum width of 110 m. I arrived at this suggested widt h as follows. Chucao territo ry size has been estimated at approximately 1 ha (Willson unpublished data) and individual patche s of approximately 1-ha are sometimes occupied and occasiona lly used as breeding territories (T.D.C.,
82 personal observations). Thus, I will assume that 1 ha represents the minimum habitat area needed to support a func tional demographic unit. A circle is the most compact shape possibl e for a habitat patch, which is typically viewed as the optimal shape for conservation because it minimizes interaction with the surrounding (potentially detrimental) matrix. I assumed that if a 1-ha circular patch is adequate to support a breeding territory, then, at a minimum, a 1-ha circular patch should meet the subsistence needs of a dispersing individual. If multiple 1-ha patches were lined-up contiguously, the resul ting corridor should be adequa te for both subsistence and movement (Fig. 4-5a). A corridor of patche s with these dimensions would have a width of approximately 110-m. It is assumed th at Chucaos would be able to move freely through such a corridor, and could subsist th ere if necessary, potentially establishing breeding territories. Applying these same principles, I assu me that a 350-m wide corridor (wide enough to support a row of contiguous 10ha circular patches; Fig. 4-5c) should provide for persistence of a continuous population across a fragmented landscape to connect isolated networks or reserves at a regional scale. The justification for this assumption is that a 10-ha patch is expected to support a sustainable population if there is adequate connectivity with other occupied patches (see the methods section detailing development of our sust ainability criteria). Alternatively, if my sustainability criteria were taken to their logical extreme, a minimal sustainable network of corridors and patches would consist of ten 10-ha patches connected by nine 500-m long corridors. Th eory predicts that flow rates among all elements would be highest if the patches we re maximally clustered, for example with a single central patch connected to the nine satellite patches vi a corridors radiating from the
83 central patch (e.g., Cantwell & Forman 1994). However, I have no empirical evidence indicating that a linear arrange ment of patches and corridors (Fig. 4-5 b) would fail to sustain long term persistence, since many occupied patches in the study landscapes occurred in linear configurati ons along riparian corridors. Thus, there is potential value in use of linear networks to create linkages among networks or reserves separated by large distances. Of course, these recommenda tions are largely speculative, but they are founded on empirical evidence a nd well-justified eco logical theory (e.g., Verboom et al 2001). When very long corridors are needed, I recommend greater widths, which I will refer to as Â“long-distanceÂ” linkages. Such connections have been proposed for the study region to link forests of the Andean slopes w ith those of the Coastal Range via corridors spanning the Chilean Central Valley (Wor ld Wildlife Fund internal report; Keitt unpublished data; Sieving unpublished data). I suggest that such linkages should have a minimum width of 1130 m, large enough to accommodate a row of contiguous 100-ha patches (Fig. 4-5d). Patches 100 ha are predicted to s upport sustainable populations requiring only minimal immigration (Verboom et al 2001). Thus, contiguous patches in this size range should provide for long-term su stainability of a continuous population that would enhance connectivity over large geogr aphic areas and provide a large source population for immigration into smaller adjacent patches occurring along the corridor length. The linkages I have suggested may be modifi ed and combined in a variety of ways, in response to local needs and constraints, fo r integration into regi onal-scale conservation designs. For instance, longdistance linkages and conti nuous populations (Fig. 4-5c,d)
84 might be used to connect multiple sustainabl e networks that are geographically isolated from each other, or to detour around areas wh ere habitat preservati on and restoration are impractical (e.g., heavily urbanized areas). In less Â“builtupÂ” landscapes where potentially sustainable networks exist (or coul d feasible be restored ), combinations of smaller-scale linkages (Fig. 4-5a,b) might be us ed to connect otherwise isolated patches. It should be noted, however, that the co rridor concept is still debated in the conservation literature (Hobbs 1992; Simberlo ff et al. 1992; Beier & Noss 1998; Haddad et al. 2000; Noss & Beier 2000; Proche et al. 2005), particularly with regard to potential negative consequences for conservation. Of pa rticular concern are indirect effects on species competition and pred ator-prey interact ions (Orrock & Damschen 2005; Proche et al. 2005). Thus, the potential for negativ e effects should be considered prior to initiating large-scale corridor development. A significant threat to bird conservation in the SATR biome may be the facilitated spread of an introduced predator, the American mink ( Mustela vison ), along riparian corridors (Iriar te et al. 2005; Sieving unpublished data). However, invasive species are inhere ntly good dispersers, so while their spread may be facilitated by corridors, it will unlikely be controlled by lack of corridors. Thus, while modifications to corridor designs (e .g., in the case of mink, focusing on corridors that do not follow river courses) may reduce the threat of invasion. Nonetheless, as pointed out by Levey et al. (2005b) , the issue practitioners face is assessing the degree to which conservation will benefit from corridor development, relative to the potential financial and ecological costs. Conclusions Due to rapid intensification of comme rcial land-uses in the SATR biome, successful conservation will likely depend on design of landscapes that retain native biota
85 along side economically productive land uses. My analysis of landscape patterns in a portion of the biome with high human population density (Osorno) indicates that habitat loss will likely outstrip the sust ainability requirements of many forest specialists, such as the Chucao Tapaculo, in portions of the biome where human land use is intensifying (Chapter 3). If forest conversion proceeds in other parts of the biome as it has in Osorno, sustainable habitat configurat ions will probably be maintained only where they are actively managed, and the species will lik ely be extirpated elsewhere. The landscape criteria suggested here are empirically based and spatially explicit, providing quantifiable planning targets for cons ervation practitioners in the biome. The suggested linkage designs and graph-theory visualization approach provide a set of building blocks for efficiently developing cons ervation strategies, which can be evaluated objectively in terms of economy and potential for success. Nonetheless, it should be recognized that potential differences in habitat quality are not explicit ly addressed by this approach. Thus, field assessment of specific sites will still be required to confirm their adequacy, and intervention to restore or ma intain habitat quality may be necessary. Finally, the criteria and linkage s I suggest are scaled to th e specific requirements of the Chucao Tapaculo. Although I pr opose the Chucao as a useful surrogate for planning the connectivity component of landscape design to meet the needs of ma ny forest vertebrates (including but not limited to the group of e ndemic tapaculos), conservation plans derived from these criteria should be examined with re gard to the needs of other SATR species of conservation concern.
86 24 ha 18 ha 14 ha 51 ha Disconnected patch: isolated by distance Articulating patches Disconnected patch: inadequate permeability in matrix Potential for restored connection Connected patchesArticulating patches Disconnected patches Figure 4-1. Example landscape graph containing six habitat patches. The four patches joined by solid lines (gra ph Â“edgesÂ”) form a single cl uster because the graph edges form a path connecting all four patches. If any one of these edges were removed, the cluster would be broken in to two sub-graphs. The two central patches are termed articulating patche s because removal of either would likewise break the cluster. The 51 ha patch is also significant to the sustainability criteria because removal of this patch woul d reduce the total combined habitat area of the cluster below the 100-ha minimum network-size criteria. Each of the two disconnected patches form individual (one-patch) sub-graphs. The patch in the upper porti on of the landscape is isolated by distance (the 300-m radius buffer does not overlap the buffer of another patch). The patch in the lower left corner lies within the appropriate movement-distance (600 m), but is isolat ed due to insufficient coverage of permeable habitat in the surrounding buffer. This patch could potentially be joined with the adjacent cluster th rough habitat management to restore connectivity (indicated by the dashed line).
87 Connected patchesArticulating patchesDisconnected patches Figure 4-2. ChiloÃ© study area landscape gra ph including 50 individua l wooded patches, 45 of which form a single highly-connect ed sub-graph (joined by solid lines) defined as a sustainable patch network acco rding to the sustainability criteria. The network has five articulating patche s whose removal would dissect one or more other patches away from the re st of connected landscape. Of the remaining patches, two are isolated by distance and three ar e disconnected due to insufficient permeable habitat in th e surrounding buffer area. Management of habitat structure in the matrix could potentially be used to join the latter three patches to the connected ne twork, indicated by dashed lines.
88 N Connected patchesArticulating patchesDisconnected patches Figure 4-3. Osorno study area landscape graph showing 19 f unctionally isolated wooded patches, each having < 25% permeable habitat cover in the surrounding matrix. None of the patches or patch clusters met the criteria for sustainability under current conditions. Even if mana gement of habitat structure in the matrix were used to join patches lyin g within movement distance from each other (potential connections indicated by dashed lines), none of the patch clusters would have sufficient total habitat area ( 100 ha) needed to support a sustainable population.
89 N Connected patchesArticulating patchesDisconnected patches Figure 4-4. Puerto Montt study area landscape graph including 52 indi vidual patches. Nineteen patches are functionally co nnected, either by a wooded corridor 500 m long, or having 25% permeable habitat cover in the surrounding matrix, forming 10 connected networks (j oined by solid lines). However, only one network, composed of eight connect ed patches (circled, upper right), is large enough to support a sustainable popul ation. Of the remaining patches, 18 are disconnected due to insufficien t permeable habitat in the surrounding matrix, but are close en ough to other patches ( 600 m) that management of habitat structure in the matrix could be used to create functional connections (indicated by dashed lines). Such mana gement could be used to increase the size of existing networks and form additional networks. Three of these (circled) would be large enough to potentially support sustainable populations after management to restore connectivity.
90 10 ha 1 ha 110 m wide 1400 m 500 m 10 m wide 350 m wide d. Long-distance linkage b. Linear network (narrow corridor) a. Linear network (wide corridor) c. Continuous population1130 m wide 10 ha 100 ha Figure 4-5. Suggested design of large-scale corridors for re gional connection of patches and patch networks. The first design (a ) uses 110-m wide corridors (equal in size to strings of c ontiguous 1-ha circular patche s) to connect habitat patches 1400 m apart (the longest observed movement within a 24-h period). The second design (b) represents a linear network of patche s predicted to support a sustainable population, if at least 10 patches 10 ha are connected by corridors 500 m long. The third design (c ) represents a 360-m wide corridor, equal to a conti guous string of 10 ha patches, predicted to support a continuous sustainable population. The fourth design (d) represents a largescale 1130-m wide Â“long-distance li nkageÂ” corridor, wide enough to accommodate a row of contiguous 100-ha patches, each large enough to independently support a su stainable population.
91 CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS Habitat fragmentation is now imposing spat ial structure on populations of wildlife species throughout the world. Many of these species originally ex isted in more-or-less continuous habitats, and their evolutionary le gacy does not equip them for the level of habitat heterogeneity they currently face. It is generally accepted that fragmentation can lead to regional extinction of sensitive species once remaining habitat patches become too small to support persistent populations inde pendently, and once patc h isolation exceeds speciesÂ’ specific dispersal capacities (Hanski 1998). However, the stages at which fragmentation effects become detrimental to i ndividual species are high ly variable. This makes understanding population responses to sp atial heterogeneity more important than ever for conservation reasons. In landscapes where formerly contiguous habitat is fragmented, conservation strategies rely on sufficient levels of move ment among habitat patches to rescue small populations from eminent extinction, or to permit recolonization once extinction has occurred (Brown & Kodric-Brown 1977; Hans ki 1998). Thus, maintenance of landscape connectivity has become a major focus of conservation planning (Forman & Godron 1986; Mann & Plummer 1993; Rosenberg et al . 1997). However, empirical data regarding habitat conditions required by dispersing animals are scarce (Ricketts 2001). Further, conservation action to increase c onnectivity may be costly, and it has been suggested that alternativ e strategies, such as protection of larger areas of primary habitat, may be a better use of conservation resources (e.g., Simberloff et al. 1992). To determine
92 when conservation action to enhance connect ivity is warranted, it is important to understand the link between movement and pe rsistence of populat ions at local and regional scales, and to unders tand the factors that determin e movement probabilities. Dissertation Summary To detect responses to area and isolati on, studies must be broad enough in spatial extent to include vacant patches of suitable habitat and a wide range of patch sizes and degrees of isolation (Hanski 1994). For this reason, fragmentation research is usually conducted at scales that are not conducive to investigating de tails of individual behavior (Lima & Zollner 1996). To address responses to habitat fragmentation at both scales, I conducted a multi-phase research project. Firs t, I experimentally tested behavioral responses of the Chucao Tapaculo ( Scelorchilus rubecula ) to common matrix-habitat components, then assessed broad scale populatio n-level effects of constrained movement on patterns of patch occupanc y at the landscape scale. The translocation experiment directly tested the relative permeability of three landscape elements (open habitat, shrubby s econdary vegetation, and wooded corridors) to Chucao movement. The numbers of days s ubjects remained in release patches prior to initial movement (a measure of habitat re sistance) was significantly longer for patches surrounded by open habitat than for patches ad joining corridors or surrounded by dense shrubs. This demonstrated significant beha vioral resistance to dispersing through open habitat. Further, their willingness to travel through wooded corridors and shrubdominated matrix indicated that both approach es (corridor protecti on and maintenance of dense vegetative cover in the matrix to enc ourage movement) may be equally viable for maintaining connectivity. This is relevant for conservation because management of
93 matrix vegetation structure may sometimes be more feasible than fully restoring wooded habitat patches or corridors. Results also provided information on attain able movement distances. For example, Chucaos routinely crossed open habitat gaps 20 m wide, but were reluctant to cross gaps 60 m, and rarely crossed gaps 80 m. In contrast, the average movement distance through dense shrubs was 100 m, and two subjects easily dispersed 500 m through a wooded corridor (the longest corridor tested ). While these distances probably do not represent movement limits for the species, th ey do provide benchmarks for conservation planning that were previously unavailable. Finally, I noted regular use of small steppingstone patches by translocated bi rds, suggesting availability of such patches might also encourage movement in fragmented landscapes. Although the translocation experiment pr ovided strong evidence that open habitat in the matrix formed a barrier to Chucao movement, it was not clear how significant this barrier effect might be for long-term persis tence at a regional scale. To determine whether barriers to movement caused by th e inhospitable open-habitat matrix would affect Chucao distribution at the landscape scale, I conducte d patch occupancy studies in two landscapes that differed in both degree a nd duration of forest fragmentation. Within each landscape, absence of Chucaos from a par ticular patch was assumed to reflect localscale patch characteristics, and characterist ics of the surrounding ma trix, that potentially inhibit long-term population viability. In the ChiloÃ© study area, located on northern ChiloÃ© Island (41o55Â’ S, 73o35Â’ W), land-use conversion began in the early 1900Â’s an d, at present, approximately 45% of the study area remains wooded. In the other study area, located on the mainland near the city
94 of Osorno (40o35Â’ S, 73o05Â’ W), deforestation was initia ted 50 to 80 years earlier, and habitat loss in this area is now extensive, with only approximately 17% of the wooded habitat remaining. In reference to the Ch iloÃ© landscape, observa tions suggest that movement is sometimes inhibited but, for the most part, the landscape still appears relatively well connected in relation to C hucao movement (most patches isolated by 80 Â– 100 m of open habitat; Chapters 2 & 3). In addition, a large portion of the deforested area is covered by dense shrubby vegetation (a pproximately 20%), which is relatively permeable to Chucao movement. Further, with an average patch size of 6.77 ha (Â± 87.38), most patches in ChiloÃ© were still large enough to suppor t several breeding territories. Many small patches that are currently occupied in the ChiloÃ© landscape may function as sink habitat, in th at the populations are too sma ll to persist without frequent immigration from larger source populations. Nonetheless, these patches may serve to augment the regional population size and ma y produce reproductive surplus in some years, further increasing the flow of dispersing individuals through the landscape. Thus, in ChiloÃ©, most patches are currently o ccupied and population structure appears to represent a single popula tion that is patchily distributed. In Osorno, however, habitat conditions were largely inadequate for long term sustainability of Chucao populations. With little wooded habitat remaining and a mean patch size of 1.44 ha (Â± 6.22), most patches < 10 ha in Osorno were unoccupie d, and population structur e was indicative of a (possibly non-equilibrium) meta population (Chapter 3). Classification tree models developed for the ChiloÃ© and Osorno study areas provided relatively accurate predictions of patch occupancy within the corresponding
95 landscapes (Chapter 3). Howeve r, predictions of the two models differed from each other considerably regarding the threshold size above which all patches were predicted occupied. The ChiloÃ© model predicted occ upancy of all patches > 3.80 ha, while the corresponding threshold in Osorno occurred at a patch size of 10.34 ha. Despite this discrepancy, the two models were remarkably consistent regarding the influence of open habitat in the surroundin g matrix. Specifically, patches with adequate habitat quality that were below the referenced threshold sizes were predicted occupied only if the immediately surrounding matrix consisted of < 72.61% (Osorno) to 77.50% (ChiloÃ©) open habitat. In both study areas, Chucaos were absent fr om isolated patches, consistent with hypothesized population-level e ffects of movement limitation. However, patch occupancy also declined as the number of y ears since initiation of the fragmentation process increased, revealing a possible exti nction time-lag following patch isolation (extinction debt; Tilman et al. 1994). Chucao demographi c data are suggestive of an extinction debt because reproductive success in forest patches is re latively high (De Santo et al. 2002). Thus, small populations could potentially persist in effectively isolated patches for many generations fo llowing fragmentation (via in situ reproductive output), although long-term persistence of small populati ons is unlikely due to environmental and demographic stochasticity. De layed extinction would be most likely in relatively large patches, which presumably support larger populat ions that are less prone to extinction in the short-term. While it was impossible to disso ciate the interacting e ffects of degree and duration of fragmentation, I expect similar reductions in patch occupancy to occur throughout the biome as forest loss and fragmentation proceed. However, the rate of land
96 cover change may be faster in landscape s currently undergoing fragmentation than occurred historically because technological a dvances have provided more efficient means of deforestation now than in the past. Based on results of the translocation expe riments, patch occupancy studies, model predictions for similarly sized vertebrates, and a few biol ogical assumptions, I developed a set of simple criteria to distinguish ha bitat configurations with reasonably high expectancy of supporting pers istent Chucao populations (C hapter 4). I predicted relatively high potential fo r persistence in patches 100 ha (supporting up to 100 breeding pairs), or in well connected networks of smaller patches ( 10 ha each) with a combined area 100 ha. Well connected patches were defined as those with 25% cover by permeable habitat types (i.e., w ooded habitat and dense shrubs) in the surrounding matrix, and lying within 600 m of one or more patches with similarly adequate permeability in the surrounding matrix, or patches that were connected via wooded corridors 500 m long. These criteria were used to predict the carrying capacity (numbers of breeding territories potentially supported within sustainable networks) of referenced landscape areas, thereby providi ng an index for comparison among real or hypothetical landscapes. I used these criteria to as sess the potential fo r persistence of C hucao populations in 100-km2 study areas in three test landscapes lo cated in ChiloÃ©, Osorno, and Puerto Montt (41o28Â’ S, 73o00Â’ W). I further assessed the poten tial benefit of conservation action to increase landscape connectivity in each lands cape, measured as the increase in numbers of breeding territories supporte d under the Â“restored connectiv ity scenarioÂ” compared to conditions without conservati on intervention. In the ChiloÃ© study area most of the 50
97 suitable patches were functiona lly connected, forming a single large network with habitat area adequate to support up to 3480 breeding te rritories. If restoration of habitat permeability were undertaken to connect the three isolated patches lying within movement range, 45 additional territorie s could potentially be supported. At the opposite extreme, in the Osorno la ndscape, none of the pa tch configurations were predicted sustainable, either with or without management to increase connectivity. In the Puerto Montt landscape, however, with an intermediate fragmentation level, one sustainable patch network (eight connected pa tches) was identified that could potentially support up to 246 breeding pairs. However, re storation of connectiv ity could potentially increase the sustainable populat ion size in that landscape to 782 breeding pairs, within 34 patches forming three sustainable networks. Thus, our criteria al lowed us to identify landscape conditions where pers istence was highly likely wi thout habitat restoration (ChiloÃ©), conditions where ex tinction was probable regardle ss of efforts to enhance connectivity (Osorno), and condi tions where restoration of c onnectivity could potentially triple the sustainable populati on size (Puerto Montt). These criteria provide a relatively simple tool for assessing poten tial sustainability in different landscape configurations, thereby providing a basis for objective comparison among alternative conservation strategies. Conservation Implications and Recommendations In the South American temperate rainfo rest biome, successful conservation may depend on our ability to predict fragmenta tion effects and take preemptive action to prevent local, regional, and ultimately globa l extinction of endemic biota. Currently, development pressure is strong throughout the biome and further deforestation is eminent. Since protected areas cover only a ti ny fraction of the forests, these areas alone
98 will be insufficient to maintain the diversity of forest dwelling species. Thus, it is clear that conservation must be accomplished within the context of a highly fragmented and densely populated landscape. Large-scale conservation planning is underway (World Wildlife Fund internal report), but basic ecology of most species in the biome is currently unknown. Therefore, data are urgently needed to provide guidelines for conservation decision-making. To date, conservation in fr agmented landscapes has focused largely on protection or restoration of vegetated corridors, wh ich are thought to provide passageways for movement among otherwise isolated patches (Forman & Godron 1986; Mann & Plummer 1993; Fahrig & Merriam 1994; Rose nberg et al. 1997). Translocation experiment results suggests that wooded corridors and shrubby vegetation function similarly as movement habitat for dispersi ng Chucaos. Thus, these elements may be similarly viable for use in landscape manageme nt to enhance connectivity. However, my analysis suggests that conservation benefits potentially derived from efforts to enhance connectivity may be highly advantageous in some circumstances, but ineffectual in others. Chucao populations in ChiloÃ© (and other la ndscapes with levels and duration of fragmentation similar to ChiloÃ©) probably are not in imminent danger of extinction. Therefore, conservation action to increase connec tivity may not be necessary at this time. Instead, it may be more advantageous to fo cus conservation efforts on ensuring long term protection of large tracts of primary forest . Landowners in ChiloÃ© currently manage properties in ways that main tain wooded corridors and shr ub-fields, which appear to provide functional connections among remnan t patches. Such habitat management by
99 landowners is widely recognized as a critical part of conservation strategies for vulnerable species (Wilcove & Lee 2004). Therefore, landowne r-based incentives (educational and economic) may provide a mean s for implementing the scale-appropriate conservation efforts. Voluntary or opportuni stic efforts at corridor protection, prior to further deforestation, may be helpful as a proactive conservation approach, although potential negative affects (e.g., spre ad of invasive species; Proche et al. 2005) should be considered. In contrast, landscapes with fragmentati on levels similar to Osorno may require immediate conservation action to prevent se rious population declin es or extinction. Although Chucao populations are currently extant in some Osorno patches (Chapter 3), my analysis suggests these populations may be on a path toward extinction due to nonviable metapopulation sizes, ev en if all remaining habi tat patches are protected (Chapter 4). Further, becaus e habitat availability in Osor no is so low, patch networks potentially created through restored connections would still be too small to support viable populations. Therefore, assuming the sustainab ility criteria are valid, my results suggest that management to enhance connectivity may be ineffectual in the Osorno landscape without complementary effort to increase the sizes of existing patches or restore wooded patches where none currently exist. In Puerto Montt, however, where the leve l of forest loss an d fragmentation was intermediate between the two extremes observed in ChiloÃ© and Osorno, a single sustainable patch network was identified that could potentia lly support a viable population. However, the conservation status of this landscape c ould be dramatically increased by conservation action to restor e landscape connectivity. Thus, my results
100 illustrate the importance of recognizing circumstances under which a particular conservation strategy will produ ce the desired outcome. Based on my research, I suggest that large investments in land management to increase connectiv ity may best serve conservation if reserved for landscapes with characteristics in which increased connectivity will produce the greatest proportio nal effect. In landscapes with lower levels of cover and contiguity, alternative management strategies, such as efforts to increase the overall percent c over of habitat, would also be required and, without this effort, corridors would probably be of little use.
101 LIST OF REFERENCES Aars, J., and R. A. Ims. 1999. The effect of habitat corridors on rates of transfer and interbreeding between vole demes. Ecology 80 :1648-1655. Allen, C. R., L. G. Pearlstine, and W. M. Kitchens. 2001. Modelling viable mammal populations in gap analysis . Biological Conservation 99 :135-144. AndrÃ©n, H. 1994. Effects of habitat fragment ation on birds and ma mmals in landscapes with different proportions of su itable habitat: a review. Oikos 71 :355-366. Arendt, R. 2004. Linked landscapes Creating greenway corridors th rough conservation subdivision design strategies in the northeastern a nd central United States. Landscape and Urban Planning 68 :241-269. Armesto, J. J., R. Rozzi, C. Smith, and M. T. K. Arroyo. 1998. Conservation targets in South American temperate forests. Science 282 :1271-1272. Bakker, V. J., and D. H. Van Vuren. 2004. Ga p-crossing decisions by the red squirrel, a forest-dependent small ma mmal. Conservation Biology 18 :689-697. Balmford, A., and A. Long. 1994. Avian e ndemism and forest loss. Nature 372 :623-624. Batschelet, E. 1981. Circular statistics in biology. Academic Press, New York. Beier, P. 1995. Dispersal of juvenile cougars in fragmented habitat. Journal of Wildlife Management 59 :228-237. Beier, P., and R. F. Noss. 1998. Do habitat corridors provide conn ectivity? Conservation Biology 12 :1241-1252. BÃ©lisle, M., and A. Desrochers. 2002. Gapcrossing decisions by forest birds: an empirical basis for parameterizing spatia lly-explicit, individual-based models. Landscape Ecology 17 :219-231. BÃ©lisle, M., A. Desrochers, and M. J. For tin. 2001. Influence of forest cover on the movements of forest birds: a homing experiment. Ecology 82 :1893-1904. BÃ©lisle, M., and C. St. Clair. 2001. Cumulativ e effects of barriers on the movements of forest birds. Conservation Ecology 5: 9. Available from the internet URL:http//www.consecol.org/vol 5/iss2/art9. (accessed June 2005).
102 Berggren, A., B. Birath, and O. Kindvall. 2002. Effect of corridors a nd habitat edges on dispersal behavior, movement rates, and m ovement angles in Roesel's Bush-Cricket ( Metrioptera roeseli ). Conservation Biology 16 :1562-1569. Bowman, J. 2003. Is dispersal distance of bird s proportional to territory size? Canadian Journal of Zoology 81 :195-202. Bowman, J., and L. Fahrig. 2002. Gap crossing by chipmunks: an experimental test of landscape connectivity. Cana dian Journal of Zoology 80 :1556-1561. Bowne, D. R., J. D. Peles, and G. W. Barre tt. 1999. Effects of landscape spatial structure on movement patterns of the hispid cotton rat ( Sigmodon hispidus ). Landscape Ecology 14 :53-65. Breiman, L., J. H. Friedman, R. A. Olsh en, and C. J. Stone. 1984. Classification and regression trees. Wadsworth, Belmont, California. Bright, P. W. 1998. Behavior of specialist spec ies in habitat corrido rs: arboreal dormice avoid corridor gaps. Animal Behaviour 56 :1485-1490. Brooker, L. C., M. G. Brooker, and P. Ca le. 1999. Animal dispersal in fragmented habitat: measuring habitat connectivity, corridor use, and dispersal mortality. Conservation Ecology 3 :4. Available from the internet URL:http//www.consecol.org/vol3/iss1/art4 . (accessed June 2001). Brown, J. H., and A. Kodric-Brown. 1977. Turnover rates in insular biogeography: effects of immigration on extinction. Ecology 58 :445-449. Buddle, C. N., and A. L. Rypstra. 2003. Factors influencing emigration of two wolf spider species (Araneae: Lycosidae) in an agroecosystem. Environmental Entomology 32 :88-95. Cantwell, M. D., and R. T. T. Forman. 1993. Landscape graphs Â– ecological modeling with graph-theory to detect configur ations common to diverse landscapes. Landscape Ecology 8 :239-255. CastellÃ³n, T. D., and K. E. Sieving. 2006. An experimental test of matrix permeability and corridor use by an endemic und erstory bird. Cons ervation Biology 20 :135-145. Carroll, C., R. F. Noss, P. C. Paquet, a nd N. H. Schumaker. 2004. Extinction debt of protected areas. Conservation Biology 18 :1110-1120. Caughley, G. 1994. Directions in conserva tion biology. Journal of Animal Ecology 63 :215-244. Collar, N. J., L. P. Gonzaga, N. Krabbe, A. MadroÃ±o Nieto, L. G. Naranjo, T. A. Parker, and D. C. Wege. 1992. Threatened birds of the Americas: the ICBP red data book, part 2. International Council for Bird Preservation, Cambridge, United Kingdom.
103 Cornelius, C., H. CofrÃ©, and P. A. Marque t. 2000. Effects of ha bitat fragmentation on bird species in a relict te mperate forest in semiarid Chile. Conservation Biology 14 :534-543. Davis, S. D., V. H. Heywood, O. Herrera-MacB ryde, J. Villa-Lobos, and A. C. Hamilton, editors. 1997. Centers of plan t diversity: a guide and stra tegy for their conservation (Vol. 3, The Americas). World Wildli fe Fund for Nature and the World Conservation Union of Natu re, Cambridge, United Kingdom. De'ath, G., and K. E. Fabricius. 2000. Cla ssification and regressi on trees: a powerful yet simple technique for analysis of complex ecological data. Ecology 81 :3178-3192. De Santo, T. L., M. F. Willson, K. E. Si eving, and J. J. Armesto. 2002. Nesting biology of tapaculos (family Rhinocryptidae) in fr agmented south-temperate rainforests of Chile. Condor 104 :482-495. Desrochers, A., and S. J. Hannon. 1997. Gap crossing decisions by forest songbirds during the post-fledging period. Conservation Biology 11 :1204-1210. Diamond, J. M. 1975. The island dilemma: lessons of modern biogeography studies for the design of natural reserv es. Biological Conservation 17 :129-145. Dias, P. C. 1996. Sources and sinks in population biology. Trends in Ecology and Evolution 11 :326-330. Donoso, C. 1993. Bosques templados de Chile y Argentina: variac iÃ³n, estructura, y dinÃ¡mica. Editorial Universitaria, Santiago, Chile. Donoso, C., and A. Lara. 1995. UtilizaciÃ³n de los bosques nativos en Chile: pasado, presente y futuro. Pages 363-404 in J. J. Armesto, C. VillagrÃ¡n, and M. K. Arroyo, editors. EcologÃa de los bosques nativos de Chile. Editorial Univ ersitaria, Santiago, Chile. ESRI (Environmental Systems Research Institute). 1999. ArcView 3.2: Geographic Information System Software. ES RI, Redlands, California. Fahrig, L. 1998. When does fragmentation of breeding habitat affect population survival? Ecological Modelling 105 :273-292. Fahrig, L., and G. Merriam. 1994. Conservati on of fragmented popul ations. Conservation Biology 8 :50-59. Ford, H. A., G. W. Barrett, D. A. Saunders , and H. F. Recher. 2001. Why have birds in the woodlands of Southern Australia declined? Biol ogical Conservation 97 :71-88. Forman, R. T. T., and M. Godron. 1986. La ndscape ecology. John Wiley and Sons, New York, New York.
104 Franklin, J. F. 1993. Preserving biodivers ity: species, ecosystems, or landscapes? Ecological Applications 3 :202-205. Franklin, J. F. 1998. Predicting the distributi on of shrub species in southern California from climate and terrain-derived vari ables. Journal of Vegetation Science 9 :733748. Glade, A., editor. 1988. Red book of Chilean terrestrial vertebrates. CONAF, Santiago, Chile. Gobeil, J. -F., and M. -A. Villard. 2002. Pe rmeability of three boreal forest landscape types to bird movements as determined from experimental translocations. Oikos 98 :447-458. Greenwood, P. J., and P. H. Harvey. 1982. The natal and breeding dispersal of birds. Annual Review of Ecology and Systematics 13 :1-21. Grubb, T. C., Jr., and P. F. Doherty, Jr . 1999. On home-range gap-crossing. Auk 116 :618-628. Guisan, A., and N. E. Zimmermann. 2000. Pred ictive habitat distri bution models in ecology. Ecological Modelling 135 :147-186. Gustafson, E. J., and L. Hansson. 1997. Corri dors as a conservation tool. Ecological Bulletin 46 :182-190. Haas, C. A. 1995. Dispersal and use of co rridors by birds in wooded patches on an agricultural landscape. Conservation Biology 9 :845-854. Haddad, N. M. 1999a. Corridor and distance effects on interpatch movements: a landscape experiment with butterf lies. Ecological Applications 9 :612-622. Haddad, N. M. 1999b. Corridor use predicted from behaviors at habitat boundaries. American Naturalist 153 :215-227. Haddad, N. M., D. R. Bowne, A. Cunningham, B. J. Danielson, D. J. Levey, S. Sargent, and T. Spira. 2003. Corridor use by diverse taxa. Ecology 84 :609-615. Haddad, N. M., D. K. Rosenberg, and B. R. Noon. 2000. On experimentation and the study of corridors: response to Beier and Noss. Conservation Biology 14 :15431545. Hanski, I. 1994. A practical model of me tapopulation dynamics. Journal of Animal Ecology 63 :151-162. Hanski, I. 1998. Metapopulation dynamics. Nature 396 :41-49.
105 Harrell, F. E. 2001. Regression modeling strategi es: with applications to linear models, logistic regression, and survival anal ysis. Springer, New York, New York. Harris, L. D., and J. Scheck. 1991. From imp lications to applications: the dispersal corridor principal a pplied to the conservation of biological diversity. Pages 189-220 in D. A. Saunders, and R. J. Hobbs, editors. Nature conservation 2: the role of corridors. Surrey Beatty and S ons, Chipping Norton, Australia. Harrison, S., and A. D. Taylor. 1996. Empi rical evidence for metapopulation dynamics. Pages 27-42 in I. Hanski, and M. E. Gilpin, editors. Meta population dynamics: ecology, genetics and evolution. Academic Press, San Diego, CA. Haufler. J. B., C. A. Mehl, and G. J. Rolo ff. 1996. Using a coarse-f ilter approach with species assessment for ecosystem mana gement. Wildlife Society Bulletin 24 :200208. Hein, S., J. Gombert, T. Hovestadt, and H. J. Poethke. 2003. Movement patterns of the bush cricket Platycleis albopunctata in different types of habitat: matrix is not always matrix. Ecological Entomology 28 :432-438. Hobbs, R. L. 1992. The role of corridors in conservation: solution or bandwagon? Trends in Ecology and Evolution 7 :389-392. Hudgens, B. R., and N. M. Haddad. 2003. Pred icting which species will benefit from corridors in fragmented landscapes from population growth models. The American Naturalist 161 :808-820. Ims, R. A. 1995. Movement pa tterns related to spatial st ructures. Pages 85-109 in L. Hansson, L. Fahrig, and G. Merriam, ed itors. Mosaic landsca pes and ecological processes. Chapman and Hall, London, United Kingdom. Inglis, G., and A. J. Underwood. 1992. Comments on some designs proposed for experiments on the biological importan ce of corridors. Conservation Biology 6 :581-586. Iriarte, J. A., G. A. Lobos, and F. M. Jaks ic. 2005. Invasive vertebrate species in Chile and their control and monitoring by govern mental agencies. Revista Chilena de Historia Natural 78 :143-154. Jax, K., and R. Rozzi. 2004. Ecological theory and values in the determination of conservation goals: examples from temperat e regions of Germany, United States of America, and Chile. Revista Chilena de Historia Natural 77 :349-366. Jimenez, J. E. 2000. Effect of sample size, pl ot size, and counting time on estimates of avian diversity and abundance in a Chilean rainforest. Journal of Field Ornithology 71 :66-87.
106 Johnson, A. W. 1965, 1967. The birds of Chile and adjacent regions of Argentina, Bolivia and Peru. Platt Establecimientos Gr aficos, Buenos Aires, Argentina. Johnson, M. L., and M. S. Gaines. 1990. Evolut ion of dispersal: theoretical models and empirical tests using birds a nd mammals. Animal Behaviour 28 :1140-1162. Keitt, T. H. 1997. Stability and complexity on a lattice: coexistence of species in an individual-based food web m odel. Ecological Modelling 102 :243-258. Korzukhin, M. D., M. T. Ter-Mikaelian, and R. G. Wagner. 1996. Process versus empirical models: which approach for fo rest ecosystem management? Canadian Journal of Forest Research. 26 :879-887. Kovach Computing Services. 2004. Oriana circ ular statistics so ftware version 2.01c. Pentraeth, United Kingdom. Available from the internet. URL:http://www.kovcomp.co.uk . (accessed May 2005). Kozakiewicz, M. 1993. Habitat isolation and ecological barriers: the effect on small mammal populations and communities. Acta Theriologica 38 :1-30. Lande, R., and G. F. Barrowclough. 1987. Eff ective population size, genetic variation, and their use in population management. Pages87-123 in M. E. SoulÃ©, editor. Viable populations for conservation. Ca mbridge University Press, Cambridge, United Kingdom. Levey, D. J., G. M. Bolker, J. T. Tewk sbury, S. Sargent, and N. M. Haddad. 2005a. Effects of landscape corridors on seed dispersal by birds. Science 309 :146-148. Levey, D. J., G. M. Bolker, J. T. Tewk sbury, S. Sargent, and N. M. Haddad. 2005b. Response to Proche et al. Science 310 :779-783. Lima, S. L. 1993. Ecological and evolutio nary perspectives on escape from predatory attack: a survey of North American birds. Wilson Bulletin 105 :1-47. Lima, S. L., and L. M. Dill. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68 :619-640. Lima, S. L., and P. A. Zollner. 1996. Towa rds a behavioral ecology of ecological landscapes. Trends in Ecology and Evolution 11 :131-135. Lindenmayer, D. B., and R. C. Lacy. 1995. A simulation study of the impacts of population subdivision on th e mountain brushtail possum Trichosurus caninus ogilby (Phalangeridae: marsupialia) in sout h-eastern Australia I. Demographic stability and population persiste nce. Biological Conservation 73 :119-129.
107 Lovejoy, T. E., R. O. Bierregaard, Jr., A. B. Rylands, J. R. Malcom, C. E. Quintela, L. H. Harper, K. S. Brown, Jr., A. H. Powell, G. V. N. Powell, H. O. R. Shubart, and M. B. Hayes. 1986. Edge and other effects of isolation on Amazon forest fragments. Pages 257-285 in M. E. SoulÃ©, editor. Cons ervation biology: the science of scarcity and diversity. Sinauer Associat es, Sunderland, Massachusetts. MacArthur, R. H., and E. O. Wilson 1967. The theory of island biogeography. Princeton University Press, Princeton, New Jersey. Machtans, C. S., M. A. Villard, and S. J. Hannon. 1996. Use of riparian buffer strips as movement corridors by forest birds. Conservation Biology 10 :1366-1379. Mann, C. C., and M. L. Plummer. 1993. The high cost of biodiversity. Science 160 :18681871. Margules, C. R., and R. L. Pressey. 2000. Systematic conservation planning. Nature 405 :243-253. McGarigal, K., S. A. Cushman, M. C. N eel, and E. Ene. 2002. FRAGSTATS: spatial pattern analysis program for categorical maps. University of Massachusetts, Amherst, Massachusetts. Available from the internet URL:http://www.umass.edu/landeco /research/fragst ats/fragstats (accessed July 2004). Mittermeier, R. A., N. Myers, J. B. Thorse n, G. A. B. de Fonseca, and S. Olivieri. 1998. Biodiversity hotspots and majo r tropical wilderness areas: approaches to setting conservation priorities. Conservation Biology 12 :516-520. Myers, N., R. A. Mittermeier, C. G. Mitterme ier, G. A. B. de Fonseca, and J. Kent. 2000. Biodiversity hotspots for cons ervation priorities. Nature 403 :853-858. Nantel, P., A. Bouchard, L. Brouillet, and S. Hay. 1998. Selection of areas for protecting rare plants with integration of land use conflicts: a case study fo r the west coast of Newfoundland, Canada. Biological Conservation 84 :223-2345. Nicholls, A. O., and C. R. Margules. 1991. The design of studies to demonstrate the biological importance of corri dors. Pages 49-61 in D. A. Saunders, and R. J. Hobbs, editors. Nature conservation 2: the role of corridors. Surrey Beatty, Chippng Norton, New South Wales, Australia. NiemelÃ¤, J. 2001. The utility of movement corridors in forested landscapes. Scandinavian Journal of Forest Research 3 (Suppl):70-78. Noss., R. F., P. Beier. 2000. Arguing over l ittle things: response to Haddad et al. Conservation Biology 14: 1546-1548.
108 Noss., R. F., C. Carroll, K. Vance-Borl and, and G. Wuerthner. 2002. A multicriteria assessment of the irreplaceability and vul nerability of sites in the Greater Yellowstone Ecosystem. Conservation Biology 16 :895-908. Olson, D. M., and E. Dinerstein. 1998. The global 200: a represen tation approach to conserving the earth's most biologically valuable ecoregions. Conservation Biology 12 :502-515. Orrock, J. L. and E. I. Damschen. 2005. Corridors cause differential seed predation. Ecological Applications 15 :793-798. Ovaskainen, O. 2002. Long-term persistence of species and the SLOSS problem. Journal of Theoretical Biology 218 :419-433. Ovaskainen, O., and I. Hanski. 2004. From individual behavior to metapopulation dynamics: unifying the patchy population and classic metapopulation models. American Naturalist 164 :364-377. Pressey, R. L., and R. M. Cowling. 2001. Re serve selection algorithms and the real world. Conservation Biology 15 :275-277. Proche , S, J. R. U. Wilson, R. Veldtman, J. M. Kalwij, D. M. Richardson, and S. L. Chown. 2005. Landscape corridors: possible dangers? Science 310 :779-783. Pulliam, R. 1988. Sources, sinks, and populat ion regulation. American Naturalist 132 :652-661. Purtauf, T., J. Dauber, and V. Wolters. 2004. Carabid communities in the spatio-temporal mosaic of a rural landscape. Landscape and Urban Planning 67 :185-193. Rail, J. F., M. Darveau, A. Desrochers, and J. Huot. 1997. Territorial responses of boreal forest birds to habitat gaps. Condor 99 :976-980. Raphael, M. G., and B. G. Marcot. 1994. Species and ecosystem viability Â– key questions and issues. Journal of Forestry 92 :45-47. Recher, H. F., and D. L. Serventy. 1991. L ong term changes in relative abundances of birds in KingÂ’s Park, Perth, west ern Australia. Conservation Biology 5 :90-102. Redford, K. H., and B. H. Richter. 1999. Conser vation of biodiversity in a world of use. Conservation Biology 13 :1246-1256. Reid, S., C. Cornelius, O. Barbosa, C. Me ynard, C. Silva-GarcÃa, and P. Marquet. 2002. Conservation of temperate forest birds in Chile: implications fr om the study of an isolated forest relict. Bi odiversity and Conservation 11 :1975-1990. Reid, S., I. A. Diaz, J. J. Armesto, and M. F. Willson. 2004. Importance of native bamboo for understory birds in Chil ean temperate forests. Auk 121 :515-525.
109 Richter-Dyn, N., N. S. Goel. 1972. On the exti nction of a colonizing species. Theoretical Population Biology 3 :406-423. Ricketts, T. H. 2001. The matrix matters: e ffective isolation in fragmented landscapes. American Naturalist 158 :87-99. Ridgely, R. S., and G. Tudor. 1994. The birds of South America, vol. II. the suboscine passerines. University of Texas Press, Austin, Texas. Ries, L., and D. M. Debinski. 2001. Butterfly responses to habitat edges in the highly fragmented prairies of central Iowa. Journal of Animal Ecology 70 :840-852. RodrÃguez, A. S. L., S. J. Andelman, M. I. Bakarr, L. Biotani, T. M. Brooks, R. M. Cowling, L. D. C. Fishpool, G. A. B. da Fonseca, K. J. Gaston, M. Hoffmann, J. S. Long, P. A. Marquet, J. D. Pilgrim, R. L. Pressey, J. Schipper, W. Sechrest, S. N. Stewart, L. G. Underhill, R. W. Walle r, M. E. J. Watts, and X. I. Yan. 2004. Effectiveness of the global protected area network in representing species. Nature 428 :640-643. RodrÃguez, A., J. AndrÃ©n, and G. Janss on. 2001. Habitat-mediated predation risk and decision making of small birds at forest edges. Oikos 95 :383-396. Roland, J., N. Keyghobadi, and S. Fownes. 2000. Alpine Parnassius butterfly dispersal: effects of landscape and population size. Ecology 81 :1642-1653. RSI (Research Systems Inc.). 1999. ENVI 3.2 Remote Sensing Software. RSI, Boulder, Colorado. Rosenberg, D. K., B. R. Noon, and E. C. Meslow. 1997. Biological corridors: form, function, and efficacy. Bioscience 47 :677-687. Rosenzweig, M. L. 2003. Win-win ecology: how the EarthÂ’s species can survive in the midst of human enterprise. Oxford University Press, New York, New York. Rykiel, E. J., Jr. 1996. Testing ecological m odels: the meaning of validation. Ecological Modelling 90 :229-244. St. Clair, C. C., M. BÃ©lisle, A. Desroche rs, and S. J. Hannon. 1998. Winter responses of forest birds to habitat corridor s and gaps. Conservation Ecology 2 :13. Available from the internet. URL:http//www.consecol.org/vol12/iss12/art13 . (accessed June 2001). Schlaepfer, M. A., M. C. Runge, and P. W. Sherman. 2002. Ecological and evolutionary traps. Trends in Ecology and Evolution 17 :474-480. Sherrod, P. H. 2003. Decision tree regression analysis softwa re (DTREG). Available at URL:http//www.dtreg.com. (accessed July 2005).
110 Sibley, C. G., and B. L. Monroe, Jr. 1990. Distribution and taxonomy of birds of the world. Yale University Pre ss, New Haven, Connecticut. Sieving, K. E., T. A. Contreras, and K. L. Maute. 2004. Heterospecific facilitation of forest-boundary crossing by m obbing understory birds in North-Central Florida. Auk 121 :738-751. Sieving, K. E., and J. R. Karr. 1997. Avian extinction and persistence mechanisms in lowland Panama. Pages 156-170 in W. F. Laurance, and R. O. Bierregaard, Jr., editors. Tropical forest remnants: eco logy, management, and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois. Sieving, K. E., M. F. Willson, and T. L. De Santo. 1996. Habitat barriers to movement of understory birds in fragmented south-temperate rainforest. Auk 113 :944-949. Sieving, K. E., M. F. Willson, and T. L. De Santo. 2000. Defining corridor functions for endemic birds in fragmented south-te mperate rainforest. Conservation Biology 14 :1120-1132. Simberloff, D. S., J. A. Farr, J. Cox, and D. W. Mehlman. 1992. Movement corridors: Conservation bargains or poor investments? Conservation Biology 6 :493-504. SPSS, 2001. SPSS advanced statistics softwa re, version 11.0.1. Chicago, Illinois. Stacey, P. B., and M. L. Taper. 1992. Envir onmental variation and the persistence of small populations. Ecological Applications 2 :18-29. Stamps, J. A., M. Buechner, and V. V. Krishnan. 1987. The effect of habitat geometry on territorial defense costs: intruder pressure in bounded populations. American Zoologist 27 :307-325. Stattersfield, A. J. 1998. Endemic bird areas of the world: priorities for biodiversity conservation. Birdlife conservation series ; no. 7. Birdlife Intern ational, Cambridge, United Kingdom. Suhonen, J. 1993. Predation risk influences the use of foraging sites by tits. Ecology 74 :1197-1203. Sutcliffe, O. L., and C. D. Thomas. 1996. Open corridors appear to fa cilitate dispersal by ringlet butterflies ( Aphantopus hyperantus ) between woodland clearings. Conservation Biology 10 :1359-1365. Sutherland, G. D., A. S. Harestad, K. Price, and K. P. Lertzman. 2000. Scaling of natal dispersal in terrestrial birds and mammals. Conservation Ecology 4 :16. Available from the internet. URL:http//www.consecol.org/vol4/iss1/art16. (accessed June 2001).
111 Taylor, B. 1991. Investigating species incidenc e over habitat fragments of different areas a look at error estimation. Biologica l Journal of the Linnean Society 42 :177-191. Tear, T. H., P. Kareiva, P. L. Angermeier, P. Comer, B. Czech, R. Kautz, L. Landon, D. Mehlman, K. Murphy, M. Ruckelshaus, J. M. Scott, and G. Wilhere. 2005. How much is enough? The recurrent proble m of setting measurable objectives in conservation. Bioscience 55 :835-849. ter Braak, C. J. F., I. Hanski, and J. Ver boom. 1998. The incidence function approach to modeling of metapopulation dynamics. Pages 167-188 in J. Bascompte, and R. V. SoulÃ©, editors. Modeling spatiotemporal dynamics in ecology. Springer-Verlag, New York, New York. Tewksbury, J. J., D. J. Levey, N. M. Haddad, S. Sargent, J. L. Orrock, A. Weldon, B. J. Danielson, J. Brinkerhoff, E. I. Da mschen, and P. Townsend. 2002. Corridors affect plants, animals, and their interacti ons in fragmented landscapes. Proceedings of the National Academy of Science 99 :12923-12926. Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371 :65-66. Tischendorf, L., D. J. Bender, and L. Fahrig. 2003. Evaluation of patch isolation metrics in mosaic landscapes for specialist vs. generalist dispersers. Landscape Ecology 18:41-50. Tischendorf, L., and L. Fahrig. 2000. On the usage and measurement of landscape connectivity. Oikos 90 :7-19. Todd, I. A., and R. J. Cowie. 1990. Measuring the risk of predation in an energy currency: field experiments with foraging blue tits, Parus caeruleus . Animal Behaviour. 40 :112-117. Turchin, P. B. 1998. Quantitative analysis of movement. Sinauer Associates, Sunderland, Massachusetts. Urban, D., and T. Keitt. 2001. Landscape connec tivity: a graph-theoretic perspective. Ecology 82 :1205-1218. Van Breemen, N. 1995. How Sphagnum bogs down other plants. Trends in Ecology and Evolution 10 :270-275. Vandermeer, J., and R. Carvajal. 2001. Metapop ulation dynamics and the quality of the matrix. American Naturalist 158 :211-220. Vayssieres, M. P., R. E. Plant, and B. H. Allen-Diaz. 2000. Classification trees, an alternative non-parametric a pproach for predicting species distributions. Journal of Vegetation Science 11 :679-694.
112 Veblen, T. T., F. M. Schlegel, and J. V. Oltremari. 1983. Temperate broad-leaved evergreen forests of South America. Pages 5-31 in J. D. Ovington, editor. Temperate broad-leaved evergreen forest s. Elsevier, Amsterdam, Netherlands. Verboom, J., R. Foppen, P. Chardon, P. Opda m, and P. Luttikhuizen. 2001. Standards for persistent habitat networks for vertebrate populations: the key patch approach. An example for marshland bird popul ations. Biological Conservation 100 :89-101. Warton, D. I., and G. M. Wardle. 2003. Site-to-si te variation in the demography of a fire affected perennial, Acacia suaveolens , at Ku-ring-gai Chase National Park, New South Wales, Australia. Austral Ecology 28 :38-47. Wiens, J. A. 1994. Habitat fragmentation: island v landscape pe rspectives on bird conservation. Ibis 137 :s97-s104. Wilcove, D. S., and J. Lee. 2004. Using econom ic and regulatory in centives to restore endangered species: lessons learned fr om three new programs. Conservation Biology 18 :639-645. Willson, M. F. 2004. Loss of habitat connectivity hinders pair formation and juvenile dispersal of Chucao Tapaculos in Chilean rainforest. Condor 106 :166-171. Willson, M. F., and J. J. Armesto. 1996. The natural history of ChiloÃ©: on Darwin's trail. Revista Chilena de Historia Natural 69 :149-161. Willson, M. F., T. L. De Santo, C. Sabag, and J .J. Armesto. 1994. Avian communities in fragmented south-temperate rainfore st in Chile. Conservation Biology 8 :508-520. Willson, M. F., J. L. Morrison, K. E. Sieving, T. L. De Santo, L. Santisteban, and I. Diaz. 2001. Patterns of predation risk and survival of bird nests in a Chilean agricultural landscape. Conservation Biology 15 :447-456. Willson, M. F., K. E. Sieving, T. L. De Santo. 2004. Aves de ChiloÃ©: diversidad, amenazas y estrategias de conservaciÃ³n. Pages 442-450 in C. Smith-RamÃrez, J. J. Armesto, and C. Valdovinos, editors. Hi story, biodiversity and ecology of the Coastal Rainforest, Chile. Editori al Universitaria, Santiago, Chile. Witt, W. C., and N. Huntly. 2001. Effect s of isolation on red-backed voles ( Clethrionomys gapperi ) and deer mice ( Peromyscus maniculatus ) in a sage-steppe matrix. Canadian Journal of Zoology 79 :1597:1603. WWF (World Wildlife Fund). 2000. Preliminary results from the biodiversity vision for Valdivian Temperate Rainforest of Chile and Argentina. Internal report. (see http://www.worldwildlilfe.org/global200/spaces.cfm ; accessed October 2005). Yoder, J. M., E. A. Marschall, and D. A. Swanson. 2004. The cost of dispersal: predation as a function of movement and site familia rity in ruffed grouse. Behavioral Ecology 15 :469-476.
113 Zollner, P. A., and S. L. Lima. 1997. Landscapelevel perceptual abil ities in white-footed mice: perceptual range and the dete ction of forested habitat. Oikos 80 :51-60.
114 BIOGRAPHICAL SKETCH Traci Darnell CastellÃ³n was born July 10, 1967, in Acworth, Georgia. She graduated from Pebblebrook High School, Mableton, Georgia, in 1985. She then attended West Georgia College in Carrollton, Georgia, receiving a Bachelor of Science degree in biology in 1990. After several y ears working on a variety of conservation projects, Traci returned to academics to pursue a graduate degree at Texas A & M University Â– Corpus Christy. She received her Master of Science in environmental science in 1998. She then continued her gr aduate education in the Department of Wildlife Ecology and Conservation at the Univer sity of Florida, pursuing a Doctor of Philosophy degree, which was awarded in May 2006.