30 Sep 2003 15:53 AR AR200-ES34-18.tex AR200-ES34-18.sgm LaTeX2e(2002/01/18) P1: GCE EFFECTS OF HABITAT FRAGMENTATION 489 TABLE 1 Summary of 100 recent fragmentation studies * Biodiversity (response) variables Abundance/ Richness/ Presence/ Fitness Genetic Species Extinction/ Individual Movement/ Population Fragmentation (predictor) density diversity absence measures variability interactions turnover habitat use dispersal growth variables (35) (28) (26) (15) (12) (10) (8) (5) (4) (3) Patch size a (63) 26 21 20 11 3 7 3 3 3 3 Habitat loss/amount (60) 21 17 13 9 8 5 5 3 2 1 Patch isolation a (35) 14 7 11 2 6 3 0 0 1 0 Edge a (22) 11 5 3 2 0 4 1 2 0 1 Number of patches (10) 2 1 0 2 0 2 3 1 0 1 Structural connectivity b (8) 3 1 1 2 0 2 1 0 1 1 Matrix quality (7) 3 2 1 1 2 0 0 2 0 0 Patch shape a (4) 0 1 2 0 0 1 0 0 0 0 Qualitative only (28) 13 9 7 10 4 0 1 3 0 1 Patch scale c (42) 17 14 16 6 7 4 0 2 1 1 Landscape scale d (37) 7 7 4 4 3 3 8 2 3 1 Patch and landscape 10 6 6 5 2 3 0 1 0 1 scales (21) a Predictor variables that can be measured at either the patch scale (individually for each patch) or at the landscape scale (averaged or summed across all patches in the landscape). b Includes both connectivity studies and corridor studies. c Each data point in the analysis represents information from a single patch. d Each data point in the analysis represents information from a single landscape. * Table entries are the numbers of papers that studied the given combination of predictor (fragmentation) variable or scale and response (biodiversity) variable. Numbers in parentheses after variable names are the total number of papers (of 100) using that variable. Columns and rows do not add to 100 because each study may contain more than one fragmentation variable and more than one biodiversity variable. Annu. Rev. Ecol. Evol. Syst. 2003.34:487-515. Downloaded from arjournals.annualreviews.org by Ontario Council of Universities Libraries on 05/02/08. For personal use only.

30 Sep 2003 15:53 AR AR200-ES34-18.tex AR200-ES34-18.sgm LaTeX2e(2002/01/18) P1: GCE 490 FAHRIG CONCEPTUALIZATION AND MEASUREMENT OF HABITAT FRAGMENTATION Fragmentation as Process Habitat fragmentation is often defined as a process during which ���a large expanse of habitat is transformed into a number of smaller patches of smaller total area, isolated from each other by a matrix of habitats unlike the original��� (Wilcove et al. 1986) (Figure 1). By this definition, a landscape can be qualitatively categorized as either continuous (containing continuous habitat) or fragmented, where the fragmented landscape represents the endpoint of the process of fragmentation. Many studies of the effect of habitat fragmentation on biodiversity conform to this definition by comparing some aspect(s) of biodiversity at ���reference��� sites within a continuous landscape to the same aspect(s) of biodiversity at sites within a fragmented landscape (e.g., Bowers & Dooley 1999, Cascante et al. 2002, Diaz et al. 2000, Groppe et al. 2001, Laurance et al. 2001, Mac Nally & Brown 2001, Mahan & Yahner 1999, Morato 2001, Mossman & Waser 2001, Renjifo 1999, Walters et al. 1999). From my sample of 100 recent studies, 28% conducted such comparisons of continuous versus fragmented landscapes (Table 1). In these stud- ies, the continuous landscape represents a landscape before fragmentation (time 1 in Figure 1) and the fragmented landscape represents a landscape following fragmentation (time 2 or time 3 in Figure 1). Although this approach conforms to the definition of fragmentation as a process, it has two inherent weaknesses. First, because habitat fragmentation is a landscape- scale process (McGarigal & Cushman 2002), the sample size in such studies, for questions about the effects of habitat fragmentation on biodiversity, is typically Figure 1 The process of habitat fragmentation, where ���a large expanse of habitat is transformed into a number of smaller patches of smaller total area, isolated from each other by a matrix of habitats unlike the original��� (Wilcove et al. 1986). Black areas represent habitat and white areas represent matrix. Annu. Rev. Ecol. Evol. Syst. 2003.34:487-515. Downloaded from arjournals.annualreviews.org by Ontario Council of Universities Libraries on 05/02/08. For personal use only.

30 Sep 2003 15:53 AR AR200-ES34-18.tex AR200-ES34-18.sgm LaTeX2e(2002/01/18) P1: GCE EFFECTS OF HABITAT FRAGMENTATION 491 only two, i.e., one continuous landscape and one fragmented landscape. With such a design, inferences about the effects of fragmentation are weak. Apparent effects of fragmentation could easily be due to other differences between the landscapes. For example, Mac Nally et al. (2000) found consistent vegetation differences between fragments and reference sites and concluded that apparent effects of fragmentation on birds could be due to preexisting habitat differences between the two landscapes. Second, this characterization of habitat fragmentation is strictly qualitative, i.e., each landscape can be in only one of two states, continuous or fragmented. This design does not permit one to study the relationship between the degree of habitat fragmentation and the magnitude of the biodiversity response. Quantifying the degree of fragmentation requires measuring the pattern of habitat on the land- scape. The diversity of approaches in the fragmentation literature arises mainly from differences among researchers in how they quantify habitat fragmentation. These differences have significant implications for conclusions about the effects of fragmentation on biodiversity. Fragmentation as Pattern: Quantitative Conceptualizations The definition of habitat fragmentation above implies four effects of the process of fragmentation on habitat pattern: (a) reduction in habitat amount, (b) increase in number of habitat patches, (c) decrease in sizes of habitat patches, and (d) increase in isolation of patches. These four effects form the basis of most quantitative measures of habitat fragmentation. However, fragmentation measures vary widely some include only one effect (e.g., reduced habitat amount or reduced patch sizes), whereas others include two or three effects but not all four. Does it matter which fragmentation measure a researcher uses? The answer depends on whether the different effects of the process of fragmentation on habitat pattern have the same effects on biodiversity. If they do, we can draw general conclusions about the effects of fragmentation on biodiversity even though the different studies making up the fragmentation literature measure fragmentation in different ways. As I show in Effects of Habitat Fragmentation on Biodiversity, the different effects of the process of fragmentation on habitat pattern do not affect biodiversity in the same way. This has led to apparently contradictory conclusions about the effects of fragmentation on biodiversity. In this section, I review quanti- tative conceptualizations of habitat fragmentation. This is an important step toward reconciling these apparently contradictory results. FRAGMENTATION AS HABITAT LOSS The most obvious effect of the process of fragmentation is the removal of habitat (Figure 1). This has led many researchers to measure the degree of habitat fragmentation as simply the amount of habitat remaining on the landscape (e.g., Carlson & Hartman 2001, Fuller 2001, Golden & Crist 2000, Hargis et al. 1999, Robinson et al. 1995, Summerville & Crist 2001, Virg�� os 2001). If we can measure the level of fragmentation as the amount of habi- tat, why do we call it ���fragmentation���? Why not simply call it habitat loss? The Annu. Rev. Ecol. Evol. Syst. 2003.34:487-515. Downloaded from arjournals.annualreviews.org by Ontario Council of Universities Libraries on 05/02/08. For personal use only.

30 Sep 2003 15:53 AR AR200-ES34-18.tex AR200-ES34-18.sgm LaTeX2e(2002/01/18) P1: GCE 492 FAHRIG reason is that when ecologists think of fragmentation, the word invokes more than habitat removal: ���fragmentation . . . not only causes loss of the amount of habi- tat, but by creating small, isolated patches it also changes the properties of the remaining habitat��� (van den Berg et al. 2001). Habitat can be removed from a landscape in many different ways, resulting in many different spatial patterns (Figure 2). Do some patterns represent a higher degree of fragmentation than others, and does this have implications for biodi- versity? If the answer to either of these questions is ���no,��� then the concept of fragmentation is redundant with habitat loss. The assertion that habitat fragmen- tation means something more than habitat loss depends on the existence of effects of fragmentation on biodiversity that can be attributed to changes in the pattern of habitat that are independent of habitat loss. Therefore, many researchers define habitat fragmentation as an aspect of habitat configuration. FRAGMENTATION AS A CHANGE IN HABITAT CONFIGURATION In addition to loss of habitat, the process of habitat fragmentation results in three other effects: in- crease in number of patches, decrease in patch sizes, and increase in isolation of patches. Measures of fragmentation that go beyond simply habitat amount are generally derived from these or other strongly related measures (e.g., amount of edge). There are at least 40 such measures of fragmentation (McGarigal et al. 2002), many of which typically have strong relationships with the amount of habi- tat as well as with each other (B�� elisle et al. 2001, Boulinier et al. 2001, Drolet et al. 1999, Gustafson 1998, Haines-Young & Chopping 1996, Hargis et al. 1998, Robinson et al. 1995, Schumaker 1996, Trzcinski et al. 1999, Wickham et al. 1999) (Figure 3). The interrelationships among measures of fragmentation are not widely recog- nized in the current fragmentation literature. Most researchers do not separate the effects of habitat loss from the configurational effects of fragmentation. This leads to ambiguous conclusions regarding the effects of habitat configuration on bio- diversity (e.g., Summerville & Crist 2001, Swenson & Franklin 2000). It is also common for fragmentation studies to report individual effects of fragmentation measures without reporting the relationships among them, which again makes the results difficult to interpret. THE PATCH-SCALE PROBLEM Similar problems arise when fragmentation is mea- sured at the patch scale rather than the landscape scale. Because fragmentation is a landscape-scale process (Figure 1), fragmentation measurements are correctly made at the landscape scale (McGarigal & Cushman 2002). As pointed out by Delin & Andr�� en (1999), when a study is at the patch scale, the sample size at the landscape scale is only one, which means that landscape-scale inference is not possible (Figure 4 see Brennan et al. 2002, Tischendorf & Fahrig 2000). However, in approximately 42% of recent fragmentation studies, individual data points represent measurements on individual patches, not landscapes (Table 1). Similarly, using a different sample of the literature, McGarigal & Cushman (2002) Annu. Rev. Ecol. Evol. Syst. 2003.34:487-515. Downloaded from arjournals.annualreviews.org by Ontario Council of Universities Libraries on 05/02/08. For personal use only.

30 Sep 2003 15:53 AR AR200-ES34-18.tex AR200-ES34-18.sgm LaTeX2e(2002/01/18) P1: GCE EFFECTS OF HABITAT FRAGMENTATION 493 Figure 2 Illustration of habitat loss resulting in some, but not all, of the other three expected effects of habitat fragmentation on landscape pattern. Expected effects are (a) an increase in the number of patches, (b) a decrease in mean patch size, and (c) an increase in mean patch isolation (nearest neighbor distance). Actual changes are indicated by arrows. Annu. Rev. Ecol. Evol. Syst. 2003.34:487-515. Downloaded from arjournals.annualreviews.org by Ontario Council of Universities Libraries on 05/02/08. For personal use only.