Printer Friendly




Abstract. We capitalized on the anthropogenic, large-scale "experiment" of deforestation in the temperate rain forests of the Olympic Peninsula, Washington State, USA, to test whether mammalian-community structure is significantly influenced by anthropogenic transformation of this landscape (principal macrohabitats include continuous, old-growth forest; old-growth corridors; old-growth fragments; second-growth forest; and clearcuts). Species richness of all non-volant mammals was lowest in second-growth forests, while richness of the eight forest-dependent species was lowest in clearcuts.

Species composition of mammalian communities differed significantly among the five principal macrohabitats, as did environmental characteristics recorded at each survey site. These differences indicate that both natural and anthropogenic processes have resulted in the nonrandom assembly of mammalian communities in this transformed, temperate rain forest landscape. The effects of deforestation are significant, interpretable (based to a large degree on changes in local habitat), and persistent given that the structure of mammalian communities from mature second-growth forests (up to 80 yr post-harvest) have not converged on those from native, old-growth forests.

Key words: assembly of ecological communities; community assembly; community ecology; corridors, role in conserving native species; fragmentation; landscape manipulation effects on mammal community; mammalian community structure; old-growth forest; Olympic Peninsula, Washington State, (USA); species composition patterns; temperate rain forests.


Community assembly remains one of ecology's most persistent and central themes (Diamond 1975, Connor and Simberloff 1979, 1983, Diamond and Gilpin 1982, Gilpin and Diamond 1982, 1984, Weiher and Keddy 1998). Studies on community assembly ask two fundamental questions: (1) do the assemblages of species in a particular community represent biased subsets of the regional pool of species? and (2) what processes are responsible for these patterns in species composition? Such studies all address the more fundamental subject addressed by MacArthur (1972) in his monograph on geographical ecology and explored by Diamond (1975) in his seminal work on assembly of ecological communities: the existence of alternative, stable communities.

Alternative, nonrandom (or disharmonic) assemblages of species can result from a diverse suite of forces, including selective extinctions, differential immigrations, or species-specific responses to disturbance, which can operate either independently or in some complex, interactive fashion (Lomolino 1986, 1996, Patterson and Atmar 1986, Patterson and Brown 1991, Taylor 1997, Wright et al. 1998). Manipulative experiments have provided some key insights into the abiotic and biotic forces influencing community structure. Many of these experiments, however, have involved very small spatial scales-i.e., on the order of just a few square meters-and short temporal scales, typically weeks to a few years (Weatherhead 1986, Karieva and Anderson 1988, Tilman 1989, Elliot 1994). Yet many of the patterns we study, including latitudinal and altitudinal gradients in species richness, or nestedness and other forms of nonrandom assembly of isolated communities, result from forces operating over a much broader range of temporal and spatial scales (Edwards et al. 1994, Brown and Lomolino 1998).

Conducting manipulative experiments at landscape and biogeographic scales is, in almost all cases, logistically infeasible and/or ethically indefensible (Brown 1995, see also Harvey and Pagel 1991). Yet biogeographers and paleontologists have known for some time that they can capitalize on another type of experiment, i.e., natural experiments, where "treatments," although not created by the researcher, provide opportunities to study a great variety of biotic responses over more appropriate temporal and spatial scales. The Great American Interchange of the South and North American biotas is perhaps the prototypic example of this approach (Simpson 1980, Marshall et al. 1982, Webb 1991).

Over the past two or three decades, ecologists have utilized another type of large-scale "experiment," but in this case the treatments are not natural, but anthropogenic. Principal among these "unplanned" experiments are those focusing on anthropogenic habitat alteration, which has modified many of the landscape features (e.g., area, isolation, patch shape and relative edge, and local-to-regional-scale productivity) thought to influence large-scale processes and related patterns in community structure. Aside from the obvious logistical difficulties, the strategies of designing large-scale studies are similar to those of any experimental study: using or, in this case, selecting sample units so that they maximize variation in the factor(s) of interest while minimizing variation of factors extrinsic to the hypothesis.

While human activities have transformed nearly all types of native landscapes and created an overwhelming, albeit depressing, diversity of unplanned, anthropogenic experiments to study, fragmentation of once-massive tropical rain forests has received the lion's share of attention (Gentry 1990, Saunders and Hobbs 1991, Adam 1992, Laurance and Bierregaard 1997). Such studies primarily focus on responses to habitat reduction, increased isolation, and increased edge effects. A wide range of taxonomic groups have been looked at, including invertebrates (Klein 1989, Brown 1997), reptiles and amphibians (Tocher et al. 1997), birds (Lovejoy et al. 1986, Warburton 1997), and mammals (Malcolm 1988, Laurance 1989, Dunstan and Fox 1996). Each group often responds differently to fragmentation, with non-flying mammals possibly demonstrating greater susceptibility than species with better dispersal abilities such as birds (Laurance 1989). For example, in the tropical forests of Queensland, Australia, diversity of mammals w as shown to decrease with decreases in forest patch size, increases in isolation, and increases in shape irregularity (i.e., relative edge; Laurance 1989, 1990, and 1991).

While these studies on fragmented tropical rain forests have provided numerous insights for ecologists and conservation biologists, there remains a great, untapped opportunity to study the effects of landscape manipulations in a variety of other biomes. Are the biogeographic and ecological responses of tropical-rain forest species representative of those of other fragmented biomes? Do fragments of other types of biomes tend to exhibit reduced diversity of native species and be dominated by just a few invasive species? Are the biogeographic and ecological responses to fragmentation rapid enough to be detectable within ecological time scales (say just a few decades), and to what degree are such responses associated with environmental and biogeographic conditions of the fragments or the surrounding matrix? Our ability to answer these questions and, in turn, gain a more comprehensive understanding of assembly of local and regional biotas may well rest upon our ability to conduct comparable studies across a varie ty of fragmented landscapes (e.g., see Martell 1983, Raphael 1988, Kirkland 1990, Craig and Dobkin 1993, Carey 1995, Carey and Johnson 1995, Songer et al. 1997).

Here we study the effects of one of these anthropogenic, large-scale experiments--deforestation and associated alterations of temperate rain forests of the Pacific Northwest. Like their tropical counterparts, temperate rain forests have some of the largest trees and greatest accumulation of biomass of any biome. Yet, due primarily to the relatively cool climates and short growing season, primary productivity and decomposition rates in temperate rain forests are relatively low (Norse 1990). In addition, many temperate rain forests, such as those of the Olympic Peninsula in Washington State, USA (Fig. 1), occur in mountainous regions, where the rains and steep slopes combine to severely inhibit soil development. Therefore, in contrast to their tropical counterparts, temperate rain forests are much less resilient. It took centuries of relatively benign conditions to develop old-growth, temperate rain forests (see Norse 1990, Peterson et al. 1997). Once the forests are cleared and subjected to severe soil erosio n, regrowth of forests and reassembly of old-growth communities may again take many centuries--providing, of course, that climatic conditions remain favorable and anthropogenic disturbances are not recurrent. On the other hand, the response of animals, both old-growth natives and invasive species, to deforestation and associated landscape transformations may be quite rapid--perhaps on the scale of years to decades.

The purpose of this study is to investigate the assembly of old-growth forest communities of the Olympic Peninsula and to assess the ecological effects of anthropogenic transformation of landscapes once dominated by temperate rain forests. Specifically, we study the differences in richness and species composition of mammal communities across five principal components (macrohabitats) of the current landscape: the mainland of relatively continuous, old-growth forest; old-growth corridors; old-growth fragments; mono-specific stands of second-growth forest; and clearcuts. We compare the mammal communities across these five macrohabitats to test the null hypothesis that mammalian community structure does not differ significantly among the five macrohabitat treatments. In addition, we test whether any observed differences in community structure among macrohabitats could be attributed to changes in local environmental conditions (e.g., canopy closure, tree height, coverage of moss, litter and herbaceous plants).


Site description

We conducted analyses across the Hood Canal District ([sim]60 000 ha) of the Olympic National Forest (ONF) in northwest Washington State, USA. Old-growth forest in this region consists of stands having the following characteristics: (1) 3.24 trees/ha (eight trees per acre) older than 160 yr or [greater than]81.3 cm (32 inches) diameter-at-breast-height (dbh), (2) deep multi-layered canopy with at least four conifer snags of at least 50.8 cm (20 inches) dbh, and (3) at least 7.34 Mg logs/ha (20 tons of logs per acre) greater than 58.4 cm (23 inches) dbh and at least 15 m long (Old-growth Definition Task Group 1986). Fragmentation of the once-dominant old-growth forest has steadily increased from 1900 to 1990 (Figs. 1 and 2). Particularly since the 1950s, over half of the mature forests in this district has been logged.

Logging in the ONF has transformed the landscape from continuous forest to its current mosaic (Rosenberg and Raphael 1990). This configuration contains the following landscape components (macrohabitats):

1) Old-growth forest (age-class [greater than] 160 yr), further broken down into: a) continuous forest: areas of old-growth forest [greater than]50 [km.sup.2]; b) fragments: patches of old-growth forest ranging from 0.1 to 2 [km.sup.2] and separated completely from old-growth forests by clearcuts or successional habitats; and c) corridors: linear bands of old-growth forest [less than]1 km across at their greatest width, [greater than]8 km in length, and connected to continuous forest;

2) Second-growth forest: stands of monospecific, even-age-class trees, 26-80 yr following harvesting; and

3) Clearcut: sites [less than]26 yr following harvesting, with an absence of trees [greater than]3 m in height and 3 cm dbh.

Field methods

We conducted field studies June--August from 1994 to 1997. Each continuous forest site contained five survey stations; depending on size, each fragment had two to five stations; and each corridor site and each second-growth site had stations. To facilitate comparisons between corridor and second-growth sites for a complementary study (Perault and Lomolino 2000), second-growth sites were paired with, and located between 100 and 500 m of, each corridor. This spatial arrangement may have underestimated the distinctness of second-growth sites (but see Results). At each site within the five macrohabitats studied, stations were spaced 75 m apart and were situated at least 75 m from the nearest forest edge.

At each station, live traps (pitfalls, Shermans, and Tomahawks), infrared-triggered cameras, and sign surveys were used to detect local mammals. Live-trapping surveys took place over a 5-d pre-bait period followed by 7d of trapping. Live traps were set within a 6-rn radius of the station center and in a variety of available microhabitats. Five 2-L pitfall traps, four Sherman live traps (one 10.2 X 11.4 X 38.1 cm [4.0 X 4.5 X 15 inches], and three 7.6 X 7.6 X 22.9 cm [3.0 X 3.0 X 9 inches]; H. B. Sherman Traps, Tallahassee, Florida), one chipmunk-sized (12.7 X 12.7 X 40.8 cm [5 X 5 X 16 inches]) Tomahawk livetrap, and one squirrel-sized Tomahawk live-trap (15.2 X 15.2 X 61.0 cm [6 X 6 X 24 inches]; Tomahawk Live Trap, Tomahawk, Wisconsin). Sherman traps were baited with a mixture of peanut butter and oats, while Tomahawk traps were baited with raw chicken, cracked corn, apples, carrots and peanut butter and oats. The pitfalls were placed along a line [sim]1 m apart near the periphery of each trap station. She rman traps were placed along the four cardinal directions at [sim]5 m from of the station center. Tomahawk traps were set within 6 m of the station center and located near stumps, logs, and trees. The traps were locked open for the 5-d pre-bait session, then unlocked, rebaited, and checked daily for the next 7 d. All small mammals captured were weighed, measured, sexed, aged, marked by toe clipping, and released. Relative frequencies for each trapped species were determined by dividing the number of individuals captured (excluding recaptures) by the number of functional trap nights. Functional trap nights were calculated by subtracting from the total potential number of trap nights 1.0 for traps that were not functional and 0.5 for traps that were disturbed, missing bait, or contained a recaptured individual (Songer et al. 1997).

In addition to live trapping, infrared-triggered cameras and sign surveys were used to detect the presence of larger or more secretive animals. One camera station was established between every two trapping stations and no closer than 75 m from the nearest trap stations. Camera stations were baited with raw chicken, peanut butter and oats, and cracked corn and were run for the duration of each 12-d trapping session. Sign surveys also were conducted throughout each session by searching for scats, tracks, feeding signs, and dens within and along the paths between trap stations.

At each trap station, 22 environmental characteristics were recorded during the trapping session (Table 1). Two 10-m ropes, knotted at 1-m intervals, were placed along the cardinal directions crossing at 90-degree angles at the center of the site. Under each knot we recorded the presence of litter, rock, fern, moss, herbaceous plant, shrub, stump, log, or tree. We also counted the number of snags and measured the size of trees, logs, and stumps within a 10-m radius of the plot center. Size categories included trees, stumps, and logs that were [less than]20 cm dbh, 20-40 cm dbh, and [greater than]40 cm dbh. Canopy closure was measured with a spherical densiometer and a clinometer was used to estimate slope and canopy height. The distance from the site to the nearest edge of the macrohabitat was also recorded.

Statistical analyses

We used a resampling program to test for statistical significance of differences in species richness among macrohabitats. Basically, resampling refers to randomization programs that compare the results of repeated samples from two or rnore groups of data to estimate whether the observed differences among those groups are significant. One major advantage of resampling methods is that they make no assumptions about the underlying distributions of pertinent data (i.e., we felt it inappropriate to use parametric statistical approaches that assumed that species richness was normally distributed; see general discussions on randomization and resampling by Manly [1994], Simon [1995], and Gotelli and Graves [1996]). Therefore, we compared species richness among macrohabitats using RESAMPLING STATS (Simon 1995). For each pairwise combination of macrohabitats (i.e., continuous old-growth vs. corridors, continuous old-growth vs. fragments, etc.) we randomly selected five stations from each macrohabitat and then compared the species richness of these random samples from the two macrohabitats. We repeated this resampling procedure 1000 times and counted the number of times richness of samples from one macrohabitat ([C.sub.1]) exceeded that of the other ([C.sub.2]; the resampling iteration number of 1000 was used because preliminary simulations indicated that resampling means converged after iterations exceeded 100; see Simon 1995). We then calculated the mean richness for each macrohabitat (over 1000 random samples of five stations) and expressed the significance of differences in richness among these means as Min([C.sub.1], [C.sub.2])/([C.sub.1] + [C.sub.2]); i.e., the proportion of iterations for which richness of the species poor macrohabitat exceeded that of the richer habitat.

We estimated the similarity in species composition among macrohabitats using both the Jaccard index and, because species richness varied among macrohabitats, we also used the Simpson index, which is much less sensitive to variation in richness among samples. As above, we first randomly selected five stations from each of two macrohabitats and then calculated similarity indices for this pair of samples. We repeated this procedure 1000 times for each pair of macrohabitats and then calculated the means for both Jaccard and Simpson indices of similarity. We also used similarity measures to estimate between-site differences in species composition, i.e., beta diversity, within each macrohabitat. For each macrohabitat, we randomly selected two stations and then calculated Simpson's index of similarity. We repeated this resampling process 500 times and then calculated the mean similarity between stations. Beta diversity was then calculated as 1.0 minus this mean value.

We used the multiple discriminant-functions-analysis procedure provided by SYSTAT (Wilkinson 1997), which in this case conducts a canonical-variates analysis, to test whether mammalian-community structure differed significantly among the five macrohabitats. Bascially, multiple discriminant analysis tests whether samples from pre-assigned categores (in this case, macrohabitats) can be distinguished based on a set of associated characteristics (here, species or local environmental characteristics). Orthogonal combinations of variables (discriminant functions), which allow separation among groups, are obtained and the discriminatory ability of these functions can be estimated by comparing within- to between-group variances. We first calculated the proportion of stations in each site that was occupied by each of the 11 most common species, and then used these data as independent variables in the discriminant analysis. In addition to obtaining information on the statistical significance of discrimination among si tes within the five macrohabitats and the classification success for each, we saved the canonical-variate scores for each site (generated by the discriminant functions) to illustrate the separation among macrohabitat sites based on species composition of mammals. Jackknifed methods were used to estimate the success of the discriminant function in assigning each site to one of the five macrohabitats.

Biological interpretations of canonical discriminant functions were determined by inspection of standardized coefficents of the first three functions and the means for each environmental variable within each of the five macrohabitat groups. The first discriminant function of species variables on macrohabitats was a direct measure of the incidence of forest deer mice (Peromyscus oreas) and an indirect measure of the incidences of common deer mice (P. maniculatus) and black-tailed deer (Odocoileus hemionus) (standardized coefficients = 0.43, -0.90 and -0.39, respectively; all other standardized coefficients [less than]0.25; eigenvalue and percentage overall dispersion of canonical space explained by this component = 3.23 and 78%). The second discriminant function of species variables was a direct measure of the incidence of red-back voles (Clethrionomys gapperi) (standardized coefficient = 0.90 vs. [less than]0.45 for all other environmental variables; eigenvalue and percentage overall dispersion = 0.67 and 16 %). The third discriminant function increased with the incidence of Townsend's chipmunks (Eutamias townsendii), and decreased with the incidence of black bears (Ursus americanus) (standardized coefficients [less than]0.53 and -0.45, respectively; all other coefficients [less than]0.35; eigenvalue and percentage of overall dispersion = 0.18 and 4%).

We performed a second discriminant-functions analysis of macrohabitats, this time based on environmental variables at each site (see Table 1). For each site, we first calculated the means for the 22 environmental variables recorded at each station. We then used the multiple discriminant-analysis procedure provided by SYSTAT (Wilkinson 1997) to test whether local environmental conditions varied significantly among macrohabitats. Again, we recorded the statistical significance and classification success of the discriminant function and saved the canonical-variate scores to illustrate the separation among macrohabitats based on their environmental characteristics.

The first discriminant function of environmental characteristics on macrohabitats was primarily a measure of forest development, loading most strongly on canopy closure and canopy height (standardized coefficients = -1.58 and -1.40, respectively; all other standardized coefficients [less than]0.40; eigenvalue and percentage overall dispersion of canonical space explained by this component = 20.98 and 81%). The second discriminant function was a direct measure of litter accumulation (standardized coefficient = 0.96 vs. [less than]0.60 for all other environmental variables; eigenvalue and percentage overall dispersion = 4.06 and 5%). The third discriminant function was a direct measure of coverage of moss and distance to edge of the forest (standardized coefficients = 0.70 and 0.57, respectively; all other coefficients [less than]0.45; eigenvalue and percentage of overall dispersion = 0.60 and 3%).

In order to test for association between mammalian-community structure and local environmental conditions at each station, we conducted two principal-components analyses (PCA) using SYSTAT (Wilkinson 1997), one based on the 22 environmental variables and one based on incidences of the 11 most common species of mammals. We thus obtained measures of environmental and mammal community characteristics, independent of the discriminant analyses of macrohabitats. We then used the correlation procedure in SYSTAT (Wilkinson 1997) to estimate Pearson product-moment correlations between the first three factor scores derived from PCA of mammalian communities and those derived from PCA of local environmental characteristics.

To test whether particular species were randomly distributed among macrohabitats we conducted Chi-square tests of independence based on distributions of the 11 species that were recorded at [greater than]10% of the 263 stations surveyed (categories = present or absent in each of the five macrohabitats; df = 4). These species included common deer mice, forest deer mice, red-backed voles, montane shrews, Trowbridge's shrews, mountain beaver (Aplodontia rufa), Townsend's chipmunks, spotted skunks (Spilogale putorius), Douglas squirrels, black-tailed deer, and black bears.


A total of 84 sites (263 stations) were sampled: 15 in continuous old-growth forest in the Olympic National Forest (northwest Washington State, USA), 21 in old-growth corridors, 20 in old-growth forest fragments, 9 in second-growth forests, and 19 in clearcuts. Twenty-four species were detected from traps, cameras, and sign surveys (Table 2). For the 17 trapped species, nearly 20 000 functional trap nights produced 2282 unique individuals. Just under half (45.6%) of these were Peromyscus oreas. Clethrionomys gapperi was the second most common species, accounting for 18.4% of all individuals trapped. Sorex monticolus, Eutamias townsendii, Sorex trowbridgii, Peromyscus maniculatus, Spilogale putorius, and Glacomys sabrinus together made up 32.2% of the individuals captured. The remaining nine trapped species, combined, made up just 3.8% of the captures, with no single species accounting for more than 1% of the total (Table 2).

Species richness

Species richness of all mammals (measured as species density = number of species per five randomly chosen stations) differed among macrohabitats, with highest levels occurring in the native old-growth forests (continuous old-growth, old-growth fragments, and old-growth corridors; Fig. 3). Mean species density of mammals inhabiting second-growth communities was only 73% of that for continuous old-growth communities (7.2 vs. 9.9 species per five stations, respectively; P = 0.042, based on resampling tests described in Methods Statistical analyses, above). The difference between richness of communities in second-growth and corridors sites was marginally significant (7.2 vs. 8.8 species per five stations; P = 0.057).

As anticipated, species richness of forest mammals considered separately (see Table 2) tended to be highest for continuous old-growth communities, slightly lower for corridor and fragment communities, intermediate for those in second-growth forests, and lowest for mammal communities of clearcuts (Fig. 3, black bars). Mean species density of forest mammals in clearcuts was less than half that of any of the old-growth communities (continuous old-growth, corridors, or fragments; P = 0.021, 0.047, and 0.042, respectively; randomization test). Mean species density of forest mammals in clearcuts also was significantly lower than that in second-growth forests (P = 0.044).

Species composition

Patterns in species composition among the macrohabitats were similar to those of species richness. Similarity in species composition was highest among pairwise comparisons that included at least one of the old-growth macrohabitats (continuous old-growth, fragment, or corridor), intermediate for comparisons that included communities from second-growth forests, and lowest for communities from clearcut sites (Table 3). Because the Jaccard index of similarity is influenced by differences in species richness (i.e., species-poor sites tend to have relatively low similarity), we also calculated Simpson's index of similarity, which measures the proportion of species in the species-poor community that also occur in the species-rich community. Although consistently higher than Jaccard's index for all pairwise comparisons in this study, results using Simpson's index yielded qualitatively identical inferences, i.e., high similarity among communities from the three old-growth macrohabitats, intermediate values for compar isons that included second-growth, and lowest similarity for those that included clearcuts. Thus, in addition to having fewer species, second-growth forests and especially clearcuts were inhabited by the most distinct assemblages of mammals. Communities from second-growth and clearcuts also exhibited the lowest beta diversities (Table 3), i.e., species composition within these macrohabitats exhibited relatively little variation from station to station.

Discriminant analyses provided additional descriptions of differences in species composition among macrohabitats. Again, mammal communities of second-growth and clearcuts tended to be more distinct in species composition than communities from the three old-growth macrohabitats (based on higher pairwise F values and higher classification success; Table 4, values above the diagonal). Communities of clearcuts were the only ones dominated by both the common deer mouse (Peromyscus maniculatus) and by black-tailed deer (Odocoileus hemionus) (Fig. 4). In all other habitats, the forest deer mouse (P. oreas) was the most frequently detected species. Second-growth communities were distinguished by having the highest overall incidence of spotted skunks (Spilogale putorius; Fig. 5e) along with relatively low incidence levels of red-backed voles, mountain beavers, and montane shrews (Fig. 4). Species composition of communities from corridors overlapped broadly with communities from other old-growth macrohabitats (continu ous old-growth and fragments), recording the lowest classification success of all macrohabitats studied (Table 4). Corridors, however, did have the highest incidence of forest deer mice, Trowbridge's shrews, elk (Cervus elaphus), and black bears; Fig. 5).

Although communities from the two remaining macrohabitats (continuous old-growth and fragments) overlapped broadly in species composition (see Fig. 4), continuous old-growth communities had the highest overall incidence of flying squirrels (Glaucomys sabrinus), Douglas squirrels, montane shrews, and mountain beavers (Figs. 4 and 5). On the other hand, communities from fragments had the highest overall incidence of Townsend's chipmunks and red-backed voles, along with intermediate incidences of montane and Trowbridge's shrews, black bear, and black-tailed deer.

Habitat selection and differences in environmental characteristics among macrohabitats

Eleven species occurred frequently enough to test whether they were nonrandomly distributed among the five macrohabitats (Chi-square tests of independence). Of these, all but two species (black bears and Trowbridge's shrews) exhibited significantly nonrandom distributions among macrohabitats ([[chi].sup.2] = 6.02, df = 3, P [greater than] 0.110 for black bears, [[chi].sup.2] = 8.95, df = 4, 0.10 [greater than] P [greater than] 0.062 for Trowbridge's shrews; P = 0.036 for spotted skunks; P [leq] 0.01 for red-backed voles, forest deer mice, montane shrews, Douglas squirrels, mule deer, common deer mice, mountain beaver, and Townsend's chipmunk).

Results of discriminant-functions analyses of sites based on environmental features were similar to those based on species composition (Table 4; compare Figs. 4 and 6a). Clear-cut sites were easily discriminated from all others based on their relatively low canopy closure and low crown height, while those of second growth had relatively high accumulations of litter and high frequencies of small trees (dbh [less than] 40 cm; Fig. 6a). Classification success of sites based on environmental features was 100% for both clearcuts and second-growth forests. Corridors, in contrast, were the least distinctive habitats, exhibiting the poorest classification success (43%) of all five macrohabitats considered (vs. 20% for the null expectation).

Given the overlap of old-growth sites in the bivariate space of Fig. 6a, we ran a separate discriminant-functions analysis based on environmental characteristics of just the three old-growth macrohabitats considered separately. Although they continued to overlap somewhat in the bivariate, environmental space of Fig. 6b, discrimination among these old-growth sites was significant (P = 0.036, Wilk's lambda = 0.22, approximated F = 1.632, df = 44, 64; overall jackknifed classification success = 52%, vs. 33% for the null expectation). Again, corridors exhibited the greatest dispersion across the bivariate, environmental space of Fig. 6b, and the lowest classification success of the three old-growth macrohabitats (success = 43, 53, and 60% for sites in corridors, continuous forests, and fragments, respectively). Old-growth fragments tended to have relatively low levels of canopy closure, while continuous forests were characterized by being located further from edge habitats and having relatively high levels of moss.

The relationship between environmental features and assembly of mammal communities in the Olympic National Forest was more directly evidenced by correlations between factor scores based on two separate principal-components analyses--one based on species composition (SS1, SS2, and SS3) and the other based on local environmental variables (ES1, ES2, and ES3; Table 5).

SS1, which loaded negatively on common deer mice, but positively on forest deermice, Douglas squirrels, montane shrews, and Trowbridge's shrews (absolute values of these loadings [greater than]0.50) was significantly correlated with ES1 (P [less than] 0.01), which was a measure of forest development (loading negatively on exposed soil, and positively on canopy closure, canopy height, and moss; loadings [greater than] 0.65; percentage variance explained by SS1 and ES1 = 20.9 and 25.5, respectively). The suite of species associated with SS1 also were significantly (P [less than] 0.05) correlated with ES3, which was a direct measure of the incidence of large ([greater than]40 cm diameter) stumps, and medium to large logs (20-40, and [greater than]40 cm diameter; loadings of these variables on ES3 [greater than] 0.50, percentage variance explained = 8.2%). Simply put, the distributions and incidences of what are generally perceived to be old-growth species (i.e., forest deer mice, Douglas squirrels, montane shre ws, and Trowbridge's shrews) increased as the old-growth character of the local environment increases.

Finally, SS3, which described a gradient of communities dominated by Townsend's chipmunks and black bears to those dominated by spotted skunks (loadings [greater than]0.40) was significantly and positively correlated with ES2 (P [less than] 0.05), which was a direct measure of the frequency of small ([less than]20 cm diameter) logs and small stumps (loadings [greater than]0.70; percentage variance explained by SS3 and ES2 = 10.6% and 10.8%, respectively). That is, incidence of spotted skunks increased, and that of Townsend's chipmunks and black bears decreased along a gradient of increasing frequency of small woody debris.


Anthropogenic modification of the Olympic National Forest landscape has created a mosaic of distinct ecosystems and mammalian communities. Differences in mammalian-community structure across this fragmented landscape is influenced by a number of factors, including habitat selectivity of the component species and significant variation in local environmental conditions of the principal landscape components (i.e., continuous old-growth, corridors, fragments, second-growth forests, and clearcuts).

Second-growth forests, with their relatively low structural and vegetative diversity (monospecific, evenage stands of trees with understories nearly devoid of moss, ferns, and downed logs), exhibited relatively low mammalian richness (over all species) and low beta diversity. Clearcuts, which also exhibited low richness of mammals in general, had by far the lowest diversity of old-growth species. Thus, conversion of old-growth forests to clearcuts and second-growth forests, while promoting landscape diversity and high densities of common deer mice, deer, and skunks, reduces local and regional diversity of native, old-growth mammals.

Results of discriminant-functions analysis and related tests (Figs. 4 and 6, Table 5) suggest that difference in mammalian-community structure among macrohabitats at least partially derives from variation in local environmental characteristics (especially those typically associated with old-growth features--canopy closure, canopy height, large logs and stumps, and moss; see Old-growth Definition Task Group 1986). Yet, the correlations between species composition and environmental variables is far from perfect, indicating that observed differences in mammalian-community structure among macrohabitats also may be influenced by factors in addition to local environmental conditions. Among such factors are landscape and biogeographic characteristics including patch size, shape, and isolation, relative amount of edge, and nature of the juxtaposed habitats. Although beyond the scope of the present paper, the relative importance of local vs. landscape-level factors, and the potential interactive effects among scales, is an intriguing subject for future studies.

Given, however, the current state of our knowledge, we caution against overgeneralizing from our results to formulate generic prescriptions for recovery or rehabilitation of old-growth forests. Attempts at ameliorative modification of one or even a number of environmental variables (e.g., reducing litter accumulation or increasing the density of downed logs) are unlikely to guarantee full restoration of old-growth forest communities--not unless niches of the component species are much simpler then we envision. On the contrary, as Hutchinson (1959), MacArthur (1972) and most ecologists believe, niches of most species, especially those adapted to structurally complex and stable ecosystems such as temperate rain forests, are likely to be complex and multifactorial. Therefore, while the general, qualitative responses to deforestation and fragmentation may be predicted with a reasonable degree of accuracy, once temperate rain forest communities have been so altered, the challenge of restoring the structure of the se complex systems may be well beyond our current abilities to understand and manage.

It is clear from the results presented here, however, that the effects of clear-cutting are both significant and persistent. Many of the clearcuts we studied were over 15 yr old, but still showed little similarity with any other macrohabitat in terms of environmental characteristics or mammalian-community structure. Even second-growth forests, which ranged from 26 to 80 yr post-harvest, remain significantly different from other macrohabitats and communities. Based on inspection of Figs. 4 and 6a, even when second-growth stands are mature and ready for harvesting, they do not appear to be converging on the structure of native old-growth forests. In fact, second-growth forests may well be on an alternative successional trajectory (Fig. 6a: arrow). A number of factors may be responsible for this--including a combination of differences between natural disturbances and successional processes, and practices typically employed during harvesting of these forests. These practices include:

1) Clearing of large tracts of land within a relatively short period of time (in comparison to the loss of trees due to blow down or other natural processes that create snags and minor gaps, sparsely distributed across time and space).

2) Burning of remaining litter and other organic matter following harvesting in a region where rainfall levels are so high that natural burns are spatially very limited and typically occur less than once per century.

3) Planting of even-age stands of non-native, genetically selected trees, (typically of just one species and variety) across a large expanse of the disturbed landscape.

4) Frequent, wholesale loss of soil down the steeply sloped terrain, which sets succession back to primary stages. Also note that the current climatic conditions, and those predicted under existing scenarios of global warming, may well be quite different from those that existed when these forests first became established (see Norse 1990).


This landscape-scale, anthropogenic experiment in the Olympic National Forest provides some important insights into the assembly and what we call "disassembly" of native forest communities. The effects of anthropogenic transformation of this temperate rain forest landscape are significant, persistent, and interpretable. Although a significant portion of the differences in community structure among macrohabitats can be attributed to differences in their local, environmental characteristics (e.g., vegetative structure), landscape-and biogeographic-scale factors are also likely to be important. We are currently investigating the hypothesis that mammalian-community structure in these forests is strongly influenced by the independent and perhaps interactive effects of local and landscape-scale factors (Perault and Lomolino 2000, M. V. Lomolino and D. R. Perault, unpublished manuscript).

Returning to one of our initial questions, it does appear that the effects of anthropogenic transformation of this temperate rain forest landscape is qualitatively similar to that of its tropical counterpart (see excellent review and papers on fragmentation of tropical rain forests in Laurance and Bierregaard [1997]). Both biotic and abiotic conditions of alternative habitats (clearcuts and second-growth forests) differ from old-growth sites, and communities in the former are dominated by edge species, i.e., those typically not found in mature old growth. The response of animal communities to deforestation appears to be rapid for both temperate and tropical rain forests, with demonstrable changes in animal community structure occurring within just a few months following harvest. These preliminary comparisons between tropical and temperate rain forest must, however, remain tentative until more controlled, comparative studies are conducted. Such studies may do well to focus on differences in resilience of trop ical and temperate rain forests, While it may take only a decade or so to establish a gallery forest of 10-30 m high trees in the lowland tropics, once cut, temperate rain forests are likely to require centuries of favorable conditions to redevelop the character of their native forests. Again, despite the 60-80 years of succession, the environmental characteristics and mammalian-community structure of second-growth forests in the Olympics did not show any signs of converging on that of native, old-growth forests.

This and subsequent studies of this system also may provide information critical to one of the most important lines of research in modern conservation biology-- the role of corridors in conserving native species. Here, we found that corridors of this portion of the Olympic National Forest had relatively high richness of mammals and tended to overlap broadly in species composition and environmental characteristics with other old-growth forests (i.e., continuous old-growth sites and fragments). This suggests that what we termed "corridors" may indeed ameliorate some of the effects of deforestation and fragmentation, but many questions remain. Does the relative importance of these linear bands of old-growth attenuate with distance from the continuous old-growth forest? If so, at what distance do the diversities and densities of old-growth species begin to decline with increasing isolation? In addition to potentially serving as conduits for movements of old-growth species, do these corridors also serve as demogr aphic sources (i.e., do old-growth species breed frequently enough in these macrohabitats to allow net emigration into other sites)? Why, in comparison to all other macrohabitats studied, do corridors exhibit such high variability in environmental characteristics and mammalian-community structure?

We are optimistic that we can, in subsequent studies, answer some, if not most of these questions (see Perault 1998, Perault and Lomolino 2000), and we hope that other ecologists also take advantage of similar anthropogenic experiments to develop a better understanding of various processes influencing biological diversity of anthropogenically transformed landscapes.

Finally, we return to our original theme of assembly and disassembly of ecological communities. Broadly interpreted, community assembly includes any processes that result in nonrandom or unbalanced assemblages of species in comparison to the regional pool of species. Given this, we conclude that the natural, slow processes involved in the initial development of ancient temperate rain forests, with the coincident changes in local environmental characteristics, clearly are assembly processes that resulted in the formation of highly structured assemblages of non-volant mammals in old-growth forests. Anthropogenic disturbances of these otherwise stable but non-resilient ecosystems, however, have resulted in disassembly--the non-random relaxation of native communities and formation of alternative, anthropogenic communities (Fox 1987, Mikkelson 1993). Communities in clearcuts and second-growth forests constitute nonrandom subsets of the species pool, which differ significantly from assemblages of mammals inhabitin g old-growth forests, and they show little evidence of converging on the environmental characteristics and community structure of these ancient, temperate rain forests.


We thank T. A. Franklin, A. Leikam, P. Leimgruber, M. Rene, B. Rosewell, and M. A. Songer for their assistance in collecting field data. R. B. Channell, R. L. Cifelli, N. Czaplewski, M. Jakubauskas, M. E. Kaspari, G. A. Smith, C. C. Vaughn, and M. Yuan provided valuable input to this work. GIS data were supplied by W. Wettengel of the Olympic National Forest. J. Lowrie and E. Milliman contributed additional information about the research area and provided logistical support from the Olympic National Forest. Field housing was furnished by Terry Neilson at the Satsop Wells Environmental Learning Lodge of the Grays Harbor Conservation District. Funding for this project was provided by two National Science Foundation grants (DEB-9322699, DEB-9707204) to M. V. Lomolino.

(1.) Oklahoma Biological Survey, Oklahoma Natural Heritage Inventory, University of Oklahoma, Norman, Oklahoma 73072 USA

(2.) Department of Zoology, University of Oklahoma Norman Oklahoma, 73072 USA

(3.) E-mail:

(4.) Present address: Department of Biology and Environmental Science, Lynchburg College, Lynchburg Virginia, 24501 USA.


Adam, P. 1992. Australian rainforests. Oxford University Press, New York, New York, USA.

Brown, J. H. 1995. Macroecology. University of Chicago Press, Chicago, Illinois, USA.

Brown, J. H., and M. V. Lomolino. 1998. Biogeography, Sinauer Associates, Sunderland, Massachusetts, USA.

Brown, K. S. 1997. Disturbance, fragmentation, and the dynamics of diversity in Amazonian forest butterflies. Pages 91-110 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical remnants: ecology, management, and conservation of fragmented communities. University of Chicago Press, Chicago Illinois, USA.

Carey, A. B. 1995. Sciurids in Pacific Northwest managed and old-growth forests. Ecological Applications 5:648-661.

Carey, A. B., and M. L. Johnson. 1995. Small mammals in managed, naturally young, and old-growth forests. Ecological Applications 5:336-352.

Connor, E. H., and D. Simberloff. 1979. The assembly of species communities: chance or competition? Ecology 60:1132-1140.

Connor, E. H., and D. Simberloff. 1983. Interspecific competition and species co-occurrence patterns on islands: null models and the evaluation of evidence. Oikos 41:455-465.

Craig, L. J., and D. S. Dobkin. 1993. Community dynamics of small mammals in mature and logged Atlantic white cedar swamps of the New Jersey pine barrens. Bulletin Number 487. New York State Museum, Albany New York, USA.

Diamond, J. M. 1975. Assembly of species communities. Pages 342-444 in. M. L. Cody and J. M. Diamond, editors. Ecology and evolution of communities. Harvard University Press, Cambridge, Massachusetts, USA.

Diamond, J. M., and M. Gilpin. 1982. Examination of the "null" model of Connor and Simberloff for species co-occurrences on islands. Oecologia 52:64-74.

Dunstan, C. E., and B. J. Fox. 1996. The effects of fragmentation and disturbance of rainforest on ground-dwelling small mammals on the Robertson Plateau, New South Wales, Australia. Journal of Biogeography 23:187-201.

Edwards, P. J., R. M. May, and N. R. Webb. 1994. Large-scale ecology and conservation biology. Blackwell Scientific, London, UK.

Elliot, J. M. 1994. Qualitative ecology and the brown trout. Oxford University Press, Oxford, UK.

Fox, B. J. 1987. Species assembly and the evolution of community structure. Evolutionary Ecology 1:201-213.

Gentry, A. W., editor. 1990. Four Neotropical rainforests. Yale University Press, New Haven, Connecticut, USA.

Gilpin, M., and J. M. Diamond. 1982. Factors contributing to non-randomness in species co-occurrence on islands. Oecologia 52:75-84.

Gilpin, M., and J. M. Diamond. 1984. Are species co-occurrences on islands non-random, and are null hypotheses useful in community ecology? Pages 297-3 15 in D. R. Strong, D. Simberloff, L. G. Abele, and A. B. Thistle, editors. Ecological communities: conceptual issues and the evidence. Princeton University Press, Princeton, New Jersey, USA.

Gotelli, N. J., and G. R. Graves. 1996. Null models in ecology. Smithsonian Institution Press, Washington, D.C., USA.

Harvey, P. H., and M. D. Pagel. 1991. The comparative method in evolutionary biology. Oxford University Press, Oxford, UK.

Hutchinson, G. E. 1959. Homage to Santa Rosalia, or why are there so many kinds of animals? American Naturalist 93:145-159.

Karieva, P., and M. Anderson. 1988. Spatial aspects of species interactions: the wedding of models and experiments. Pages 35-50 in A. Hastings, editor. Community ecology. Springer Verlag, New York, New York, USA.

Kirkland, G. L. 1990. Patterns of initial small mammal community change after clearcutting of temperate North American forests. Oikos 59:313-320.

Klein, B. C. 1989. The effects of forest fragmentation on dung and carrion beetle (Scarabaeinae) communities in central Amazonia. Ecology 70:1715-1725.

Larrison, E. J., and A. C. Fisher. 1976. Mammals of the northwest: Washington, Oregon, Idaho and British Columbia. Seattle Audubon Society, Seattle, Washington, USA.

Laurance, W. F. 1989. Ecological impacts of tropical forest fragmentation on non-flying mammals and their habitats. Dissertation. University of California, Berkeley, California, USA.

Laurance, W. F 1990. Comparative responses of five arboreal marsupials to tropical forest fragmentation. Journal of Mammalogy 71:641-653.

Laurance, W. F. 1991. Edge effects in tropical forest fragments: application of a model for the design of nature reserves. Biological Conservation 57:205-219.

Laurance, W. F., and R. O. Bierregaard, Jr., editors. 1997. Tropical remnants: ecology, management, and conservation of fragmented communities. University of Chicago Press, Chicago Illinois, USA.

Lomolino, M. V. 1986. Mammalian community structure on islands: immigration, extinction and interactive effects. Biological Journal of the Linnaean Society 28: 1-21.

Lomolino, M. V. 1996. Investigating causality of nestedness of animal communities: selective immigrations or extinctions? Journal of biogeography 23:699-713.

Lovejoy, T. E., R. O. Bierregaard, Jr., A. B. Rylands, J. R. Malcolm, C. E. Quintela, L. H. Harper, K. S. Brown, Jr., A. H. Powell, H. O. R. Schubart, and M. B. Hays. 1986. Edge and other effects of isolation on Amazon forest fragments. Pages 257-285 in M. E. Soul[acute{e}], editor. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunderland, Massachusetts, USA.

MacArthur, R. H. 1972. Geographical ecology. Princeton University Press, Princeton, New Jersey, USA.

Malcolm, J. R. 1988. Small mammal abundances in isolated and non-isolated primary forest reserves near Manaus, Brazil. Acta Amazonica 18:67-83.

Manly, B. F. J. 1994. Randomization and Monte Carlo methods in biology. Chapman and Hall, New York, New York, USA.

Marshall, L. G., S. D. Webb, J. J. Sepkoski, Jr., and D. M. Raup. 1982. Mammalian evolution and the great American interchange. Science 215:351-357.

Martell, A. M. 1983. Demography of southern red-backed voles (Clethrionomys gapperi) and deer mice (Peromyscus maniculatus) after logging in north-central Ontario. Canadian Journal of Zoology 61:958-969.

Mikkelson, G. M. 1993. How do food webs fall apart? A study of changes in trophic structure during relaxation on habitat fragments. Oikos 67:539-547.

Norse, E. A. 1990. Ancient forests of the Pacific Northwest. Island Press, Washington, D.C., USA.

Old-growth Definition Task Group [J. F. Franklin, F Hall, W. Laudenslayer, C. Maser, J. Nunan, J. Poppino, C. J. Ralph, and T. Spies.] 1986. Interim definitions for old-growth Douglas-fir and mixed-conifer forests in the Pacific Northwest and California. Note PNW-447. USDA Forest Service, Pacific Northwest Research Station, Portland, Oregon, USA.

Patterson, B. D., and W. Atmar. 1986. Nested subsets and the structure of insular mammalian faunas and archipelagoes. Biological Journal of the Linnaean Society 28:65-82.

Patterson, B. D., and J. H. Brown. 1991. Regionally nested patterns of species composition in granivorous rodent assemblages. Journal of Biogeography 18:395-402.

Perault, D. R. 1998. Landscape heterogeneity and the role of corridors in determining the spatial structure of insular mammal populations. Dissertation. University of Oklahoma, Norman, Oklahoma, USA.

Perault, D. R., and M. V. Lomolino. 2000. Corridors and mammal community structure across a fragmented, old-growth forest landscape. Ecological Monographs 70: in press.

Peterson, D. L., E. G. Sehreiner, and N. M. Buckingham. 1997. Gradients, vegetation and climate: spatial and temporal dynamics in the Olympic Mountains, U.S.A. Global Ecology and Biogeography Letters 6:7-17.

Raphael, M. G. 1988. Long-term trends in abundance of amphibians, reptiles, and mammals in Douglas-fir forests of northwestern California. Pages 23-31 in R. C. Szaro, K. F. Severson, and D. R. Patton, technical coordinators. Proceedings of the symposium management of amphibians, reptiles, and small mammals in North America (Fort Collins, Colorado, USA). USDA Forest Service General Technical Report RM-166.

Rosenberg, K. V., and M. G. Raphael. 1990. Effects of fragmentation on vertebrates in Douglas- fir forests. Pages 263- 272 in J. Verner, M. L. Morrison, and C. T. Ralph, editors. Modeling habitat relationships of terrestrial vertebrates. University of Chicago Press, Chicago, Illinois, USA.

Saunders, D. A., and R. J. Hobbs, editors. 1991. Nature conservation: the role of corridors. Surrey Beatty & Sons, Chipping Norton, Australia.

Simon, J. L. 1995. Resampling Stats users guide. Resampling Stats, Arlington, Virginia, USA.

Simpson, G. G. 1980. Splendid isolation: the curious history of mammals in South America. Yale University Press, New Haven, Connecticut, USA.

Songer, M. A., M. V. Lomolino, and D. R. Perault. 1997. Habitat selection and niche dynamics of deermice (Peromyscus oreas and P. maniculatis) in a fragmented, old-growth forest landscape. Journal of Mammalogy 78:1027-1039.

Taylor, C. M. 1997. Fish species richness and incidence patterns in isolated and connected stream pools: effects of pool volume and spatial position. Oecologia 110:560-566.

Tilman, D. 1989. Ecological experimentation: strengths and conceptual problems. Pages 136-157 in G. E. Likens, editors. Long-term studies in ecology. Springer Verlag, New York, New York, USA.

Tocher, M. D., C. Gascon, and B. L. Zimmerman. 1997. Fragmentation effects on a Central Amazonian frog community. Pages 124-137 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropcial forest remnants: ecology, management, and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois, USA.

Warburton, N. H. 1997. Structure and conservation of forest avifauna in isolated rainforest remnants in tropical Australia. Pages 190-206 137 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropcial forest remnants: ecology, management, and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois, USA.

Weatherhead, P. J. 1986. How unusual are unusual events? American Naturalist 128:150-154.

Webb, D. S. 1991. Ecogeography and the great American interchange. Pages 357-386 in F. G. Stehli and S. D. Webb, editors. The great American biotic interchange. Plenum, New York, New York, USA.

Weiher, E. A., and P. A. Keddy. 1998. The search for assembly rules in ecological communities. Cambridge University Press, Cambridge, UK.

Wilkinson, L. 1997. SYSTAT for Windows: statistics, version 7.0.1. SYSTAT, Evanston, Illinois, USA.

Wright, D. H., B. D. Patterson, G. M. Mikkelson, A. Cutler, and W. Atmar. 1998. A comparative analysis of nested subset patterns of species composition. Oecologia 113:1-20.
COPYRIGHT 2000 Ecological Society of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2000 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Geographic Code:1U9WA
Date:Jun 1, 2000

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters