The following is our NSF proposal. It is the entire text of the document and is quite long.

Scaling up from local experiments to large complex landscapes

Patches, corridors, and the dispersal of insects and plants

Introduction - Fragmentation of habitat is claimed by some to be "..the most serious threat to biological diversity" (Wilcox and Murphy 1985) and is the greatest threat to imperiled species (Wilcove et al. 1998). Because of this, the study of fragmentation has structured much of the fields of conservation biology and landscape ecology in the last decade (Meffe and Carroll 1998). A central tenet of both fields is that patches of habitat do not exist in isolation; connectivity among them and interactions with adjacent habitats are key to understanding the impacts of habitat fragmentation on population dynamics and biodiversity (MacArthur and Wilson 1967; Levins 1969; Harrison, et al. 1988; Pulliam 1988; Gilpin and Hanski 1991; Crome 1997; Meffe and Carroll 1998). One of the most obvious management implications of fragmentation studies is that increasing exchange rates in patchy landscapes will provide a mechanism to accomplish two conservation goals: prevent extinction and increase gene flow (Noss 1991; Meffe and Carroll 1998). The most popular strategy for increasing exchange rates has been the addition or retention of habitat corridors (Wilson and Willis 1975; Noss 1987, 1992; Mann and Plummer 1993; Mann and Plummer 1995; Meffe and Carroll 1998). The popularity of corridors is clearly demonstrated by the number of published papers that focus on them (Fig. 1). Given the exponential rise in these papers, it is time to pause and ask whether the results of previous studies are compelling enough to merit such intense attention -- do corridors really matter?

Recent studies (reviewed below) generally support the positive role of corridors on plant and animal movement and population sizes. However, in all of these studies, one necessarily confounding factor has not been investigated -- the effect of patch area on interpatch movement and population sizes. Area is a confounding factor because a patch connected by a corridor can be viewed not only as a connected patch, but also as a single, larger patch. The dynamics within a patch with a connecting corridor could be identical to the dynamics of an equal-sized patch without such a corridor (Bennett et al. 1994; Rosenberg et al. 1997; Haddad and Baum 1998). Viewed in this way, the mechanisms that would lead to higher colonization rates and population sizes in patches with corridors would be the same as the well-established effects of patch area on colonization and population size. In particular, a large patch may support larger densities of plants or animals solely due to patch area effects, such as reduced edge-effects in larger patches (Sisk, et al. 1997; Haddad and Baum 1998) and reduced emigration (Karieva 1985; Hill, et al. 1996). To further complicate this issue, the smaller patch plus its corridor may be colonized more frequently than an isolated patch due to a "drift fence" effect, whereby individuals dispersing through the landscape are intercepted by corridors and funneled into connecting patches (Haddad and Baum 1998). In short, the difference in size and shape of patches with and without corridors greatly compromises the way in which published studies have interpreted corridor effects. It is time to address this confusion.

The confounding of corridor and area effects is particularly pronounced in studies of corridor effects on animal densities. Many studies have reported increased densities in patches connected by corridors compared to those that are not (MacClintock, et al. 1977; Fahrig and Merriam 1985; Forney and Gilpin 1989; LaPolla and Barrett 1993; Dunning, et al. 1995; Norton et al. 1995; Haddad and Baum 1998). Higher densities are usually ascribed to the effects of corridors on increasing animal movement (i.e., higher colonization). Higher colonization would increase average densities in connected patches, especially if unconnected patches were unable to support populations due to colonization limitation. However, in existing studies of corridor effects on densities, animals are often not dispersal limited. In such cases, other factors, such as patch area and drift fence effects, may cause higher densities in connected patches (Haddad and Baum 1998).

Less confounded are studies that show corridors increase movement rates between patches (Henderson, et al. 1985; Zhang and Usher 1991; Beier 1993; Bennett, et al. 1994; Haas 1995; Sutcliffe and Thomas 1996; Machtans, et al. 1996; Haddad 1998a). Obtaining estimates of movement between connected and unconnected patches is difficult, so few such studies exist. In addition, most studies are correlational, and other factors that likely determine movement through the landscape, like patch size, interpatch distance, and corridor width are not controlled. We stress that there are other lines of evidence for the beneficial effects of corridors (Haddad 1998b). But, they are generally less direct and their limitations have been thoroughly discussed elsewhere (see Simberloff, et al. 1992; Rosenberg, et al. 1997).

Because of these and other deficiencies with corridor studies, recent reviews have strongly argued that the role of corridors in maintaining biodiversity has been uncritically accepted (Simberloff and Cox 1987, Nicholls and Margules 1991; Hobbs and Hopkins 1991; Simberloff et al. 1992, Mann and Plummer 1995; Rosenberg, et al. 1997). For example, there are few large-scale experimental studies (but see Bierregaard, et al. 1992; Andreassen, et al. 1996; Schmiegelow, et al. 1997; Bowne, et al. 1998; Haddad 1998a,b; Haddad and Baum 1998). Furthermore, the studies that have demonstrated that corridors have no effect (Bowne, et al. 1998; Rosenberg, et al. 1998) -- or even have negative effects (Hess 1994; Burkey 1997) -- are not given much weight. The great popularity of corridors in conservation is especially worrisome, given the magnitude and rate at which corridors are being incorporated into management plans (for recent examples, see Mann and Plummer 1995; Laurence and Gascon 1997; Meffe and Carroll 1998). It is time for a rigorous test of corridor effects, one that controls for confounding factors.

This study is specifically designed to tease apart corridor and area effects. We will take advantage of a unique opportunity to create a large, replicated experiment that is composed of open patches and corridors within large areas of mature pine forest. Within eight experimental units, we will test four key hypotheses about the effects of corridors on movement and population sizes (H₁ - H₄; Table 1). We will then test the relevance of our results for predicting abundances of plants and animals at larger, regional scales (Hypothesis 5). Specifically, we will use the results from our experiment to parameterize a model that predicts the distributions and abundances of focal species within our study area, the 80,000 ha Savannah River Site (SRS). This area is well-suited for such an approach because it consists of hundreds of open areas (created by clearcut forestry) and corridors (created by roads and utility rights-of-way). We also have access to the site's GIS database of current and historical landscape patterns. To extend the generality of our results, we will study two very different taxa, butterflies and bird-dispersed seeds, in testing the five hypotheses (Table 1).

Table 1. Hypotheses about the Effects of Corridors

Hypotheses 1-4 address the importance of corridors on movement and population sizes, rather than on within-patch demographics. Then, Hypothesis 5 takes a meta-population approach by exploring the consequences of various assumptions about birth and death rates on population sizes in our landscapes. However, note our belief (justified below) that in our system of ephemeral habitats, colonization -- not persistence -- is key. The more individuals that can colonize a patch, the better the chances of establishment, the higher the population densities that are likely to result, and the higher the production of dispersing individuals to establish new populations elsewhere. Because we are studying open-habitat species (see justification, below), we focus on the behavior of dispersing individuals (or their host carriers in the case of bird-dispersed seeds), rather than on the effects of dispersal on within-patch demographics.

Site and Study Species -- Savannah River Site (SRS) is located near Aiken, SC, and its vegetation is described in Workman and McLeod (1990). Upland pine forest is managed for harvest by the Savannah River Natural Resource Institute (SRNRI), which is operated by the US Forest Service. The unique mission of SRNRI focuses on the maintenance of biodiversity in the context of managed forests. We will be working in forests of mature, mostly planted pines. Species include Pinus taeda, P. palustris, and P. elliottii. The entire site has been closed to the public since the 1950's, providing a secure location for research. Furthermore, SRNRI has promised to pay most of the costs associated with creating our experimental landscape (Appendix A), and they have demonstrated their ability to carry through on this promise by doing similar harvests for us in 1994-5 (Haddad 1998a,b; Haddad and Baum 1998; Bowne, et al. 1998). In sum, we have the opportunity to conduct large forest manipulations at a speed, flexibility, and scale that we believe cannot be matched elsewhere. (We note that previous large scale experiments on corridor effects have been more constrained in their placement of patches and corridors. For example, patch geometry is rarely held constant and corridors are often associated with streams.)

Butterflies and bird-dispersed seeds may seem like a strange combination of focal taxa. We focus on them for five reasons. (1) Within similar experimental sites that tested corridor and distance effects on interpatch movement (see below), we and others have studied several taxa, including five butterfly species, four bird-dispersed plant species, two small mammal species, bees, bee-pollinated plants, and two lizard species. Not all of these species responded positively to corridors, but butterflies and bird-dispersed seeds did. One of our primary objectives is to separate corridor from area effects on species that we know show positive responses to corridors. (2) These taxa are our areas of expertise and we have demonstrated our ability to release, mark, and recapture the butterflies and seeds that we propose to study (details provided below). Previous research by us and others at the SRS provides a solid foundation on the life histories and distributions of birds (e.g., unning, et al. 1995), butterflies (Haddad 1998a,b; Haddad and Baum 1998), and plants (e.g., Workman and McLoed 1990; Sargent and Levey, in preparation). (3) This work has been done as part of a large-scale, collaborative pilot study (see below). We would like to build upon our previous collaboration. (4) To justify the enormous effort needed to create our system of patches and corridors, we feel it is important to study responses of more than one taxon. In this respect, the more disparate the taxa, the better. This leads to the fifth reason: (5) Different species are likely to respond differently to the habitat patterns we will create. Understanding this variation is biologically important and contrasts with the more typical situation in which researchers focus on one species or taxon.

The two butterfly species that we will study, Buckeye (Junonia coenia) and Variegated Fritillary (Euptoieta claudia), strongly prefer open habitat (Scott 1975a, 1975b, 1984; Haddad 1998a; Haddad and Baum 1998). Both are likely to remain within the area of a patch, but both are also capable of long-distance, interpatch movements (Haddad 1998a). We have demonstrated our ability to mark, release, and recapture thousands of individuals (Haddad 1998a). In addition to our measures of interpatch movement and density, we will measure abundances of all butterfly host and nectar plants in the patches and corridors. Host plants include Linaria canadensis, Plantago spp., and Gerardia purpurea for J. coenia, and Passiflora incarnata and Viola sp. for E. claudia. The butterflies are generalists for nectar resources. Nectar plants include L. canadensis, Rubus sp., and Asclepias tuberosa. Seed dispersal of four species of plants will be studied. Winged Sumac (Rhus copallina), American Pokeweed (Phytolacca americana), Flowering Dogwood (Cornus florida), and Wax Myrtle (Myrica cerifera). These species were chosen because they are abundant in or alongside young clearcuts, produce numerous fruits that are rapidly eaten by birds, and fruit at different times of year (Levey and Sargent, unpublished data). Although two species, Cornus and Myrica, do not produce fruit in young clearcuts, they readily colonize such habitats and are consumed by open-habitat birds along the edges of clearcuts. Thus, their seed dispersal to clearcuts by open-habitat birds is highly relevant.

Supporting Results

Our proposal builds on previous studies that tested corridor and distance effects on interpatch movements and densities (Haddad 1998a,b; Haddad and Baum 1998; Bowne et al. 1998; Sargent and Levey, unpublished data). Here, we highlight key findings of our previous work (which, incidentally, demonstrate our ability to carry out large scale experiments), justify our choice of study species, and show that we have already estimated important movement parameters, including butterfly and seed responses to corridors and interpatch distance, that are critical for the model proposed below.

1. Corridors increase movement of butterflies and bird-dispersed seeds between patches of habitat that they connect, relative to similar, unconnected patches. Two open-habitat butterfly species and two seed species move with higher frequencies between patches connected by corridors than between similar, unconnected patches. Mark-release-recapture studies indicate that the proportion of both Eurema claudia and Junonia coenia moving between connected patches is significantly higher than the proportion moving between unconnected patches (Fig. 2; Haddad 1998a). Likewise, one species of seed, Rhus copallina, was dispersed significantly more often between connected than between unconnected patches (Fig. 3; Sargent and Levey, unpublished. data). A second species, Phytolacca americana, showed a similar pattern.

2. Corridors have greater effects at larger interpatch distances. When patches are close together (64 m), movement rates between patches are relatively high, even between unconnected patches (Figs. 2 and 3). As patch separation and corridor length increase, movement rates between patches generally decrease and the difference in movement rates between connected and unconnected patches generally becomes more pronounced (Figs. 2 and 3).

3. Corridors increase densities of habitat-restricted species. Both seeds and butterflies reached higher densities in patches connected by corridors than in similar, isolated patches (Haddad and Baum 1998; Sargent and Levey, unpublished.). While higher densities may have been due to higher movement rates caused by corridors, two other hypotheses could not be rejected: 1) butterflies and seeds reached higher densities further from the forest edge (for Rhus seeds, density under perches was 4 to 10 times higher near the center of patches than near the edge; for both butterfly species, densities were more than 10 times higher at the center of the patch (Haddad and Baum 1998)); corridors caused more of the patch to be further from the edge, which increased densities, or 2) corridors of connected patches acted as "drift fences," capturing individuals dispersing through the landscapes and directing them toward connected patches. The inability to separate these factors highlights the need for studies that tease apart corridor and area effects.

4. Corridor use can be predicted from movement behaviors. Behavioral observations of butterfly movement paths indicate that two open-habitat species (Phoebis sennae and Eurema nicippe) leave through corridors at up to 4 times the rates expected by random movement, while a habitat generalist (Papilio troilus) leaves through the corridor as often as expected if dispersal direction was random (Fig. 4; Haddad 1998b). Simulation models, parameterized with actual movement behaviors, predicted that the two open-habitat specialists are more likely to colonize a patch connected by a corridor than one that is unconnected (Haddad 1998b).

5. Connectivity influences population abundances and distributions. Anderson and Danielson (1997) have developed a model that is structurally similar to the one proposed in this study. Their model confirmed that corridors can increase the longevity and size of a metapopulation by ensuring that all the available habitat in a landscape is colonized. However, the model also demonstrated that the positive effects of corridors can be offset by a negative effect: well-connected patches may experience increased rates of immigration at the expense of poorly connected patches, resulting in colonization limitation. This causes the net effect of corridors to be negative for the metapopulation as a whole.

Proposed Research

Design of experimental sites

Our proposed experiment is designed around open-habitat patches and corridors surrounded by large areas of pine forest. We note that our experimental design creates the inverse of what people typically consider a corridor -- a strip of forest connecting blocks of forest. Instead, we have corridors of clearcut connecting patches of clearcuts. However, we believe that the key element of the design is the contrast between preferred (in this case, open) and less preferred habitats. The open habitats at our site, which support diverse and productive herbaceous vegetation, contrast strongly with the surrounding forest, which is densely planted in pine and prevents the development of an herbaceous layer. Many plant and animal species, including the butterflies, plants, and birds in this study, respond strongly to the contrast between the two habitats (Haddad 1998a,b; Haddad and Baum 1998; Sargent and Levey, unpublished. data). Additionally, there are rare or declining species in open habitats that might derive some benefit in management for the specific landscape characteristics we examine. This holds true both at our site (e.g., Bachman's Sparrow; Aimophila aestivalis; Dunning, et al. 1995) and more generally for many species in native prairies and grasslands.

An "experimental unit" contains five patches: one "source" patch (where butterflies and seeds are marked and released) in the center of four "receiver" patches, each separated from the source by the same distance (150 m; Fig. 5). The source patch is 100 x 100 m (1 ha). Patch and corridor areas were chosen to fall within the range of typical Forest Service management and to avoid shading in the core of the patch. In our design, we manipulate two key factors. First, the source patch in each experimental unit will be connected by a corridor to one receiver patch. The other three receiver patches will be unconnected. The corridor will be 150 m long, 25 m wide. This dimension was chosen based on our previous work (which demonstrated that 128 m was the smallest distance for which we expect a strong effect; Fig. 2, 3) and because a 25 m width prevents the entire corridor from being shaded. Second, we will vary the dimensions of outer patches to test hypotheses about area, edge, and drift fence effects on butterfly and plant movement and densities. The receiver patch connected by the corridor (henceforth "connected receiver") will be equal in size (1 ha) to the source patch. The three other receiver patches will be isolated and have an area equal to the connected receiver plus the corridor. The area of the remaining three patches will be increased in one of two ways. "Rectangular receiver" patches will have an area equivalent to the corridor added to the side of the patch farthest from the source patch (i.e., their cross-sectional area when viewed from the source patch will be equivalent to that of the connected receiver patch). "Winged receiver" patches will have two "blind corridors" (each 75 m, or half the interpatch distance) projecting off opposite sides, parallel to the nearest edge of the source patch. (A single, full-length corridor is not proposed as a wing because of problems with symmetry, relative to the source patch.) The remaining (4th) receiver patch will be either a rectangular or a winged receiver. This fourth patch permits additional replication and increases efficiency of our design by allowing us to capture dispersers in all directions from the source patch. Thus, the experimental units will be equal in every way, except that half will have two rectangular patches, and half will have two blind corridor patches. The relative position of receiver patches around the source patch will be determined randomly.

We will create a total of eight experimental units (n = 8, 12, 12 connected, winged, and rectangular receiver patches, respectively). If Junonia coenia behaves in the same way it did in Haddad's (1998) study at our site, this sample size will yield a statistical power of 0.96 (assuming a one-sided alpha = 0.05 and beta = 0.80; Cohen 1969). In other words, if the difference between means is at least 0.96 standard deviations, our probability of a Type II error will be < 0.20. An effect size of this magnitude are reasonable to expect, given corridor effects with butterflies and seeds in our previous design. In fact, given the effect sizes we observed (Fig. 2 and 3), we could justify a much smaller sample size (n=4 or 5). We think this would be unwise, because area and drift fence effects may be more subtle than corridor effects. Finally, note that if we increase alpha to 0.10, as is common in other large-scale experiments (Schmiegelow, et al. 1997), the necessary effect size is reduced.

A grid system will be established to enable rapid assessment of spatial location within patches. Within a 12.5 m buffer in each patch, a 3 x 3 grid of 25 m x 25 m square cells will be demarcated with a laser transit (in the rectangular patch, a 4 x 4 grid will be created, to standardize for the larger area). At each point of intersection on the grid (n = 16), a 3.3 m PVC pole will be erected. Poles will also be erected at 25 m intervals in each corridor and blind corridor. This system, supplemented by additional poles and flags, will allow location to be determined to a resolution of 6 m from any point within the patch. Grid poles also serve as experimental perch locations for birds (see below).

Level #1: Corridor effects on movement

Hypothesis 1: After controlling for patch area, movement of butterflies and seeds is greater between patches connected by a corridor of similar habitat than between unconnected patches. That corridors increase movement between the habitat areas they connect is a frequently assumed but poorly tested idea in landscape ecology (Wilson and Willis 1975; Noss 1987; Simberloff and Cox 1987; Simberloff, et al. 1992; Rosenberg, et al. 1997). By creating experimental units that contain connected and unconnected patches, we can directly test the effect of corridors on interpatch movement rates. Our study will be the first replicated, experimental test of this that controls for area. We will separate corridor from area effects by comparing movement rates between the source patches and connected patches to movement between the source patches and the rectangular patches.

Hypothesis 2: Corridors act as "drift fences," intercepting dispersers from the surrounding habitat and directing them to habitat patches. For vagile species, stepping stone movement from patch to patch through unsuitable habitat may be as important as movement through corridors. If this is the case, corridors may act to divert dispersers from the surrounding landscape into patches (Haddad and Baum 1998). Put another way, corridors may intercept open-habitat species dispersing through forest habitat and direct them into patches of open habitat, in a way that is analogous to the functioning of a drift fence. To test this hypothesis, we will compare movement rates into winged receiver patches with movement rates into rectangular receiver patches. These patches are equal in area, but winged receiver patches subtend over two times the angle in the horizon, from the source patches, than is subtended by rectangular patches. We predict that the rate of colonization of winged receiver patches from source patches will be greater than colonization of rectangular receiver patches.

Methods: Butterflies. We will transplant marked butterflies into each source patch, and sample receiver patches to recapture marked individuals. Within each source patch, 30 Junonia coenia and 10 Euptoieta claudia will be released each week for 8 consecutive weeks in May to June, 2000 and 2001 (for a total of 240 and 80 butterflies per patch. The difference in release rate between species reflects differences in actual densities within patches (Haddad and Baum 1998)). Release months correspond to the spring peak in adult abundances of our focal species (Haddad 1998a). Given the results of our previous studies (Haddad 1998a), we anticipate a total recapture rate of 25%. All released butterflies will be marked with a unique code using a felt-tip pen (Ehrlich and Davidson 1961). For one week after release, all receiver patches (4 patches/experimental unit X 8 experimental units = 32 patches) will be sampled daily for butterflies. Average adult life span of the two species is approximately one week (Scott 1975 a,b); recaptures after that time are extremely rare (Haddad 1998a). We will sample receiver patches and corridors by walking parallel transects separated by 12.5 m. [This assumes a 6 m detection distance, which we have shown to be adequate for sampling most butterflies (Haddad 1998a,b).] We will also sample corridors, to document that marked butterflies actually use them for dispersal (see Simberloff et al. 1992). Transects will be sampled at the rate of 5 min for every 100 m of transect, not including time spent capturing and marking butterflies. Locations of recaptured individuals will be recorded within 6 m of capture. Because host plant density and flower density are likely to influence butterfly abundance independent of corridor presence/absence, we will record host and flower density during separate surveys in each 6 m grid cell. Data on plant resources will be used as covariates in statistical analyses (see below). Butterflies for release will be transplanted from elsewhere on the SRS. We have demonstrated our ability to catch, mark, and release wild-caught individuals in the quantities we propose (Haddad 1998a). In addition, we have demonstrated that transplanted butterflies do not differ from naturally occurring butterflies in their use of corridors (Haddad 1998a). We will control for age and sex of released butterflies.

Seeds: We will mark fruits of our four focal species in source patches and look for recovery of their seeds in all adjacent receiver patches. Fruits will be marked with flourescent microspheres (15 micron diameter, Bangs Laboratories, IL), which we will apply in a solution of Wilt-Pruf (a non-toxic, biodegradable carrier) using an airbrush. In a series of lab and field validation trials, we have demonstrated that the microspheres: (1) adhere to the fruits for several months in the field, (2) do not undergo photodegradation in full sunlight, (3) do not affect fruit choice or consumption by frugivorous birds, (4) have the same mean retention time in bird guts as seeds, (5) are not digested by birds, and (6) can be detected with an epiflourescent microscope in fecal samples. Full details of these trials are provided in a manuscript (submitted to Ecology) available at http://www.zoo.ufl.edu/DLEVEY/microspheres.

We will mark at least 2,000 Cornus, 5,000 Phytolacca, 100,000 Myrica, and 150,000 Rhus fruits in each season. These numbers, which vary because the four fruit species vary tremendously in their probability of removal, are sufficient for expected captures of at least 100 seeds in receiver patches (Sargent and Levey, unpublished. data; and see ms at above web site). Seeds will be captured in seed traps, placed at each of the marker poles. Most birds defecate exclusively from perches (Robinson and Handel 1993; McClanahan and Wolfe 1993). Because different species of frugivorous birds use different types of perches, two types of traps will be used: (1) For birds that perch low in vegetation, we will place rectangular traps on the ground below horizontal dowel rods that protrude from the sides of each marker pole, approximately 1.5 m off the ground. Screen on these traps prevents seed removal by rodents. We have documented that seed removal by ants is negligible. (2) For birds that perch high above the vegetation, we will wrap circular traps around the poles, and suspend them from the pole tops. On the top of each pole we will place a short horizontal perch. We have used both designs with success. We will place 25 traps (16 associated with marker poles and 9 additional) in each of the 32 receiver patches. In addition, we will place 6 traps in each of the 32 corridors or corridor equivalents (i.e., the "extra" area in rectangular patches) to monitor use of those areas. Thus, seed traps will total (25 x 32) + (6 x 32) = 992. We will check traps at least twice per week during summer and winter seasons (10 wks each), when we will mark fruit. Seeds and fecal material will be preserved in 70% ethanol. Presence of a microsphere will be equated with seed dispersal by a bird that ate marked fruit in the source patch. Detection of numerous spheres in a trap from a single collection period will be counted as a single dispersal event (i.e., independent occurrence).

Our summer fruit species will be Phytolacca and our winter fruit species will be Rhus, Myrica, and Cornus. In some years, all Myrica and Cornus fruits are removed as soon as they ripen in the fall. To guard against this possibility and ensure we will have fruits to use in the winter, we will place netting over Myrica and Cornus fruits in the fall. For these two species, we will likely place branches containing fruits in the source patches to attain an adequate sample of marked fruits. We have had success with this technique.

Because fruit-eating bird movements are likely to be influenced by fruit abundance as well as by corridors, density of fruit will be recorded in each receiver patch and used as a covariate in analyses. Methodology will parallel that used with butterflies for controlling for host plant and flower density.

Statistical Analysis: We will record the number of individuals recaptured in each of the receiver patches. We describe only the analyses and interpretation for butterflies, since the methodology for seeds will be identical (except the dependent variable will be arcsine-transformed proportions of marked seeds recovered in receiver patches). Since the number of marked butterflies in each source patch will be equal, the response variable will be the number of marked butterflies reaching each of the receiver patches. We will use an ANCOVA to analyze movements, with the type of receiver patch as a main effect, and the abundance of host and nectar plants as covariates. Hypothesis 1 will be supported if recovery of marked butterflies is higher in the connected receiver patches than in the rectangular receiver patches. Hypothesis 2 will be supported if recovery is higher in the winged receiver patches than in the rectangular receiver patches.

Level #2: Corridor effects on population size

Hypothesis 3: After controlling for patch area, the presence of corridors will increase population densities in patches connected by corridors. Our design will allow us to separate corridor from area effects on butterfly and seed densities. The positive effects of corridors on densities represent some of the most compelling evidence that corridors function to increase exchange rates (MacClintock, et al. 1977; Fahrig and Merriam 1985; LaPolla and Barrett 1993; Dunning, et al. 1995; Haddad and Baum 1998). However, in every case, patch area effects may also explain increased densities (Haddad and Baum 1998). Larger patches may support higher densities for at least two reasons: 1) larger patches have lower edge:area ratios, reducing emigration (Kareiva 1985; Hill, et al. 1996), and 2) larger patches have a greater area of interior habitat, which may be preferred by many habitat restricted species (Sisk, et al. 1997; Haddad and Baum 1998). The first explanation is relevant for butterflies, but not for seeds. The second applies because we found seed rain to be highest in the center of patches (see above). We assume that the more seeds arriving in a patch, the higher the probability of establishment and, eventually, the higher the population size of adult plants. By controlling for area, we will separate corridor from area effects on densities.

Hypothesis 4: For habitat restricted species, patches with lower edge:area ratios have higher densities than patches of equal area with more edge. In addition to separating area from corridor effects, our experiment allows us to determine the contribution of edge effects to changes in density. As described under hypothesis 3, edge effects may impact the area of available habitat and the rate of emigration. The isolated rectangular patch has a lower edge:area ratio than the isolated patch with a corridor. Thus we predict that the butterflies and seeds in this study will reach higher densities in the rectangular patch than in the blind corridor patch, solely due to edge effects.

Methods: Butterflies: We will estimate densities of Junonia coenia and Euptoieta claudia in each receiver patch. Patches will be sampled as described for H₁ and H₂ above. Naturally occurring and transplanted individuals of each species will be captured and, if unmarked, marked with a unique code to prevent double counting. The location of each butterfly will be recorded at a resolution of 8 m. Seeds: Note that the seed trap data used to test H₁ and H₂ already provide a density estimate of seeds, because the trap area is known. The data are biased towards seed densities under perches, but because bird-dispersed seed rain elsewhere is extremely low (Robinson and Handel 1993; McClanahan and Wolfe 1993), this bias is biologically justified.

Statistical Analysis: The response variables will be the number of butterflies or seeds in each of the receiver patches. Because count data of both butterflies (Haddad and Baum 1998) and seeds (Sargent and Levey, unpublished data) follow a Poisson distribution, we will use Poisson regression to analyze the significance of patch area, shape, and connection on butterfly densities in each patch (contained within the Generalized Linear Model procedure in SAS (McCullagh and Nelder 1989; SAS Institute, Inc. 1996)). As with the analysis of movement, the abundances of fruits, or host and nectar plants, will be used as covariates in analyses. Hypotheses 3 will be supported if densities are higher in the connected receiver patches than in either winged or rectangular receiver patches. Hypothesis 4 will be supported if densities are higher in the rectangular patches than in the winged patches.

Level #3: Scaling-up: Models and Field Validation at Large-Scales

Hypothesis 5: Effects of corridors on movement patterns of butterflies and seeds will influence large-scale distributions and abundances of butterflies and seeds. We propose to test the implications of our experimental results at the scale of large, regional metapopulations. Our experimental sites are representative of actual landscapes at the SRS, where forest clearings are created through pine harvests, and cleared stands vary in their degrees of isolation and connectivity (Fig. 6). We will develop and test a model that will "scale-up" our experimental results to larger patches and corridors in a managed forest. The model will predict the relative abundance of butterflies and plants within cleared stands, based on such factors as the distance between stands, connectivity to other stands, and patch area. Degree of connectivity will be assessed by presence of power line right of ways and roads, which are thought to serve as dispersal corridors between clearcuts in this dynamic landscape (Fig. 6; Dunning, et al. 1995). We have already obtained estimates of several parameters for the model, including distance and corridor effects on interpatch movement rates (see Supporting Results above). But other parameters are needed -- in particular those that predict drift fence and area effects on interpatch movement and population density. We need such data because they will allow us to disentangle and incorporate corridor, drift fence, and area effects in the model and field tests.

Figure 6. Two representative 10,000 ha areas of the SRS showing clearcuts (dark) that are 1-6 yrs old. Lines between patches represent potential corridors (road, rail, power, and pipeline rights-of-way). Note that the areas differ substantially in the number and spacing of corridors and clearcuts.

Modeling Plant and Animal Abundances. -- We will model butterfly and seed metapopulations across twenty 4,000 ha subsections of the 80,000 ha SRS. We will determine colonization probabilities of open patches at the scale of the managed forest landscape using field results from this proposal (i.e., H₁-H₄ above), as well as previous results demonstrating the effects of corridors and distance on interpatch movements of the butterflies and plants in this study (Haddad 1998a,b; Sargent and Levey, unpublished data; see Supporting Results above). Models will be validated by field estimates of abundances and distributions of butterflies and plants within the same habitats that are modeled in the simulation.

We re-emphasize that the clearings created by clearcut forestry are ephemeral, and provide suitable habitat to our study species for only 6-7 years (at which time pine regeneration shades out other plant species). Because of this, colonization limitation, rather than within patch demography, is probably the dominant factor driving metapopulations in these landscapes -- a conclusion that has been reached by other researchers at the SRS (Pulliam, et al. 1992; Dunning, et al. 1995). Because dispersal limitation predominates in these landscapes, the details of within patch demography (within reasonable limits) are likely to have no qualitative influence on model predictions. However, we will conduct a rigorous test of the sensitivity of our model to this assumption.

The model involves two stages: 1) the computation of movement probabilities between all pairs of patches, and 2) simulations of metapopulation dynamics. We will first compute interpatch movement probabilities between all patch pairs within 4,000 ha regions of the SRS. We will identify all open patches and potential corridors on GIS-based images of the SRS (provided by the SRNRI). Using our experimental results, we will then calculate, for each butterfly and plant species, the probability of moving between each pair of patches. This probability will be based on our own data, and will include interpatch distance and connectivity, as well as the receiver patch area, and the distance to intervening corridors (which may act as drift fences). Even at distances for which we have not specifically estimated interpatch movements, we can predict rates based on estimates generated from our data. To complete calculations of interpatch movement probabilities, however, we need to estimate how variation in other, major attributes of the SRS landscape -- namely, area and drift fence effects -- influence interpatch movement probabilities. This is where data collected for H1-H4 will be essential to the modeling. We will array movement probabilities in a "transition matrix."

After computing movement probabilities, we will simulate metapopulation dynamics in a simple model. The initial within patch population size will be estimated using density estimates from our experimental patches, as well as the effective area model (sensu Sisk, et al. 1997), which accounts for the diminishing effects of edge in larger patches. Because we do not estimate birth and death rates as part of this study, we will simulate a large range of rates that are biologically justified for these species, to determine the sensitivity of our model results to assumptions about within-patch demography. Between each reproductive period, we will move butterflies and seeds between patches in proportions specified by the transition matrix. Butterflies generally move further than bird-dispersed seeds, requiring that the plant model be simulated over many generations. Thus the plant model will be run on dynamic landscapes incorporating the senescence of old and the creation of new clearcuts. This is readily done by using the current age structure of forest stands and knowledge of recent harvest rotations to construct historical landscapes. A somewhat similar method of modeling future landscapes for the SRS has been successful (Pulliam, et. al. 1992).

Field Methods: In May and June, 1999, we will census our study species in each of 10 randomly-chosen clearcuts in each of the twenty 4,000 ha regions (20 regions x 10 clearcuts/region = 200 total clearcuts) for our study species. We will census an entire clearcut along transects with lengths determined by the area and shape of the patch. We will standardize the census rate at 15 ha/hr. A 15 ha area would be divided into transects separated by equal distances that extend a total of 1,000 m. The total number of Junonia coenia and Euptoieta claudia butterflies will be counted. Seedling and established plants of Rhus copallina, Phytolacca americana, Cornus florida, Myrica cerifera will be counted along 20, 10 x 1 m transects in each clearcut (4,000 transects, total). Transect locations within clearcuts will be determined at random.

Matching Model Predictions and Field Abundances: Model output will consist of butterfly and plant abundances in each clearcut; however, generating reliable point-to-point comparisons of abundances is not an objective of this study. Doing so would require much more detailed demographic and behavioral information than we propose in the experiment. Instead, we will use model output to rank each of the 10 randomly selected clearcuts (patches) in each region by order of their relative abundances of butterflies and plants. We will rank the same patches by order of their relative abundances as determined by field surveys. Only if the behaviors that move seeds and butterflies are important at this larger (4,000 ha) scale can we reasonably expect strong qualitative correspondence between these ranked data.

While we expect to find good within-region correspondence between the model and each of the censused 4,000 ha regions of the SRS, other unknown environmental variables may become increasingly important at larger spatial scales. Thus, we will determine the model's ability to predict abundances among 4,000 ha regions by comparing model and field estimates of abundances averaged from the 10 random selected clearcuts. If interpatch movement is important at the scale of the entire 80,000 ha SRS, then there should also be strong correspondence between these regional averages computed from the model and from the field censuses. We will compare expected rank orders of abundance with observed rank orders of abundance with Spearman's rank correlations.

Significance of Proposal

The major contribution of our study lies in its integrative and experimental approach, using a model to extend a large-scale experiment (relative to most previous studies) to even larger-scale patterns. The experiment will document whether corridors increase interpatch movements and population sizes, and will allow us to avoid the pitfalls and assumptions associated with correlative studies (Saunders and Hobbs 1991; Simberloff et al. 1992). In particular, we will tease apart corridor and area effects on butterfly and plant movement and population sizes. The modeling component will allow us to move beyond the traditional focus on processes within patches and corridors to one that acknowledges the role of larger elements of the landscape in patterns of species abundances (Saunders et al. 1991; Anderson and Danielson 1997). The uniqueness and need for our approach are explained in the Introduction; we won't reiterate them further. Instead, we would like to note two additional features of our project.

First, the effort associated with implementing our design is so great and the opportunity so unique that we will encourage participation by other researchers. In particular, we aim to forge strong collaborative ties with faculty at the Savannah River Ecology Lab, the US Forest Service, and the University of Georgia (this list is not restrictive). We also will apply for NSF REU supplements to involve undergraduates in the project. (Several of D. Levey's REU students have published and received university-wide awards)

Second, conservation biology is sometimes criticized for lack of rigor (Norton 1988; Murphy 1990; Meffe and Carroll 1998). Yet, the experiments most urgently needed -- those at large-scales -- are precisely those most difficult to conduct. Compromises are necessary, and frustration, controversy, and confusion result (Simberloff and Cox 1987; Noss 1987; Nicholls and Margules 1991; Inglis and Underwood 1992; Simberloff et al. 1992). An obvious, yet rarely pursued way to alleviate this problem is for scientists to work closely with land managers, who regularly alter landscapes on large scales (Saunders et al. 1991; Machtans et al. 1996; Schmiegelow, et al. 1997). Not only do we have such an opportunity, but we have strong encouragement from the US Forest Service and Department of Energy to take advantage of it (see letter by J. Blake; appendix A). Furthermore, we have enormous support from the Forest Service; they have agreed to cover most costs associated with creating the patches and to restrict use of the land until we are finished. We believe this collaborative effort, permitting experimental research at spatial scales relevant to forest management, and incorporating area and drift fence effects, will provide the most comprehensive assessment to date of the conservation value of corridors.

Literature Cited

Anderson, G.S., and B.J. Danielson. 1997. The effects of landscape composition and physiognomy on metapopulation size: the role of corridors. Landscape Ecology, in press.

Andreassen, H.P., S. Halle, and R.A. Ims. 1996. Optimal width of movement corridors for root voles: not too narrow and not too wide. Journal of Applied Ecology 33:63-70.

Beier, P. 1993. Determining the minimum habitat areas and habitat corridors for cougars. Conservation Biology 7:94-108.

Bennett, A.F., K. Henein, and G. Merriam. 1994. Corridor use and the elements of corridor quality: chipmunks and fencerows in a farmland mosaic. Biological Conservation 68:155-165.

Bierregaard, R.O., T.E. Lovejoy, V. Kapos, A.A. Santos, and R.W. Hutchings. 1992. The biological dynamics of rainforest fragments. Bioscience 42:859-866.

Bowne, D.R., J.D. Peles, and G.W. Barrett. 1998. Effects of landscape spatial structure on movement patterns of the hispid cotton rat (Sigmodon hispidus). Landscape Ecology, in press.

Burkey, T.V. 1997. Metapopulation extinction in fragmented landscapes: using bacteria and protozoa communities as model ecosystems. The American Naturalist 150:568-591.

Chrome, F.H.J. 1997. Researching tropical forest fragmentation: shall we keep on doing what we're doing? In W.F. Laurance and R.O. Bierregaard, editors. Tropical forest remnants: ecology, management, and conservation of fragmented communities. Chicago University Press, Chicago, IL.

Cohen, J. 1969. Statistical Power Analysis for the Behavioral Sciences. Academic Press. New York, NY.

Dunning, J.B., R. Borgella, K. Clements, and G.K. Meffe. 1995. Patch isolation, corridor effects, and colonization by a resident sparrow in a managed pine woodland. Conservation Biology 9:542-550.

Ehrlich, P.R. and S.E. Davidson. 1960. Techniques for capture-recapture studies of Lepidoptera populations. Journal of the Lepidopterists' Society 14:227-229.

Fahrig, L. and G. Merriam. 1985. Habitat patch connectivity and population survival. Ecology 66:1762-1768.

Forney, K.A. and M.E. Gilpin. 1989. Spatial structure and population extinction: a study with Drosophila flies. Conservation Biology 3:45-51.

Gilpin, M.E. and I. Hanski, editors. 1991. Metapopulation dynamics. Academic Press, London, England.

Haas, C.A. 1995. Dispersal and use of corridors by birds in wooded patches on an agricultural landscape. Conservation Biology 9:845-854.

Haddad, N.M. 1998a. Corridor and distance effects on interpatch movements: a landscape experiment with butterflies. Ecological Applications, in press.

Haddad, N.M. 1998b. Corridor use predicted by behaviors at habitat boundaries. The American Naturalist, in review (resubmission invited).

Haddad, N.M. and K. Baum. 1998. An experimental test of corridor effects on butterfly densities. Ecological Applications, in press.

Harrison, S, D.D. Murphy, and P.R. Ehrlich. 1988. Distribution of the Bay Checkerspot butterfly, Euphydryas editha bayensis: evidence for a metapopulation model. American Naturalist 132:360-382.

Henderson, M.T., G. Merriam, and J. Wegner. 1985. Patchy environments and species survival: chipmunks in an agricultural mosaic. Biological Conservation 31:95-105.

Hess, G.R. 1994. Conservation corridors and contagious disease: a cautionary note. Conservation Biology 8:256-262.

Hill, J.K., C.D. Thomas, and O.T. Lewis. 1996. Effects of habitat patch size and isolation on dispersal by Hesperia comma butterflies: implications for metapopulation structure. Journal of Animal Ecology 65:725-735.

Hobbs, R. J. and A. J. M. Hopkins. 1991. The role of conservation corridors in a changing climate. Pages 281-290 in D. A. Saunders and R. J. Hobbs, eds. The role of corridors. Chipping Norton, Surrey Beatty, New South Wales, Australia.

Inglis, G. and A.J. Underwood. 1992. Comments on some designs proposed for experiments on the biological importance of corridors. Conservation Biology 6:581-586.

Kareiva, P.M. 1985. Finding and losing host plants by Phyllotreta: patch size and surrounding habitat. Ecology 66:1809-1816. Kemp, J.C. and G.W. Barrett. 1989. Spatial patterning: impact of uncultivated corridors on arthropod populations within soybean agroecosystems. Ecology 70:114-128.

LaPolla, V.N., and G.W. Barrett. 1993. Effects of corridor width and presence on the population dynamics of the meadow vole (Microtus pennsylvanicus). Landscape Ecology 8:25-37.

Laurance, W. F. and C. Gascon. 1997. How to creatively fragment a landscape. Conservation Biology 11:577-579.

Levins, R. 1969. Some genetic and demographic consequences of environmental heterogeneity for biological control. Bulletin of the Entomological Society of America 15:237-240.

MacArthur, R.H. and E.O. Wilson. 1967. The theory of island biogeography. Princeton University Press, Princeton, New Jersey.

MacClintock, L., R.F. Whitcomb, B.L. Whitcomb. 1977. Island biogeography and the "habitat

islands" of eastern forest. II. Evidence for the value of corridors and minimization of

isolation in preservation of biotic diversity. American Birds 31:6-12.

Machtans, C.S., M. Villard, and S.J. Hannon. 1996. Use of riparian buffer strips as movement corridors. Conservation Biology 10:1366-1379.

Mann, C.C., and M.L. Plummer. 1993. The high cost of biodiversity. Science 260:1868-1871.

Mann, C.C., and M.L. Plummer. 1995. Are wildlife corridors the right path? Science 270:1428-1430.

McClanahan, T. R. and R. W. Wolfe. 1993. Accelerating forest succession in a fragmented landscape: the role of birds and perches. 1993. Conservation Biology 7:279-288.

McCullagh, P. and J. A. Nelder. 1989. Generalized linear models. Monograph on Statistics and Applied Probability No. 37. Chapman and Hall, New York.

Meffe, G.K., and C.R. Carroll. 1998. Principles of Conservation Biology, second edition. Sinaur Associates, Sunderland, Massachusetts.

Murphy, D.D. 1990. Conservation biology and the scientific method. Conservation Biology 4:203-204.

Nicholls, A.O. and C.R. Margules. 1991. The design of studies to demonstrate the biological importance of corridors. Pp. 49-64 in D.A. Saunders and R.J. Hobbs editors. Nature conservation 2: The role of corridors. Hobbs, Surrey, Beatty & Sons, Australia.

Norton, B. G. 1988. What is a conservation biologist? Conservation Biology 2:237-238.

Norton, D. A., R. J. Hobbs, and L. Atkins. 1995. Fragmentation, disturbance, and plant distribution: mistletoes in woodland remnants in the western Australian wheatbelt. Conservation Biology 9:426-438.

Noss, R.F. 1987. Corridors in real landscapes: a reply to Simberloff and Cox. Conservation Biology 1:159-164.

Noss, R. F. 1991. Landscape connectivity: different functions at different scales. Pages 27-39 in W. E. Hudson, ed. Landscape linkages and biodiversity. Island Press, Washington, DC.

Noss, R.F. 1992. Wildlife corridors. In D. Smith and P. Hellmund, editors. Ecology of

Greenways. University of Minnesota Press, Minneapolis, Minnesota.

Pulliam, H.R. Sources, sinks, and population regulation. American Naturalist 132:757-785.

Pulliam, H.R., J.B. Dunning, and J. Liu. 1992. Population dynamics in complex landscapes: a case study. Ecological Applications 2:165-177.

Robinson, G. R., and S. N. Handel. 1993. Forest restoration on a closed landfill: rapid addition of new species by bird dispersal. Conservation Biology 7:271-279.

Rosenberg, D.K., B.R. Noon, and E.C. Meslow. 1997. Ecological role of linear conservation areas for maintaining biological diversity. Bioscience 47:677-687.

Rosenberg, D.K., B.R. Noon, J.W. Megahan, and E.C. Meslow. 1998. Compensatory behavior of Ensatina eschscholtzii in biological corridors: a field experiment. Canadian Journal of Zoology 76:117-133.

SAS Institute Incorporated. 1996. SAS/STAT software: changes and enhancements. SAS Institute Inc., Cary, NC. Saunders, D.A. and R.J. Hobbs, editors. 1991. The Role of Corridors. Surrey Beatty, Chipping Norton, New South Wales, Australia.

Saunders, D. A., Jr., R. J. Hobbs, and C. R. Margules. 1991. Biological consequences of ecosystem fragmentation: a review. Conservation Biology 5:18-32.

Schmiegelow, F.K.A., C.S. Machtans, and S.J. Hannon. 1997. Are boreal birds resilient to forest fragmentation? An experimental study of short-term community responses. Ecology 78:1914-1932.

Scott, J.A. 1975a. Movement of Precis coenia, a 'pseudoterritorial' submigrant (Lepidoptera: Nymphalidae). Journal of Animal Ecology 44:843-850.

Scott, J.A. 1975b. Flight patterns among eleven species of diurnal Lepidoptera. Ecology 56:1367-1377.

Scott, J.A. 1984. The butterflies of North America. Stanford University Press, Stanford, CA.

Simberloff, D. and J. Cox. 1987. Consequences and costs of conservation corridors. Conservation Biology 1:63-71.

Simberloff, D., J.A. Farr, J. Cox, D.W. Mehlman. 1992. Movement corridors: conservation or poor investment? Conservation Biology 6:493-504.

Sisk, T.D., N.M. Haddad, and P.R. Ehrlich. 1997. Bird assemblages in patchy woodlands: modeling the effects of edge and matrix habitats. Ecological Applications 7:1170-1180.

Sutcliffe, O.L. and C.D. Thomas. 1996. Open corridors appear to facilitate dispersal by ringlet butterflies (Aphantopus hyperantus) between woodland clearings. Conservation Biology 10:1359-1365.

Wilcove, D.S., D. Rothstein, J. Dubow, A. Phillips, and E. Losos. 1998. Quantifying threats to imperiled species in the United States. Bioscience, in press.

Wilcox, D.A., and D.D. Murphy. 1985. Conservation strategy: the effects of fragmentation on extinction. The American Naturalist 125:879-887.

Wilson, E.O., and E.O. Willis. 1975. Applied biogeography. Pages 522-534 in M.L. Cody and J.M. Diamond, editors. Ecology and evolution of communities. The Belknap Press, Cambridge, Massachusetts.

Workman, S.W., and K.W. McLoed. 1990. Vegetation of the Savannah River Site: major community types. A Publication of the Savannah River Site, Aiken, SC.

Zhang, Z. and M.B. Usher. 1991. Dispersal of wood mice and bank voles in an agricultural landscape. Acta Theriologica 36:239-245.