Experimental evidence does not support the Habitat Amount Hypothesis

Ecography 40: 48 55, 2017 doi: 10.1111/ecog.02535 2016 The Authors. Ecography 2016 Nordic Society Oikos Subject Editor: Robert Holt. Editor-in-Chief: Miguel Araújo. Accepted 5 October 2016 Experimental evidence does not support the Habitat Amount Hypothesis Nick M. Haddad, Andrew Gonzalez, Lars A. Brudvig, Melissa A. Burt, Douglas J. Levey and Ellen I. Damschen N. M. Haddad (http://orcid.org/0000-0002-1591-9322) (haddad@ncsu.edu), Dept of Applied Ecology, North Carolina State Univ., Raleigh, North Carolina 27695, USA. A. Gonzalez, Dept of Biology, McGill Univ., Montreal, QC H3A 1B1, Canada. L. A. Brudvig, Dept of Plant Biology and Program in Ecology, Evolutionary Biology and Behavior, Michigan State Univ., East Lansing, Michigan 48824, USA. M. A. Burt and E. I. Damschen, Dept of Zoology, Univ. of Wisconsin, Madison, Wisconsin 53706, USA. D. J. Levey, National Science Foundation, Division of Environmental Biology, Arlington, Virginia 22230, USA. For a half century, habitat configuration the arrangement of habitat patches within a landscape has been central to theories of landscape ecology, population dynamics, and community assembly, in addition to conservation strategies. A recent hypothesis advanced by Fahrig (2013) would, if supported, greatly diminish the relevance of habitat configuration as a predictor of diversity. The Habitat Amount Hypothesis posits that the sample area effect overrides patch size and patch isolation effects of habitat fragmentation on species richness. It predicts that the amount of habitat in a local landscape, regardless of configuration, is the main landscape-level determinant of species richness. If habitat amount is indeed the major, landscape-level driver of species richness, the slopes of the species area relationship (SAR) for otherwise similar fragmented and unfragmented landscapes should be indistinguishable. We tested the Habitat Amount Hypothesis with data from two replicated and controlled habitat fragmentation experiments that disentangle the effects of habitat amount and configuration. One experiment provided time-series data on plant species richness and the other on micro-arthropod species richness. We found that, relative to less fragmented habitats, the SARs for fragmented habitats have significantly higher slopes and that the magnitude of the difference in slopes increased over time. Relatively more species were lost in smaller areas when fragments were more isolated. In both experiments, the proportion of species lost due to increased habitat fragmentation was nearly identical to the proportion lost due to reduced habitat amount. Our results provide a direct and experimentally derived refutation of the Habitat Amount Hypothesis, supporting the long-held view that in addition to area, patch isolation and configuration are important determinants of species richness. Differences in species richness between fragmented and non-fragmented habitats increase over time, demonstrating that long-term studies are needed to understand the effects of fragmentation, above and beyond the amount of habitat lost. Since the publication of Applied biogeography (Wilson and Willis 1975), the sizes and configurations of habitat fragments have been at the center of landscape conservation efforts. Expanding the implications of the Theory of island biogeography (MacArthur and Wilson 1963, 1967), Wilson and Willis (1975) predicted reductions to species richness in small and isolated fragments, analogous to patterns observed for oceanic islands. They then advanced perspectives on how alternative landscapes with the same habitat area but different configurations might assist biodiversity conservation through management actions, such as increasing fragment area, reducing the amount of habitat edge, and connecting fragments with landscape corridors. Central to this perspective is the notion that both the area of individual fragments and configuration of those fragments across a landscape are core determinants of local (e.g. fragment-scale) species richness. Such applications of island biogeography theory to conservation have been questioned since their inception. Simberloff and Abele (1976) asserted that the cost and irreversibility of large-scale conservation programs demand a prudent approach to the application of an insufficiently validated theory. For example, enlarging and connecting habitats within reserves might have unintended consequences, including reducing the diversity of species conserved (Simberloff and Cox 1987). Emerging empirical evidence raised further questions when Quinn and Harrison (1988) showed identical areas of small islands and fragments can have higher species richness than large islands and fragments. Renewed skepticism later accompanied the increased attention to spatially distributed populations (Doak and Mills 1994, Harrison 1994). In parallel with this dialogue, the effects of habitat fragmentation (i.e. isolation, edge, or area effects) on species richness have been supported by mechanisms arising from both theory and observation (Leroux et al. 2017). The original theory relied on just two processes: rates of island coloniza- 48

tion and extinction. Increased recognition of spatial subdivision of populations (i.e. Huffaker 1958, Ehrlich 1961) motivated new theory of the spatial dynamics of metapopulations (Levins 1969). Although perhaps slower to spread its reach into conservation, concern about populations in fragmented landscapes accelerated application of metapopulation research (Lande 1987, Hanski 1999). Metacommunity models have provided a mechanistic bridge between spatial population models and island biogeography theory, Leroux et al. 2017), incorporating multiple species and their interactions to deepen our understanding of how habitat loss and fragmentation impact species richness (Tilman et al. 1994, Holyoak et al. 2005, Gonzalez et al. 2009, Vellend 2010, Thompson et al. 2017). The body of empirical evidence though not uniform has affirmed the relevance of both fragment area and configuration for the diversity of species within fragments and across fragmented landscapes (Didham et al. 2012, Haddad et al. 2015). Habitat fragmentation and total habitat area act in concert; in landscapes where habitat has been lost, possibly beyond a threshold that has been estimated around 70%, fragmentation reduces species richness and degrades ecosystem function even further (Hanski 2015). Despite the theoretical and empirical support for the dual importance of total habitat area and fragmentation, skepticism remains. Following logic nearly indistinguishable from Simberloff and Abele (1976), Fahrig (2013) recently advanced the Habitat Amount Hypothesis: the main characteristic of landscapes that affects species richness in a given habitat is the total amount of that habitat in the landscape. This hypothesis is premised on a species area relationship (SAR) that is independent of the configuration of habitat fragments across a landscape (Fahrig 2013). Fahrig questioned whether SAR slopes for fragments would differ in any detectable way from SAR slopes formed by sampling the same areas of contiguous habitat because of the lower barrier to dispersal in terrestrial landscapes, compared to the oceanic islands upon which fragmentation theory is predicated (Fig. 1a). If supported, the Habitat Amount Hypothesis would fundamentally transform how landscapes are conserved by shifting focus away from considerations of habitat configuration and connectivity, toward total habitat area alone. The Habitat Amount Hypothesis states that SARs can be explained by a simple sampling effect of identical numbers of species in identical areas, and that this relationship is necessary and sufficient to explain all fragmentation effects. It assumes that all patch size effects are simply linked to how increasing area (whether or not within fragments) represents increasingly larger samples of the regional species pool. Whether those areas are surrounded by identical or contrasting habitats, they sample from an identical regional pool of species (Fig 4 in Fahrig 2013). This hypothesis thus ignores evidence of oceanic island effects, where extinction rates are greater than colonization rates in small habitat fragments. Since there is evidence for island effects among small fragments (Gonzalez 2000a, Chisholm et al. 2011), an unresolved tension is whether these island-like effects, independent of sampling effects, can cause the SAR to steepen. A strong test of the Habitat Amount Hypothesis would control habitat amount and vary fragmentation; areas sampled in fragmented and continuous habitats would then be Figure 1. (a) Species area relationship (SAR) hypotheses adapted from Fahrig (2013, Fig. 2) The Samples line represents the relationship for areas sampled within contiguous habitat, the Islands line represents the relationship for islands surrounded by ocean. The difference in slope between the horizontal line and Samples represents the pure area effect expected from sampling smaller areas of habitat. The Habitat Amount Hypothesis posits that the slopes of Samples and Fragments may be indistinguishable. (b) SARs for the SRS Corridor Experiment (2014 data). The relationship for Connected fragments is compared to that for Isolated fragments. In one test (Isolated areas, solid line and white diamonds, Table 1a), the largest sampled area for connected and isolated fragments is the entire area encompassed by the two connected fragments. In a second test (Isolated areas, dashed line and circles, Table 1c), the largest isolated areas are pooled among isolated fragments and are independent of connected areas. (c) SARs for the Moss Fragmentation Experiment (month 12 data; Table 1b). The relationship for samples within contiguous areas is compared to the same area fragments. Error bars are 95% CIs. In every test, there are significant effects of both area and fragmentation (Table 1). In the SRS Corridor Experiment, there is a significant interaction between area and fragmentation only in the test where the largest connected patch and the largest fragment are identical areas (Table 1). used to construct comparable SARs. The ability to detect a steeper SAR depends (among other things) upon how well one samples the communities, how strong the isolation effect is, and the temporal scale over which fragmentation effects 49

Figure 2. (a) One experimental block in the SRS Corridor Experiment. Connected areas all occur in the two connected fragments. Unconnected areas occur within areas in individual isolated fragments. Further, the largest area is either the entire connected fragment (largest connected and isolated fragments areas are identical, last row of left two columns) or pooled areas of isolated fragments that are completed independent of connected fragments (last row of left and right columns). (b) One experimental block in the Moss Experiment. Fragments were destructively sampled every two months for one year. Control areas were sampled from within large contiguous areas of moss, and samples were equal to the area of fragments. The edge:area ratio of the fragments was identical for all replicates within the small and large fragment treatments. are allowed to unfold (Chase and Knight 2013). The Habitat Amount Hypothesis would be supported if SARs were indistinguishable between landscapes with identical amounts of habitat, but different configurations of fragments (Fig. 1a). Conversely, evidence against the Habitat Amount Hypothesis would result from a steeper SAR in fragmented landscapes (where patches are isolated), relative to less or unfragmented landscapes with an equivalent total amount of habitat. Using data from two different experiments in which landscape-scale habitat amount and fragment isolation are controlled, we construct SARs for connected and isolated fragments to test the primary prediction of the Habitat Amount Hypothesis (Fahrig 2013): if habitat amount is controlled, SARs should not be affected by landscape configuration. Our Fig. 1a is adapted from Fig. 2 in that article (with precedent in Fig. 3 of MacArthur and Wilson 1967, Holt 1996). The two experiments overcome the limitation of observational studies that typically confound habitat amount and configuration (Didham et al. 2012). In the Savannah River Site (SRS) Corridor Experiment, data on plant species richness have been collected continuously for 15 years, and fragments connected by landscape corridors support greater species richness than in isolated fragments of equivalent area (Damschen et al. 2006). The amount of habitat in replicate local landscapes is controlled and fragments are identical in every way except in their connectivity, providing a valid test of the Habitat Amount Hypothesis. In the Moss Fragmentation Experiment, data on the community of microarthropods responding to changes in moss habitat area and isolation were collected for 12 months, and have also shown that species richness is higher in equal-sized areas of connected (continuous) as compared to isolated fragments and when moss fragments are connected by corridors (Gonzalez et al. 1998). By comparing the slopes (predicted by the Habitat Amount Hypothesis to be equal) of SARs within each experiment, our analysis provides a direct test of the hypothesis that habitat amount, not configuration, mediates species richness. Our primary analysis focuses on the last sampling period in both experiments. Yet, SARs would be expected to change through colonization and extinction processes that could take many generations to materialize. Temporal change is expected Figure 3. The change in the slope of the SAR for continuous habitat and fragments through time: (a) the SRS Corridor Experiment comparing z between Connected and Isolated fragments and (b) the value of z for Isolated habitat was greater than the Continuous habitat at six and twelve months in the Moss Experiment. Asterisks denote individual time points when the relationship was statistically significant. In the SRS Corridor Experiment, the slopes of Isolated fragments became significantly steeper over time, relative to the slopes of Connected fragments. 50

based on the theory of island biogeography (Simberloff and Wilson 1969), the extinction debt (Tilman et al. 1994), and empirical responses to fragmentation (Haddad et al. 2015). Both experiments occurred over long time periods relative to species generation times, permitting us to test a second hypothesis: in fragments, the slope of the SAR increases relative to the SAR of continuous habitat over time. Methods SRS Corridor Experiment This experiment was created in 2000 at the Savannah River Site (SRS), a National Environmental Research Park in Aiken and Barnwell Counties, South Carolina, USA. Each of eight blocks contained five fragments of open habitat created by clearing pine (Pinus taeda and P. palustris) plantation forest (Fig. 2), which formed the matrix between patches and extended at least 50 m beyond the edge of peripheral fragments. Within each block, a central fragment was 1 ha in area. The central fragments were connected by a 150 m long by 25 m wide corridor to another 1 ha fragment. Three other fragments were each equal in area to a connected fragment plus the corridor (1.375 ha). These fragments varied in shape to isolate effects of corridors that might have been attributed to edge:area ratio (Tewksbury et al. 2002). As isolation, and not other fragment characteristics, explained differences in vascular plant species richness (Damschen et al. 2006), we do not discuss the effects of fragment shape further. In sum, this experiment manipulated patch connectivity while controlling for landscape-scale habitat amount i.e. the total area of habitat within each block is equal (6.5 ha), the area of each isolated fragment is equal (1.375 ha), and the area of two connected fragments is 2.375 ha. Within each fragment, we have sampled plant species richness annually since 2000 (Damschen et al. 2006). Fragments were sampled systematically by searching the entire area of each fragment. Surveys occurred each June, when herbaceous plants were most easily identified. Fragments have been maintained as open savannas with few shrubs and P. palustris, the historical community type, with controlled burns and hardwood removal to replicate natural rates of disturbance, providing high and consistent detection of species among fragments. During surveys, each unique plant species was cumulatively recorded such that species richness for a 1ha area within each fragment and the full fragment area (e.g. core area plus corridor) could be determined. Moss Fragmentation Experiment The Moss Fragmentation Experiment created fragmented and continuous areas of moss (Hypnum compressiforme and Tortella tortuosa) growing on boulders in Derbyshire Park District, UK. Eight boulders were selected to represent eight replicate landscapes. These landscapes were treated as statistical blocks and in a randomized design. Each landscape included a continuous area (0.25 m 2 ) and an adjacent fragmented area. The fragmented area was composed of six large fragments (200 cm 2 ) and six small fragments (20 cm 2 ). Moss fragments were created and isolated by removing moss from around a circular wire template. Each fragment was a minimum of 15 cm from the nearest fragment, a distance considered to be large relative to the dispersal abilities of the resident micro-arthropods. The data analyzed from this experiment were derived from a previously published paper (Gonzalez 2000b). Fragments were sampled every two months for one year. Destructive sampling at each time point was performed on one large and one small fragment per block, as well as equal areas of continuous habitat. Removed moss was placed in Tullgren funnels, and all micro-arthropods (mainly soil mites and collembola) were identified to morphospecies and counted. This method extracts 90% of individuals (Gonzalez et al. 1998) and so represents an almost complete census of all the individuals in the moss area. Variation in species richness is calculated from a complete census as opposed to a rarified estimate based on a small sample. Over the 12-month duration of the experiment, more than 98 000 individuals were identified and species richness was quantified. Analysis Fragmentation effects on slopes and intercepts We fit general linear mixed models to test the effects of sampling area and fragment connectivity on species richness. To analyze the significance of terms in the Species Area Relationship, we converted S ca z to log(s) z loga c, log-transforming the number of species and habitat area. For analyses, sample area was treated as a fixed effect, and experimental block was treated as a random effect, providing estimates of slope (z) and intercept (c). Although the intercept does not directly test the Habitat Amount Hypothesis, it enables us to calculate the relative impacts of area and fragmentation (Analysis). Fragment type (connected or unconnected in the SRS experiment and continuous or fragmented in the Moss Experiment) and the interaction between fragment type and area were also included as fixed effects to determine differences in c and z, respectively. Whereas the Moss Experiment was comprised of independent treatments (fragmented or continuous) in which we were able to analyze each area as an independent sample, the SRS Experiment pooled areas that were connected or isolated, in which two connected fragments were also the largest fragment. These constitute island and nested sampling schemes, respectively, which are both valid methods for constructing SAR s (Scheiner 2003). Connected areas were (Fig. 2a): 1) the area of individual (1 ha) fragments connected by a corridor (small), 2) one fragment plus the corridor connected to it (medium), and 3) the summed area of two fragments connected by a corridor (large). Isolated areas were assessed using 1) individual isolated fragments (each 1.375 ha; medium) and 2) the largest fragment, the summed area of two fragments connected by a corridor (large; Fig. 2a). In the SRS Experiment, habitat amount could be accrued in a different way, with larger sampling areas derived from fragment areas that were independent (Fig. 2a). To test the effects of fragmentation in independent areas, the area 51

of isolated fragments was: 1) one unconnected fragment (small), 2) parts of two unconnected fragments (medium), or 3) all of two unconnected fragments (large), pooling species richness from larger and larger unconnected areas to create species area relationships. The analysis was otherwise identical. Fragmentation effects over time To determine the temporal effects of fragmentation on SARs, we analyzed the slope of the relationship between area and number of species (z) among fragment types every year in the SRS Experiment over the entire time series 2001 2014. In addition, to test change over time, we created a mixed model with fragment type as a fixed effect, block as a random effect, year as a repeated measure, and z as the response variable. In the Moss Experiment, repeated measures analysis was not appropriate because destructively sampling meant a new fragment was used at each time point; confidence intervals around values of z were compared every-other-month for one year (Gonzalez 2000b). Relative effects of habitat amount and habitat fragmentation We compared the relative effects of habitat amount and fragmentation on species richness, with the y-intercept, c, as the standard for comparison of effect sizes among SARs. To do this, we calculated the proportional effect size of habitat amount as (S max c connected )/S max, where S max is species richness of the largest connected fragment and c is the y-intercept. We calculated the proportional effect size of habitat fragmentation as (c connected c isolated )/c connected. Data available from the Dryad Digital Repository: < http://dx.doi.org/10.5061/dryad.k88h6 > (Haddad et al. 2017). Results Fragmentation effects on slopes and intercepts We first analyzed the final data points from each experiment, 2014 in the SRS Experiment and after 12 months in the Moss Experiment. In an analysis of the SRS Experiment in which two connected fragments were also the largest fragment, fragmentation effects (as interpreted from log-transformed data) caused a 39% increase in z (slope) and a 3% reduction in c (y-intercept; Fig. 1b, Table 1a). We can quantify the relative impact of habitat amount and fragmentation by back-transforming our results. This yields, over the range of our fragment sizes, S max 129 species c connected 100 species and c isolated 84.5 species. Using these empirically derived estimates, we find that habitat amount effects reduced species richness by 22% and fragmentation effects reduced species richness by 16%. In a second analysis of effects of fragmentation in independent areas, we considered species richness in larger areas of unconnected habitat by pooling richness across separate unconnected fragments. Fragmentation had a significant, negative effect on species richness (p 0.01), resulting in a 2% reduction in c (Fig. 1b, Table 1b). We found effects of fragment type and area (p 0.001) on species richness, but no significant interaction between these variables (although there was a positive effect of area on number of species, z was indistinguishable between connected and unconnected fragments). In the Moss Experiment, fragmentation caused a 34% increase in z and a 33% reduction in c (Fig. 1c, Table 1c). Again after back-transforming results over the range of fragment sizes, S max 24.5 species, c connected 10 species, and c isolated 5 species. As in the SRS experiment, habitat amount and fragmentation reduced species richness by 59% and 50%, respectively. Table 1. (a) Random effects from year 2014, SRS Corridor Experiment, when two connected fragments are equal to the largest isolated fragment. All categorical effects estimate connected relative to unconnected fragments. (b) Random effects from month 12, Moss Experiment, for which all categorical effects estimate continuous habitats relative to unconnected fragments. (c) Random effects from year 2014, SRS Corridor Experiment, when connected fragments are independent of isolated fragments. (a) SRS Corridor Experiment (two connected fragments largest isolated fragment). Effect Fragment type Estimate Standard error DF t-value Pr t Intercept 1.9272 0.01953 7 98.70 0.0001 Fragment type Connected 0.07636 0.02069 45 3.69 0.0006 Area 0.4906 0.07275 45 6.74 0.0001 Area Fragment type Connected 0.1925 0.09160 45 2.10 0.0412 (b) Moss Experiment. Effect Fragment type Estimate Standard error DF t-value Pr t Intercept 0.6805 0.07667 6 8.88 0.0001 Fragment type Continuous 0.3289 0.1084 16 3.03 0.0079 Area 0.2529 0.04023 16 6.29 0.0001 Area Fragment type Continuous 0.08604 0.05689 16 1.51 0.1500 (c) SRS Corridor Experiment (analysis of independent areas). Effect Fragment type Estimate Standard error DF t-value Pr t Intercept 1.9638 0.01477 7 132.97 0.0001 Fragment type Connected 0.03975 0.01558 85 2.55 0.0125 Area 0.3012 0.03441 85 8.75 0.0001 Area Fragment type Connected 0.00314 0.06233 85 0.05 0.9600 52

Fragmentation effects over time Over the duration of the SRS Experiment, fragmentation yielded a significant (p 0.001) effect of area on species richness (z in SAR) every year from 2001 2014. There was a significant interaction between fragment type and area (different values of z in connected and unconnected fragments) on species richness (p 0.05 each year) in 2005, 2006, 2008, 2009, 2011 2014 (Fig. 3). A repeated measures analysis revealed a significant effect of time; z increased in unconnected fragments (p 0.03) and thus the difference in z between connected and unconnected fragments grew over the duration of the study. Over the duration of the Moss Experiment, there was a significant effect of area on species richness for months 2 12 (p 0.01; Gonzalez 2000b). Samples in continuous areas had significantly greater species richness in large fragments in months 8 and 12 (p 0.001) and in small areas in months 6 and 12. The slope of the SAR (z) was significantly greater in fragments compared to continuous habitat in months 6 and 12 (p 0.05; Fig. 3, adapted from Fig. 3 in Gonzalez 2000b). Discussion Our results reject the Habitat Amount Hypothesis (Fahrig 2013); both amount and configuration of habitat affect species richness. Using controlled, randomized, and wellreplicated experiments with vastly different spatial scales and taxa, we show that the slope of the SAR is steeper (Fig. 1b, largest fragment 2 connected fragments; Fig. 3) in fragmented than in more continuous landscapes. By isolating fragmentation effects, our experimental approach overcomes a common limitation in natural landscapes, where habitat amount and connectivity/isolation for individual fragments are typically confounded (Fahrig 2013). Despite finding no support for the central hypothesis, we were able to find support for an exception expected from the Habitat Amount Hypothesis (Fig. 10 in Fahrig 2013): that fragments embedded in lower-quality matrix habitats will yield equivalent slopes but lower y-intercepts of SARs (Fig. 1b, Independent Areas). Fundamentally, the predictions of the Habitat Amount Hypothesis only concern slope. However, if the same total area of fragments is surrounded by different matrix habitats, the fragments would necessarily be expected to differ in their resistance to dispersal and in their ability to support demographically healthy populations of plants or animals (Laurance 2008, Kuefler et al. 2010). Our results from the SRS Fragmentation Experiment for isolated fragments which are completely independent of connected fragments (Fig. 2) are consistent with this expectation. Yet, considered in this way, differences between high-quality matrix and corridor are simply semantic. The factors that cause species richness to increase in connected fragments in our two tests from the SRS Corridor Experiment can be attributed only to habitat configuration. This is an example of how the caveats and delimitations (Fahrig 2013) of the Habitat Amount Hypothesis are used to construct a hypothesis that does not present clear, independent predictions. Another limitation of the Habitat Amount Hypothesis regards edge effects. The hypothesis treats habitat near and far from an edge within a fragment as distinct. However, by acknowledging edge effects, the hypothesis recognizes the existence of a fragment, its ecological properties, and its functional impacts on individuals and populations as the edge:area ratio increases. An extensive literature on edge effects (summarized in Ries et al. 2004) provides yet another line of evidence for why habitat amount alone is inadequate for understanding patterns of diversity across a landscape. Regardless, both the Moss and SRS Experiments control for edge effects by keeping edge:area ratios among connected and unconnected fragments constant. We show that the effects of fragmentation increase over time, and in both experiments never reached an asymptote (Fig. 3). More generally, a variety of fragmentation experiments conducted worldwide have shown that fragmented ecosystems lose species and ecosystem function continuously over decades, with no plateau to these effects even after 25 years (Haddad et al. 2015). An emerging lesson is that ecological tests conducted soon after fragmentation are inadequate because of extinction debt (Tilman et al. 1996, Mouquet et al. 2011), which is apparently common in fragmented landscapes (Haddad et al. 2015). More broadly, and counter to the Habitat Amount Hypothesis, not only is habitat fragmentation an important contributor to patterns of species richness, but its temporally dynamic effects must be explicitly considered in human-altered landscapes. Counter to many recent studies that explicitly quantify fragmentation effects, the Habitat Amount Hypothesis ignores mechanisms underlying changing patterns of species diversity. Considered over decades, the body of literature in spatial ecology has moved beyond describing spatial patterns alone, to address mechanisms underlying population and community processes (Tilman and Kareiva 1997, Holyoak et al. 2005). Our experiments build on knowledge of the mechanisms hypothesized to underlie the effects of fragmentation on SARs. These mechanisms have arisen from theory of spatial populations (Lande 1988, Pulliam 1988, Hanski 1994a, 1994b, Hanski et al. 1995) and communities (Holyoak et al. 2005, Vellend 2010), and have been tested and supported in detailed empirical studies (Davies et al. 2000, Haddad et al. 2003, Cook et al. 2005, Damschen et al. 2008, Fáveri et al. 2008). Furthermore, the focus on only one dimension of communities, species richness, ignores a broader examination of the effects of habitat configuration ecosystem-wide effects (Wilson et al. 2016) for example, on aspects of ecosystem functioning (with links to services, Mitchell et al. 2013, 2014, Chaplin-Kramer et al. 2015, Haddad et al. 2015), species composition, functional diversity, and foodwebs (Pillai et al. 2011, Collins et al. 2017). The Habitat Amount Hypothesis is a mechanism-free statement of correlation between area and species richness. Data from our two experiments offer no support for the hypothesis expectations, resulting in its rejection for its expected impacts on species richness. Our results are consistent with those of previous studies that recognize habitat fragmentation as a driver of species loss, with habitat fragmentation rising in importance as remaining habitat area declines (Rybicki and Hanski 2013, Banks-Leite et al. 2014). 53

In conservation, only considering one attribute of landscapes (i.e. habitat amount) and only one aspect of changing species diversity (i.e. sampling effect) could misguide protected area management. Other effects of fragmentation, including varying combinations of fragment area, edge, matrix quality, and isolation, are known to drive changes in species diversity (Didham et al. 2012). Consistent with applied conservation (Resasco et al. 2017), our results also emphasize the importance of connectivity, the value of prioritizing habitat and corridors for connectivity (Pascual-Hortal and Saura 2006), and configuring landscapes to reduce habitat isolation (Gilbert-Norton et al. 2010). We advocate for research that focuses on understanding mechanisms that impact species diversity and ecosystem processes in fragmented landscapes above and beyond habitat amount. Acknowledgments We thank Bill Morris, Lenore Fahig and the Fahrig lab for comments on the manuscript. Lenore Fahrig provided independent comments from the rest of Fahrig lab. AG is supported by an NSERC Discovery grant, a Killam Fellowship, and the Canada Research Chair program. We thank the Theoretical and Experimental Ecology Station in Moulis, France, for hosting a workshop on fragmentation workshop. NH, ED, LB, and MB are supported by the National Science Foundation (DEB-1354218, DEB-1354085, DEB-1354101). NH, ED, LB, MB, and DL thank the USDA Forest Service-Savannah River for support in creating and maintaining the SRS Corridor Experiment. References Banks-Leite, C. et al. 2014. Using ecological thresholds to evaluate the costs and benefits of set-asides in a biodiversity hotspot. Science 345: 1041 1045. Chaplin-Kramer, R. et al. 2015. Degradation in carbon stocks near tropical forest edges. Nat. Comm. 6: doi 10.1038/ ncomms10158 Chase, J. M. and Knight, T. M. 2013. Scale-dependent effect sizes of ecological drivers on biodiversity: why standardised sampling is not enough. Ecol. Lett. 16: 17 26. Chisholm, C. et al. 2011. Metacommunity diversity depends on connectivity and patch arrangement in heterogeneous habitat networks. Ecography 34: 415 424. Collins, C. D. et al. 2017. Fragmentation affects plant community composition over time. Ecography 40: 119 130. Cook, W. M. et al. 2005. Secondary succession in an experimentally fragmented landscape: community patterns across space and time. Ecology 86: 1267 1279. Damschen, E. I. et al. 2006. Corridors increase plant species richness at large scales. Science 313: 1284 1286. Damschen, E. I. et al. 2008. The movement ecology and dynamics of plant communities in fragmented landscapes. Proc. Natl Acad. Sci. USA 105: 19078 19083. Davies, K. F. et al. 2000. Which traits of species predict population declines in experimental forest fragments? Ecology 81: 1450 1461. Didham, R. K. et al. 2012. Rethinking the conceptual foundations of habitat fragmentation research. Oikos 121: 161 170. Doak, D. F. and Mills, L. S. 1994. A useful role for theory in conservation. Ecology 75: 615 626. Ehrlich, P. R. 1961. Intrinsic barriers to dispersal in checkerspot butterfly. Science 134: 108 109. Fahrig, L. 2013. Rethinking patch size and isolation effects: the habitat amount hypothesis. J. Biogeogr. 40: 1649 1663. Fáveri, S. B. et al. 2008. Effects of Amazonian forest fragmentation on the interaction between plants, insect herbivores, and their natural enemies. J. Trop. Ecol. 24: 57 64. Gilbert-Norton, L. et al. 2010. A meta-analytic review of corridor effectiveness. Conserv. Biol. 24: 660 668. Gonzalez, A. 2000a. Community relaxation in fragmented landscapes: the relation between species richness, area and age. Ecol. Lett. 3: 441 448. Gonzalez, A. 2000b. Community relaxation in fragmented landscapes: the relation between species richness, area and age. Ecol. Lett. 3: 441 448. Gonzalez, A. et al. 1998. Metapopulation dynamics, abundance, and distribution in a microecosystem. Science 281: 2045 2047. Gonzalez, A. et al. 2009. Biodiversity as spatial insurance: the effects of habitat fragmentation and dispersal on ecosystem functioning. Biodiv. Ecosyst. Func. Ecosyst. Serv., pp. 134 146. Haddad, N. M. et al. 2003. Corridor use by diverse taxa. Ecology 84: 609 615. Haddad, N. M. et al. 2015. Habitat fragmentation and its lasting impact on Earth s ecosystems. Sci. Adv. 1: doi:10.1126/ sciadv.1500052 Haddad, N. M. et al. 2017. Data from: Experimental evidence does not support the Habitat Amount Hypothesis. Dryad Digital Repository, < http://dx.doi.org/10.5061/dryad. k88h6 >. Hanski, I. 1994a. Patch-occupancy dynamics in fragmented landscapes. Trends Ecol. Evol. 9: 131 135. Hanski, I. 1994b. A practical model of metapopulation dynamics. J. Anim. Ecol. 63: 151 162. Hanski, I. 1999. Metapopulation dynamics and conservation biology. In: Hanski, I. (ed.), Metapopulation ecology. Oxford Univ. Press, pp. 179 203. Hanski, I. 2015. Habitat fragmentation and species richness. J. Biogeogr. 42: 989 993. Hanski, I. et al. 1995. Metapopulation persistence of an endangered butterfly in a fragmented landscape. Oikos 72: 21 28. Harrison, S. 1994. Metapopulations and conservation. In: Edwards, P. J. et al. (eds), Large-scale ecology and conservation biology. Blackwell Sci. Publ., pp. 111 128. Holt, R. D. 1996. Temporal and spatial aspects of food web structure and dynamics. In: Polis, G. A. and Winemiller, K. O. (eds), Food webs: integration of patterns and dynamics. Chapman and Hall, pp. 255 257. Holyoak, M. et al. 2005. Metacommunities: spatial dynamics and ecological communities. Univ. of Chicago Press. Huffaker, C. B. 1958. Experimental studies on predation: dispersion factors and predator prey oscillations. Hilgardia 27: 343 383. Kuefler, D. et al. 2010. The conflicting role of matrix habitats as conduits and barriers for dispersal. Ecology 91: 944 950. Lande, R. 1987. Extinction thresholds in demographic models of territorial populations. Am. Nat. 130: 624 635. Lande, R. 1988. Demographic models of the northern spotted owl (Strix occidentalis caurina). Oecologia 75: 601 607. Laurance, W. F. 2008. Theory meets reality: how habitat fragmentation research has transcended island biogeographic theory. Biol. Conserv. 141: 1731 1744. Leroux, S. et al. 2017. Structural uncertainty in models projecting the consequences of habitat loss and fragmentation on biodiversity. Ecography 40: 36 47. Levins, R. 1969. Some demographic and genetic consequences of environmental heterogeneity for biological control. Bull. Entomol. Soc. Am. 15: 237 240. MacArthur, R. H. and Wilson, E. O. 1963. An equilibrium theory of insular zoogeography. Evolution 17: 373 387. 54

MacArthur, R. H. and Wilson, E. O. 1967. The theory of island biogeography. Princeton Univ. Press. Mitchell, M. G. et al. 2013. Linking landscape connectivity and ecosystem service provision: current knowledge and research gaps. Ecosystems 16: 894 908. Mitchell, M. G. et al. 2014. Forest fragments modulate the provision of multiple ecosystem services. J. Appl. Ecol. 51: 909 918. Mouquet, N. B. et al. 2011. Extinction debt in source sink metacommunities. PLoS ONE 6: e17567. Pascual-Hortal, L. and Saura, S. 2006. Comparison and development of new graph-based landscape connectivity indices: towards the priorization of habitat patches and corridors for conservation. Landscape Ecol. 21: 959 967. Pillai, P. et al. 2011. Metacommunity theory explains the emergence of food web complexity. Proc. Natl Acad. Sci. USA 108: 19293 19298. Pulliam, H. R. 1988. Sources, sinks, and population regulation. Am. Nat. 132: 652 661. Quinn, J. F. and Harrison, S. P. 1988. Effects of habitat fragmentation and isolation on species richness: evidence from biogeographic patterns. Oecologia 75: 132 140. Resasco, J. et al. 2017. Theory, experiments, and the conservation of fragmented landscapes. Ecography 40: 109 118. Ries, L. et al. 2004. Ecological responses to habitat edges: mechanisms, models, and variability explained. Annu. Rev. Ecol. Syst. 35: 491 522. Rybicki, J. and Hanski, I. 2013. Species area relationships and extinctions caused by habitat loss and fragmentation. Ecol. Lett. 16: 27 38. Scheiner, S. M. 2003. Six types of species area curves. Global. Ecol. Biogeogr. 12: 441 447. Simberloff, D. and Cox, J. 1987. Consequences and costs of conservation corridors. Conserv. Biol. 1: 63 71. Simberloff, D. S. and Abele, L. G. 1976. Island biogeography theory and conservation practice. Science 191: 285 286. Simberloff, D. S. and Wilson, E. O. 1969. Experimental zoogeography of islands: the colonization of empty islands. Ecology: 278 296. Tewksbury, J. J. et al. 2002. Corridors affect plants, animals and their interactions in fragmented landscapes. Proc. Natl Acad. Sci. USA 99: 12923 12926. Thompson, P. L. et al. 2017. Loss of habitat and connectivity erodes species diversity, ecosystem functioning, and stability in metacommunity networks. Ecography 40: 98 108. Tilman, D. and Kareiva, P. M. 1997. Spatial ecology: the role of space in population dynamics and interspecific interactions. Princeton Univ. Press. Tilman, D. et al. 1996. Species fragmentation or area loss? Nature 382: 215 216. Tilman, D. et al. 1994. Habitat destruction and the extinction debt. Nature 371: 65 66. Vellend, M. 2010. Conceptual synthesis in community ecology. Quart. Rev. Biol. 85: 183 206. Wilson, E. O. and Willis, E. O. 1975. Applied biogeography. In: Cody, M. L. and Diamond, J. M. (eds), Ecology and evolution of communities. The Belknap Press, pp. 522 534. Wilson, M. C. et al. 2016. Habitat fragmentation and biodiversity conservation: key findings and future challenges. Landscape Ecol. 31: 219 227. 55