Differential effects of habitat fragmentation on birds and mammals in Valdivian temperate rainforests

Four recent studies on Chilean vertebrates underscore the very different effects that habitat fragmentation may have under different conditions. In southern Chile, birds exhibited significant species-area relationships, whereas the more depauperate small mammal community did not. The avifauna of highly isolated rainforest remnants in Fray Jorge National Park (IV Region) presented steeper species-area relations than those in southern Chile, reminiscent of landbridge islands after faunal relaxation. The small mammal results were unexpected but may reflect the reduced vagility and immigration potential of this group. Additionally, the inter-remnant matrix habitat may support large populations of only some species, allowing them to capitalize on resources that are available in forest remnants. Such an effect is less likely for birds due to their greater vagility. Although the influence of remnant area is more clearly demonstrated for birds in Chile, it appears that small mammal assemblages may be more immediately impacted by habitat fragmentation than are birds, and that further fragmentation of temperate rainforests will seriously impoverish the small mammal assemblage, many of whose members are endemic to this region. Further research should emphasize the role of the inter-remnant matrix, of habitat corridors, and of the relative sensitivity of different species to matrix/corridor quality and remnant isolation.


INTRODUCTION
The biota of the temperate rainforests of southern South America is diverse and of great biogeographic interest.The herpetofauna alone includes 20 species of anurans, six lizards, and two snakes (Formas 1979, these numbers are dated and likely underestimate the true diversity).The avifauna is moderately diverse (Vuilleumier 1985 reports 46 species in 40 genera) but highly endemic (66 % of species [Rozzi et a!. 1995], 10 % of genera [Vuilleumier 1985]).And while the mammal fauna is not highly diverse (Meserve & Jaksic 1991), it is similarly unique (six of nine genera are endemic, Patterson 1992).Overall, this area is notable for levels of endemicity that rival some of the world's hotspots such as Madagascar.
This region, the Nothofagus center of endemicity (Cracraft 1985), has been given high priority for conservation by the International Council on Bird Preservation (International Council for Bird Conservation 1992), and overall is considered "endangered" and of outstanding regional value (Biodiversity Support Program 1995).The World Wildlife Fund has included the Valdivian temperate rainforest in its "Focal 25" areas for immediate conservation action and impact (see http: lwww.world wildlife.org/ glo bal200/ spaces.cfm).Unfortunately, protected areas in temperate Chile generally are at high elevation and higher latitudes, leaving most species in this region susceptible to ecologically unfriendly land management (Armesto et al. 1998, Defensores del Bosque Chileno 1998).The future of this region, therefore, hangs on the balance of political decisions that are being made often without sufficient biological background (e.g., Fuentes 1994, Lara et al. 1995).The future of the animals of these forests is directly dependent on the future of the forests themselves, and enlightened management decisions to guide the future of these global treasures rely on an understanding of the consequences of pending management decisions.
Four recent research programs have provided steps towards understanding the consequences of forest fragmentation in this system, but all suffer from taxonomically restricted focus.Willson et al. ( 1994), Gantz & Rau ( 1999), and Cornelius et al. (2000) studied avian communities, and Kelt (2000) studied small mammal communities, in fragmented forest habitats in Chile.Three of these studies (Willson et al. 1994,Gantz & Rau 1999and Kelt 2000) were conducted in anthropogenic forest fragments within areas recently covered by Valdivian rainforest, whereas the fourth study (Cornelius et al. 2000) was conducted in forest remnants that have been isolated from other temperate rainforests since the Quaternary.Perhaps surprisingly, the relationship between area and local species richness differed greatly between these three studies.The purpose of this commentary is to compare these four studies and to suggest avenues of future research.et al. ( 1994) studied the avifauna of 11 forest remnants on Isla Chiloe during the austral spring of 1992.Sites ranged in area from less than 0.5 ha to over 100 ha, but were not evenly distributed by size class.Unable to find any intermediate size fragments (e.g., 10-100 ha in area), their sample included three large patches (> 100 ha), five small remnants (0.1-10 ha), and three tiny remnants ( < 0.5 ha).They also studied one secondary forest greater than 100 ha in area.They reported 25 regularly occurring bird species, most of which (60 %) were endemic to southern temperate forests.Species diversity declined markedly with remnant size (Table 1, Fig. 1), and with the exception of three edge species, virtually all species had lower abundances in smaller remnants.

Willson
Gantz & Rau (1999) evaluated avian diversity in 18 forest remnants in the province of Osorno in 1996 and 1997.Remnants here ranged from 3 to 344 ha with a fairly even distribution of sizes.Species richness ranged from eight to 19 species, and a total of 32 species were recorded.These authors also demonstrated a clear relationship Estadfsticas de regresi6n para aves en fragmentos de bosque remanentes del sur y centro-norte, y para mamfferos pequeiios en fragmentos de bosque remanentes del sur de Chile.Fuentes: (I)   (2000) analizaron la avifauna en parches de bosque de Chile centro-norte.Se incluye un resumen de los datos y de las relaciones especie-area.La leyenda incluye las pendientes (valores de z) de estas relaciones.
between remnant area and species richness (Table 1, Fig. 1).Cornelius et al. (2000) censused the avifauna of six forest fragments located in north-central Chile.Over the course of a year these patches, ranging from 0.5 to 22.5 ha in area, were sampled with methods similar to those employed by Willson et al. (1994).Unlike the forest remnants studied in southern Chile, the forests studied by Cornelius et al. (2000) were protected from human intervention by their presence within Parque N acional Fray Jorge.Also unlike other studies summarized here, the forests at Fray Jorge have been isolated from continuous temperate rainforest since the Quaternary, but have persisted at this locality because the coastal hills trap extensive fog (the "camanchaca") and provide a substantial supplement to the mean annual rainfall of 85 mm.Because of this extensive period of isolation the forests at Fray Jorge provide insights into the long-term impacts of forest fragmentation, and may bear important insights for managers of the forests found in southern Chile and Argentina.Cornelius et al. (2000) encountered 21 species of birds in their forest samples (another 16 species were recorded in adjacent matorral habitats but did not venture into the forests), and they found species richness to be strongly and positively correlated with remnant area in all seasons studied (Table 1 ).Moreover, the species-area relationship was similar in all four seasons studied (ANCOV A, Table 2A).
Only one study has been conducted on the distribution of mammals in temperate rainforest TABLE2 Analysis of covariance results comparing the regression slope (species-area relation, z-values) across seasons for birds at (A) Fray Jorge and (B) across all three studies with all data pooled within studies Resultados del analisis de covarianza utilizado para comparar las pendientes (de relaciones especie-area, valores de z) entre estaciones en (A) Fray Jorge y (B) entre los tres estudios con los datos agrupados dentro de cada estudio remnants.Kelt (2000) sampled the small mammal assemblages in 14 remnants in the same area as Gantz & Rau (1999).Located in a similar system to that studied by Willson et al. (1994), both Kelt (2000) and Gantz & Rau (1999) studied forest remnants within a largely agricultural mosaic.Kelt (2000) reported on remnants ranging from ca. 0.5 to 125 ha in size, all of which were sampled with extensive trapping during two winter seasons ( 1985 and 1986), when populations generally are greatest in this system (Meserve et al. 1991 ).Additionally, four remnants were sampled in both years of study (for a total of 18 samples).In contrast to the relatively speciose avifaunas of these habitats, the total small mammal fauna included only six native rodents and two marsupials.In marked contrast to both avian studies, and in spite of a larger number of remnants sampled than in two of the avian studies, Kelt found no indication of any area effects, either on species richness (Table 1 ), estimated population  2000).
Although six of the bird species recorded at Fray Jorge were not recorded by Willson et al. (1994 ), and the latter recorded nine species not reported by Cornelius et al. (2000), the Fray Jorge avifauna was a subset of the Valdivian avifauna, both in terms of species composition and relative abundances (Cornelius et al. 2000).Gantz & Rau ( 1999) did not provide a list of the species in their study, but it was undoubtedly typical of the Valdivian avifauna as well.Consequently it seems reasonable to compare these studies in terms of general responses to habitat fragmentation.The broad overlap in the study areas of Kelt (2000) and Gantz & Rau (1999) also support cross-taxon comparisons between these studies.

Comparison of responses to fragmentation
The species-area relations across these studies differ significantly (ANCOV A), and a priori contrasts demonstrated that most studies differ significantly from each other (Table 2B).The zvalue calculated for the avifauna studied by Gantz & Rau ( 1999) does not differ from that observed at Fray Jorge (Table 2B), although it is significantly steeper than that observed by Willson et al. (1994) on Isla de Chiloe.Although these contrasts should be evaluated cautiously, as it is not technically valid to perform all of these contrasts, it is apparent that dramatic differences exist among these studies.
In comparing their results with those from Chiloe, Cornelius et al. (2000) noted that the forest habitat at Fray Jorge has been restricted to its present distribution since the Quaternary.The nearest extensive Valdivian rainforests presently are > 1,200 km south of Fray Jorge, leading Cornelius et al. (2000) to predict that the Fray Jorge fauna would be depauperate, that it would yield a species-area relationship more similar to true (oceanic) islands than to land-bridge islands, and that this system had "already paid its extinction debt" (pp.535).Implicit in the prediction of an oceanic z-value is that most bird species present in these remnants have arrived since isolation, rather than representing a relictual post-isolation fauna (Lawlor 1986).In fact, three of the four seasonal z-values obtained by these authors were> 0.20, more similar to the mammal faunas of land-bridge islands than to the oceanic islands summarized by Lawlor (1986), but intermediate between the "island curves" (z 0.25-0.33)and the "mainland curves" (z 0.13-0.18)outlined by Rosenzweig (1995), and substantially lower than the z = 0.34 exhibited by birds in 23 Paramo "islands" (Viulleumier & Simberloff 1980).And while the assemblages at Fray Jorge (S = 21 species) had fewer species than those in southern Chile (S = 25 and 32), it is not clear if this is "depauperate" relative to the presumed source fauna (S = 46, Vuilleumier 1985).Certainly the remnants at Fray Jorge include some species that are not members of the Valdivian rainforests avifauna (e.g., Patagona gigas (Vieillot), Asthenes humicola (Kittlitz), but most species encountered are represented in the southern avifauna (Cornelius et al. 2000).At any rate, the remnants at Fray Jorge do not seem inordinately depauperate relative to remnant forests near Osorno (Gantz & Rau 1999) or on Isla de Chiloe (Willson et al. 1994 ).
Because true oceanic islands experience low rates of immigration, regardless of size, Lawlor (1986) noted that they should have low species richness as well as low species-area relations.Land-bridge islands, in contrast, should be driven by differential extinction on large vs. small islands, with smaller islands losing more species.Consequently, these should have steeper z-values and range from low to high richness on small and large remnants, respectively.Thus, higher zvalues are to be expected either when postisolation extinction has been greater in small than large remnants, or when immigration is sufficiently high that large remnants can accumulate species that cannot survive on smaller remnants (a third possibility, operating over evolutionary time scales, invokes greater in situ speciation on larger islands, but this is not relevant to the habitat remnants under consideration here).Of particular interest is that the former explanation invokes differential extinction on large vs. small remnants, whereas the latter invokes differential immigration.Data from Fray Jorge are somewhat inconclusive, as z-values fall in a rather vague area between that expected for oceanic islands and that for land-bridge islands (see above), but it seems likely that the extant avifauna has indeed relaxed from an initially higher species richness.Further work aimed at teasing apart these two explanations, perhaps involving genetic comparisons of Fray Jorge populations with those in contiguous Valdivian forests of southern Chile, would be particularly interesting.In contrast to the avifauna of Fray Jorge, bird faunas in fragmented habitats of southern Chile (Willson et al. 1996, Gantz & Rau 1999) yielded lower slopes but high species richness that generally are more typical of continental habitat fragments (Lawlor 1986).
The lack of a significant species-area relationship for small mammals was initially surprising, as such relationships have been reported elsewhere (e.g., Dunstan & Fox 1996, Rosenblatt et al. 1999, Nupp & Swihart 2000).Mammal faunas typically are depauperate relative to avifaunas, however, and it is possible that the limited mammal fauna of the Valdivian forests precluded a significant relationship.It seems more likely, however, that mixed management of "matrix" habitat allowed for extensive dispersal by common species (especially Abrothrix olivaceus (Waterhouse, 1837), A. longipilis (Waterhouse, 1837), and Oligoryzomys longicaudatus (Bennett, 1832), which occurred in all remnants and comprised over 98 % of captures) but only moderate dispersal by most other species (see Malcolm 1997, Kelt 2000).In the structurally similar rainforests of the Olympic Peninsula of Washington, Lomolino & Perault (200 1) studied a mammal fauna with 18 species, also reporting a non-significant sp-ecies-area relationship.Although the largest remnant reported by Lomolino & Perault (200 1) was less than 60 ha in area, Kelt (2000) surveyed one remnant in excess of 120 ha.Both papers suggested that the lack of a significant relationship exemplified the "small island effect," which occurs when no sites are sufficiently large to generate a positive slope (Brown & Lomolino 1998).Both studies also echoed earlier suggestions (Malcolm 1997) that matrix habitat may support habitat generalists, thereby increasing the representation of these species in isolated habitats, as well as the impact of these species on any individuals of more specialist species.
Superficially these results might suggest that the avifauna of southern temperate rainforests are more likely to be adversely impacted by continued habitat fragmentation than is the mammal fauna.This is not necessarily true, however, as the distribution of small mammals reported by Kelt (2000) was really quite heterogeneous.Reliance upon the species-area relationship may obfuscate a number of details that may be important, and it also ignores the broader ecological context in which habitat remnants exist.Although most bird species declined in habitat remnants on Chiloe, three edge species (Turdus falcklandii Quoy and Gaimard, Troglodytes aedon Veillot, and Zonotrichia capensis (Miiller)) increased significantly.Similarly, three of the seven native species encountered by Kelt (2000) (excluding Loxodontomys micropus (Waterhouse, 183 7) which was only captured in intervening habitats) were present in all remnants, and constituted the vast majority of captures.The remaining species included the endemic Irenomys tarsal is (Philippi, 1900) (12 captures in seven remnants), Geoxus valdivianus (Philippi, 1858) (nine captures in seven remnants), Dromiciops gliroides Thomas, 1894 (three captures in two remnants), and Rhyncholestes raphanurus Osgood, 1924 (one capture).Moreover, four mammal species found in this region had lower abundances in remnants than in nearby contiguous habitat (Kelt 2000).Clearly, reliance only upon the simplistic speciesarea relationship fails to account for factors that influence the abundance and distribution of these taxa (e.g., Harrison & Bruna 1999).Importantly, these observations suggest that ongoing fragmentation is highly likely to have a negative impact on the mammalian assemblage of this region, and possibly more so than for birds, and the ecology of these species should be considered in ongoing management decisions.

The future
In their assessment of the conservation status of terrestrial mammals in Chile, Cofr & Marquet (1999) noted that the Matorral region contains a larger number of endemic species than any major biotype in Chile, is particularly threatened by habitat loss and degradation, and is poorly protected by the national system of parks and reserves.Although the temperate rainforests of southern Chile did not emerge as highly threatened in their analysis, this likely was a consequence of the fact that the southern region has fewer species than the matorral region of Chile, and therefore did not yield large summary statistics.Mediterranean regions are threatened globally (e.g., Runde! eta!. 1988, Mittermeier eta!. 1998, Olsen & Dinerstein 1998), and it might be argued that future efforts to protect additional parts of these biomes from human encroachment are tightly constrained.Fortunately, the case for temperate rainforests is not quite as severe; although these ecosystems also are poorly represented in Chile's system of protected areas (Armesto et a!. 1998) and they are heavily impacted by extractive industries (Ecotrust/Conservation International 1992, Dono so & Lara 1995), it may yet be possible to plan for their biotic maintenance.With respect to the Valdivian forests of southern Chile and Argentina, it is clear that further work is needed in many facets of the ecology of various species.We presently have no information on the responses of most temperate groups to habitat fragmentation or degradation.Several points and recommendations emerge from this commentary.
First, the importance of the habitat matrix in which forest remnants are isolated appears to be critical to the post-isolation dynamics of the remnants.The importance of matrix habitat has been emphasized both in tropical (e.g., Lynam 1997, Malcolm 1997) and temperate (e.g., Kelt 2000, Lomolino & Perault 2001) systems, and may greatly influence the character of remnant faunas, both in terms of the species present and their relative numbers.Further research on how matrix characteristics influence both the composition and the demography of remnant faunas should be given high priority.
Second, for most taxa we remain woefully ignorant of the role of habitat edge on the activity patterns of various species.While many predators are most active along habitat edges and may greatly influence the composition of the remnant fauna (e.g., Reske 1995, Robinson eta!. 1995), some non-predatory taxa also prefer edge habitats (e.g., Willson et a!. 1996, Lomolino & Perault 2001 ), which may provide an agreeable combination of cover and food.Of course, edge is perceived very differently by different taxa (e.g., butterflies vs. birds vs. small mammals in Amazonia; see papers in Laurance & Bierregaard 1997), so it may be difficult, if not impossible, to generalize any specific effects across taxa.
Third, synergistic effects of remnant area, isolation, and habitat on the faunal composition of habitat remnants have been documented in Australia (e.g., Dunstan & Fox 1996) and North America (Lomolino & Perault 2001).Further work in Chilean and Argentine forests should employ a broader geographic scale than the studies summarized here, employing GIS and remote imaging to better quantify remnant features as well as details of matrix habitat.The basic island biogeography paradigm is no longer suitable for understanding the complexities of fragmented, often multiply interconnected habitats (Harrison & Bruna 1999).
Fourth, patterns of individual movement need to be investigated to understand how species use fragments of differing sizes and shapes.Different species may exhibit unexpected patterns of movement among and within habitat remnants (e.g., Diffendorfer et al. 1995), rendering predictions based on research in primary forest unreliable at best.
Fifth, differential sensitivity to secondary habitat degradation is likely to influence some species more than others.Habitat changes often associated with fragmentation in South America include invasion by cattle and removal of wood for fuel.Arboreal species (e.g., many birds, the Chilean tree mouse Irenomys, the monito del monte Dromiciops) and species that forage in the litter (e.g., Geoxus, Rhyncholestes, birds in the family Rhynocryptidae) may be more heavily impacted by such activities than by the size or isolation of the remnant or by characteristics of the habitat matrix.It is likely that some form of management will be needed to allow partial exploitation of forest remnants while maintaining critical structural and biotic features that allow the survival of these species.
Finally, and perhaps most important in the face of certain continued forest exploitation, is the question of how habitat remnants and corridors should be designed to maximize the probability of inter-remnant movement.While research has begun to quantify habitat barriers (Sieving et al. 1996) and corridor needs by Valdivian forest birds (Sieving et al. 2000), no information is available on the requisite structure or size of corridors for mammalian habitat specialists such as Irenomys tarsalis, Dromiciops gliroides, Geoxus valdivianus, and Rhyncholestes raphanurus, nor for larger mammals such as the the guifia (Oncifelis guina) or the endemic pudu (Pudu puda).These species are likely to be threatened by extensive habitat alteration, and should be the focus of research while time remains to develop sound management plans.
The temperate forests of Chile and Argentina constitute an ideal system to compare against the extensive results from fragmented environments in North America, Europe, and Australia (e.g., Burgess & Sharpe 1981, Harris 1984, Wilcove et al. 1986, Robinson et al. 1995, and papers in Lidicker 1995, Dunston & Fox 1996, McCullough 1996, Barrett & Peles 1999, Lomolino & Perau1t 2000, 2001, Perault & Lomolino 2000), and to validate or modify conclusions drawn there.Substantial progress has been made towards such objectives (e.g., Donoso 1993, Armesto et al. 1995), but much remains to be done.Further efforts towards the questions outlined here will greatly increase the ability of landscape and economic managers alike to plan for rational, and hopefully sustainable, utilization of the Valdivian (and Fuegian) temperate rainforests.None of these are logistically difficult problems, and all may be addressed relatively easily.Without a concerted effort, however, it is virtually certain that the biodiversity of temperate South America will become impoverished in the face of the global economy.

TABLE I
Regression statistics for birds in temperate rainforest remnants in southern and north-central Chile, and for mammals in rainforest remnants in southern Chile.Sources: (1) Willson et al.