Samuel A. Cushman

COMPARATIVE EVALUATION OF EXPERIMENTAL APPROACHES TO THE STUDY OF HABITAT FRAGMENTATION EFFECTS

Ecologists have used a variety of comparative mensurative and manipulative experimental approaches to study the biological consequences of habitat fragmentation. In this paper, we evaluate the merits of the two major approaches and offer guidelines for selecting a design. Manipulative experiments rigorously assess fragmentation effects by comparing pre- and post-treatment conditions. Yet they are often constrained by a number of practical limitations, such as the difficulty in implementing large-scale treatments and the impracticality of measuring the long-term (decades to centuries) responses to the imposed treatments. Comparative mensurative studies generally involve substituting space for time, and without pre-treatment control, can be constrained by variability in ecological characteristics among different landscapes. These confounding effects can seriously limit the strength of inferences. Depending on the scale of the study system and how “landscape” is defined, both approaches may be limited by the difficulty of replicating at the landscape scale. Overall, both mensurative and manipulative approaches have merit and can contribute to the body of knowledge on fragmentation. However, from our review of 134 fragmentation studies published recently in three major ecological journals, it is evident that most manipulative and mensurative fragmentation experiments have not provided clear insights into the ecological mechanisms and effects of habitat fragmentation. We discuss the reasons for this and conclude with recommendations for improving the design and implementation of fragmentation experiments.

Hierarchical, Multi-scale decomposition of species-environment relationships

We present an adaptation of existing variance partitioning methods todecompose species-environment relationships in hierarchically-structured,multi-scaled data sets. The approach translates a hierarchical, multi-scaleconceptual model into a statistical decomposition of variance. It uses a seriesof partial canonical ordinations to divide the explained variance inspecies-environment relationships into its independent and confoundedcomponents, facilitating tests of the relative importance of factors atdifferent organizational levels in driving system behavior. We discuss themethod in the context of an empirical example based on forest bird communityresponses to multiple habitat scales in the Oregon Coast Range, USA. Theexamplepresents a two-tiered decomposition of the variation in the bird community thatis explainable by a series of habitat factors nested within three spatialscales(plot, patch, and landscape). This method is particularly suited for theproblems of hierarchically structured landscape data. The explicit multi-scaleapproach is a major step forward from conducting separate analyses at differentscale levels, as it allows comprehensive analysis of the interaction of factorsacross scales and facilitates ecological interpretation in theoretical terms.

Behavior of class-level landscape metrics across gradients of class aggregation and area

Habitat loss and fragmentation processes strongly affect biodiversity conservation in landscapes undergoing anthropogenic land use changes. Many attempts have been made to use landscape structure metrics to quantify the independent and joint effects of these processes. Unfortunately, ecological interpretation of those metrics has been plagued by lack of thorough understanding of their theoretical behavior. We explored behavior of 50 metrics in neutral landscapes across a 21-step gradient in aggregation and a 19-step gradient in area using a full factorial design with 100 replicates of each of the 399 combinations of the two factors to assess how well metrics reflected changes in landscape structure. Metric values from real landscapes were used to determine the extent of neutral landscape space that is represented in real landscapes. We grouped metrics into three major behavioral classes: strongly related to focal class area (n=15), strongly related to aggregation (n=7), and jointly responding to area and aggregation (n=28). Metrics strongly related to class area exhibited a variety of distinct behaviors, and many of these metrics have unique interpretations that make each of them particularly useful in certain applications. Metrics strongly related to aggregation, independent of class area, are particularly useful in assessing effects of fragmentation. Moreover, metrics in this group exhibited a range of specific behaviors, highlighting subtle but different aspects of landscape aggregation even though we controlled only one aspect of aggregation. The non-linear behavior exhibited by many metrics renders interpretation difficult and use of linear analytical techniques inappropriate under many circumstances. Ultimately, comprehensive characterization of landscapes undergoing habitat loss and fragmentation will require using several metrics distributed across behavioral groups.

HIERARCHICAL ANALYSIS OF FOREST BIRD SPECIES–ENVIRONMENT RELATIONSHIPS IN THE OREGON COAST RANGE

Species in biological communities respond to environmental variation si- multaneously across a range of organizational levels. Accordingly, it is important to quantify the effects of environmental factors at multiple levels on species distribution and abundance. Hierarchical methods that explicitly separate the independent and confounded influences of environmental variation across several levels of organization are powerful tools for this task. Our study used a hierarchical approach to partition explained variance in an Oregon Coast Range bird community among plot-, patch-, and landscape-level factors. We used a series of partial canonical ordinations to decompose species-environment relationships across these levels of organization to test four hypotheses about the importance of envi- ronmental control over community structure. We found that plot-level factors were better predictors of community structure than patch- or landscape-level factors. In addition, al- though landscape-level variables contributed substantial independent explanatory power, there was little evidence that patch-level environmental variability provided additional explanation of community structure beyond that provided by plot- and landscape-level factors. At higher levels of the hierarchical analysis, we found that, among plot-level factors, vegetation cover type was as powerful a predictor of community structure as detailed floristics, and more powerful than vegetation structure. At the landscape level, we found that landscape composition and configuration both provided substantial independent ex- planatory power, with landscape composition being the better overall predictor. Our results have a number of implications for sampling, analysis, and conservation. For example, misleading results could be obtained by studies conducted at a single organizational level. Also, the high degree of confounding among several levels of our analysis suggests that there is a lack of independence between the influences of environmental structure at different organizational levels. Due to this confounding, our results suggest that patch-based studies of forest-bird ecological relationships in the Oregon Coast Range may be equivocal. In addition, the power of mapped cover class as a plot-level predictor variable suggests that coarse-filter, multiscale approaches utilizing remote sensing and GIS may be nearly as effective at predicting local patterns as expensive field surveys of habitat conditions at the plot level, and more effective at predicting patterns continuously across large regions.

A multi-scale analysis of species-environment relationships: breeding birds in a pitch pine-scrub oak (Pinus rigida-Quercus ilicifolia) community

Analyses of species–environment relationships are critically needed to guide conservation of many declining species. Comprehensive investigations of these relationships must incorporate environmental variables at multiple spatial scales since species responses to the environment vary with the scale of observation. We used partial canonical correspondence analysis to associate a bird assemblage in a threatened pitch pine–scrub oak (Pinus rigida–Quercus ilicifolia) community in Plymouth County, Massachusetts, USA, with plot, patch, and landscape level variables. The first level of analysis partitioned the amount of variance in the bird community explained by plot, patch, and landscape factors from that explained by spatial autocorrelation. The second level partitioned the amount of variance explained by plot, patch, and landscape factors alone and in combination. All three levels of environmental variables together accounted for 43% of the variance in the species data and only 5% of the variance was explainable by spatial factors alone. Landscape level factors accounted for a slightly larger amount of the explained variation (12%) than plot (11%) or patch (8%) level factors. We also examined the cumulative fit of each species to the plot, patch, and landscape partial models. Nine species of regional and/or national conservation concern had distributions that fit one model considerably better than the others. The Great Crested Flycatcher (Myiarchus crinitus) and Black-and-white Warbler (Mniotilta varia) were predominantly associated with plot level factors; the Whip-poor-will (Caprimulgus vociferus) was predominantly associated with patch level factors; and the Purple Finch (Carpodacus purpureus), Gray Catbird (Dumetella carolinensis), Scarlet Tanager (Piranga olivacea), Ovenbird (Seiurus aurocapillus), Brown Thrasher (Toxostoma rufum), and Eastern Kingbird (Tyrannus tyrannus) were predominantly associated with landscape level factors. This study suggests that focusing conservation efforts at the landscape level would provide the most effective protection for the largest number of sensitive species. # 2002 Elsevier Science Ltd. All rights reserved.

Landscape-level patterns of avian diversity in the Oregon Coast Range

We used a comparative mensurative landscape-level experiment to quantify the relative importance of mature forest area and fragmentation and differences among watersheds in influencing avian community diversity in the Oregon Coast Range, USA. Our study design included three large hydrological basins, two levels of fragmentation, and six levels of mature forest area. We recorded 82 species of birds in a total of 1046 plots in 30 landscapes. Our four response variables were species richness, species density, species evenness, and total bird density. We used a combination of factorial ANOVA and partial canonical variates analysis to quantify the relative importance of differences in mature forest area, fragmentation, and basin in influencing each response variable and community diversity overall. Bird community diversity was influenced by both the extent and fragmentation of mature forest at the landscape level. Species richness and density responded more strongly to mature forest area than to fragmentation and were significantly lower in landscapes that were completely dominated by mature forest than in landscapes with a mixture of seral stages. This reflected a selective loss of a number of early-seral species when mature forest completely blanketed the landscape. Species evenness was more strongly related to frag- mentation than to area of mature forest at the landscape level. In fragmented landscapes, the most dominant species decreased in abundance, whereas moderately abundant species increased in relative abundance. This resulted in bird communities that were more even in fragmented landscapes. Unlike several other studies, we did not find that the relative strength of fragmentation effects increased as habitat area decreased. Total bird density, in contrast, was not related to either mature forest area or fragmentation, but it varied significantly among basins. In summary, although both area and fragmentation of mature forest at the landscape level were strongly related to bird community diversity, they had qualitatively different effects. These results must be interpreted within the scope of this study, which took place in a forest-dominated landscape. Our results provide an interesting view of the factors that are related to bird diversity in a spatially complex seral mosaic but may not reflect the patterns that would exist in a high-contrast forest-nonforest landscape.

Mapping Landscape Resistance to Identify Corridors and Barriers for Elephant Movement in Southern Africa

One of Africas greatest conservation successes is the recovery of elephant (Loxodonta africana) populations within protected areas (e.g. Aleper and Moe 2006), such as those in northern Botswana. This recovery poses several challenges, however. First, habitat within protected areas is becoming degraded from high intensity elephant browsing. Second, the increasing elephant and human populations in the region have led to large increases in human—elephant conflict along the periphery of protected areas (Sitati et al. 2005; Lee and Graham 2006). Management options include facilitating natural dispersal, active relocation, and culling. Relocation is prohibitively expensive as a population-level solution given the high per capita cost. Culling is politically unpopular given Botswanas booming wildlife tourist industry. Simultaneously, large areas of the neighboring countries of Namibia, Zambia and Angola have low elephant densities. Some of these governments desire to increase elephant populations within their protected areas to promote the growth of wildlife tourism. Thus, facilitated dispersal of elephants from high density areas of northern Botswana to protected areas in other countries with low elephant densities is an attractive potential solution.

Separating the effects of environmental, spatial and disturbance factors on forest community structure in the Russian Far East

We used partial canonical correspondence analysis to decipher the relationships among abiotic environmental variables, disturbance history and spatial gradients and the composition of the forest communities in the Sikhote-alin Mountains in the Russian Far East. Through a series of partial canonical ordinations, we decomposed the variance in the forest tree community that is explainable by these three sets of factors into its seven discrete components of partial effects and interactions. We determined that the forests of the Sikhote-alin Range are structured primarily by the interaction of three things: elevation, physiographical position and disturbance history. Over the past 100 years, the disturbance regime in the study area has changed dramatically, with vast increases in the frequency of catastrophic wildfires. The effects these changes in disturbance regime have had are seen at different strengths across the altitudinal gradient. At lower elevations, the change in disturbance regime occurred very early on, and the community is now in a fire-maintained condition that is dominated by Quercus mongolica. Mature Pinus koraiensis, which formerly was co-dominant, has been nearly eliminated from the lowland forest. The increase in fire disturbance has moved in a wave across the study area and now is impacting upper-elevation primary forest as well. In the near-future, it is likely that P. koraiensis forest remnants at middle elevations will be drastically reduced and that large proportions of the upper-elevation Picea and Abies primary forest will be converted to Betula, Larix and Populus tremula.

Habitat Fragmentation Effects Depend on Complex Interactions Between Population Size and Dispersal Ability: Modeling Influences of Roads, Agriculture and Residential Development Across a Range of Life-History Characteristics

Habitat loss and fragmentation are widely believed to be the most important drivers of extinction (Leakey and Lewin 1995). The habitats in which organisms live are spatially structured at a number of scales, and these patterns interact with organism perception and behavior to drive population dynamics and community structure (Johnson et al. 1992). Anthropogenic habitat loss and fragmentation disrupts these patterns and is expected to have large, negative effects on biodiversity (Flather and Bevers 2002; Haila 2002; Fahrig 2003). The majority of theoretical studies suggest that the effect of habitat fragmentation is weak relative to the effect of habitat loss (Fahrig 1997; Henein et al. 1998; Collingham and Huntley 2000; Flather and Bevers 2002; Fahrig 2003), although some studies have predicted larger fragmentation effects (Boswell et al. 1998; Burkey 1999; Hill and Caswell 1999; Urban and Keitt 2001). In addition, some theoretical studies suggest that the effects of fragmentation per se should become apparent only at low levels of habitat amount, for example below approximately 20–30% of the landscape (Fahrig 1998; Flather and Bevers 2002), although there is little empirical evidence available to test this prediction (Fahrig 2003).