See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/225842575

Effects of spatial heterogeneity on butterfly species richness in Rocky Mountain National Park, CO, USA

Article in Biodiversity and Conservation · March 2008 DOI: 10.1007/s10531-008-9536-8

CITATIONS READS 51 285

3 authors:

Sunil Kumar Sara E. Simonson United States Department of Agriculture Colorado State University

107 PUBLICATIONS 2,969 CITATIONS 25 PUBLICATIONS 250 CITATIONS

SEE PROFILE SEE PROFILE

Thomas J. Stohlgren Colorado State University

243 PUBLICATIONS 13,198 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Semi-retirement! View project

The Entomologist's Monthly Magazine Volume 154, pg: 65-78. Published January 26 2018 View project

All content following this page was uploaded by Sunil Kumar on 02 June 2014.

The user has requested enhancement of the downloaded file. Biodivers Conserv (2009) 18:739–763 DOI 10.1007/s10531-008-9536-8

ORIGINAL PAPER

EVects of spatial heterogeneity on butterXy species richness in Rocky Mountain National Park, CO, USA

Sunil Kumar · Sara E. Simonson · Thomas J. Stohlgren

Received: 14 May 2008 / Accepted: 17 November 2008 / Published online: 11 December 2008 © Springer Science+Business Media B.V. 2008

Abstract We investigated butterXy responses to plot-level characteristics (plant species richness, vegetation height, and range in NDVI [normalized diVerence vegetation index]) and spatial heterogeneity in topography and landscape patterns (composition and conWgu- ration) at multiple spatial scales. StratiWed random sampling was used to collect data on butterXy species richness from seventy-six 20 £ 50 m plots. The plant species richness and average vegetation height data were collected from 76 modiWed-Whittaker plots overlaid on 76 butterXy plots. Spatial heterogeneity around sample plots was quantiWed by measur- ing topographic variables and landscape metrics at eight spatial extents (radii of 300, 600 to 2,400 m). The number of butterXy species recorded was strongly positively correlated with plant species richness, proportion of shrubland and mean patch size of shrubland. Patterns in butterXy species richness were negatively correlated with other variables including mean patch size, average vegetation height, elevation, and range in NDVI. The best predictive model selected using Akaike’s Information Criterion corrected for small sample size X (AICc), explained 62% of the variation in butter y species richness at the 2,100 m spatial extent. Average vegetation height and mean patch size were among the best predictors of butterXy species richness. The models that included plot-level information and topographic variables explained relatively less variation in butterXy species richness, and were improved signiWcantly after including landscape metrics. Our results suggest that spatial heterogeneity greatly inXuences patterns in butterXy species richness, and that it should be explicitly considered in conservation and management actions.

Keywords Akaike’s information criterion · ButterXy species richness · FRAGSTATS · Landscape context · Landscape metrics · Model selection · Plant species richness · Spatial autocorrelation · Spatial heterogeneity · Spatial scale

S. Kumar (&) · S. E. Simonson Natural Resource Ecology Laboratory, Colorado State University, 1499 Campus Delivery, A204 NESB Building, Fort Collins, CO 80523-1499, USA e-mail: [email protected]; [email protected]

T. J. Stohlgren U.S. Geological Survey, Fort Collins Science Center, Fort Collins, CO 80526-8118, USA 1 C 740 Biodivers Conserv (2009) 18:739–763

Abbreviations

AICc Akaike’s information criterion corrected for small sample size DEM Digital elevation model ESRI Environmental Systems Research Institute GIS Geographical information system GPS Global positioning system MODIS Moderate resolution imaging spectroradiometer NAD North American datum NASA National Aeronautics and Space Administration NDVI Normalized diVerence vegetation index NLCD National land cover dataset USGS United States Geological Survey

Introduction

Understanding how spatial heterogeneity aVects ecological patterns and processes is one of the major focuses of landscape ecology (Risser et al. 1984; Pickett and Cadenasso 1995; Turner et al. 2001; Fortin and Agrawal 2005; Turner 2005). Spatial heterogeneity can be deWned as the complexity and variability in ecological systems’ properties of interest in space (Li and Reynolds 1994). Quantifying spatial heterogeneity is needed to understand its eVects on the diversity and distributions of diVerent organisms and their species-speciWc responses (Gustafson 1998; Turner et al. 2001; Thies et al. 2003; Kumar et al. 2006). However, the decision at which scale to quantify spatial heterogeneity is one of the challenging questions that ecologists face because spatial heterogeneity is a complex phenomenon and is highly scale dependent (Kolasa and Rollo 1991; Gustafson 1998; Fortin and Agrawal 2005; Wagner and Fortin 2005). Spatial heterogeneity in ecological systems is caused by spatial interactions between many biotic and abiotic factors and the diVerential responses of organisms to these factors (Milne 1991) and the organisms themselves (Huston 1994). DiVerent organisms may have diVerential responses to spatial heterogeneity at multiple scales depending on their grain of perception (Levins 1968), the grain of the landscape (Forman and Godron 1986) and their natural history. Therefore, identiWcation of the factors that most inXuence species diversity, and the dominant scale (i.e., the scale that explains highest variation in the diversity and abundance of organisms) of response of species to these factors is important (Turner 2005) for maintaining and managing biodiversity. ButterXies and plants are generally predicted to show congruent patterns in species diversity due to ecological interactions involving herbivory and pollination (Opler and Krizek 1984; Scoble 1992), and their long history of mutual evolutionary inXuence (Ehrlich and Raven 1964). At local scales, many species of butterXies are restricted to one or a few closely related species of host plants that provide suitable food resources for the larvae (caterpillar; Opler 1999). Although butterXies require the suitable host plants as larval food resources, the geo- graphic distribution of a given butterXy is typically less extensive than the distribution of its potential host plants. A butterXy species may often be found near certain species of plants, but a butterXy will rarely be present in every area that the plant occurs. Across their ranges, butterXy species can show dramatic diVerences in breadth of host plant use and preferences for host plant species. The relative importance of individual plant species as butterXy host plants can vary in space, time, and even among individuals in the same population. 1 C Biodivers Conserv (2009) 18:739–763 741

The availability of Xoral nectar plant resources for adult butterXies can also be an impor- tant characteristic of the local vegetation. In a given area there is generally a greater variety of plant species that can potentially be used by butterXies as larval host plants than as nectar plants (Opler 1999). Nectar plants used by adult butterXies tend to be less speciWc than the host plants required by their larvae. A showy display of abundant nectar Xowers can be an attractive food resource for the adults of many diVerent butterXy species. However, there are examples of butterXy species that are linked closely with particular species of plants that they frequently visit for nectar resources. The diversity and distribution of plant species and characteristics of the vegetation may also inXuence patterns of butterXy species diversity by aVecting their movement and searching behavior (Kareiva 1983; Ricketts 2001). ButterXies can be sensitive to changes in environmental conditions such as vegetation structure, solar radiation, climate variabil- ity, weather events, and patterns in land-use (Wood and Samways 1991; Parmesan 1996; Fleishman et al. 2002; Luoto et al. 2006). Although it can be diYcult to determine the causal mechanisms underlying observed Xuctuations in butterXy populations, the rapid response of butterXies to changes in local vegetation and climate conditions suggests that they may be useful as indicators to monitor ecosystem properties and local habitats (Murphy and Weiss 1992; Kremen 1992; Pollard and Yates 1993; Parmesan 1996). For both practical and ecological reasons, butterXies have been suggested as potential indicator taxa to monitor ecosystems, habitat loss, non-native species invasion, fragmenta- tion, and climate change. Among , butterXies are well-studied, relatively easy to monitor, and have relatively short generation periods (Fleishman et al. 2002; Thomas 2005). For example, butterXies may exhibit rapid responses to disturbance events such as Wre and management activities such as logging because of their short generation time (Fleishman et al. 2002). ButterXy resources and their habitats often co-occur with other ecosystem properties of management interest, thus butterXies have also been proposed as indicator species for other taxa (Kremen 1992; Fleishman et al. 2002; Thomas 2005). Habitats used by butterXies may also support other lesser-known groups of conservation interest, including valuable plant pollinators such as moths, bees, ants, and Xies. ButterXies can also have important and diverse trophic links to other taxa, such as plant pollination services, butterXy–ant symbio- ses, and as food resources for bird predators (Gilbert and Singer 1975). Although the diversity and distribution of butterXies in a given area is determined in part by regional processes and biogeographic history, local environmental conditions can also have a large inXuence on observed species diversity. Biotic and abiotic factors of potential importance to butterXies can include the abundance and spatial distribution of host and larval plants, co-occurring plant and species, successional stages of vegetation, landscape composition, habitat conWguration, topography, irradiance, moisture availability, disturbance, climate, and weather events (Thomas and Mallorie 1985; Wood and Samways 1991; Kremen 1992; Pollard and Yates 1993; Parmesan 1996; Simonson et al. 2001; Fleishman et al. 2002; Krauss et al. 2003; Bergman et al. 2004; Hogsden and Hutchinson 2004; Davis et al. 2007). The responses of butterXies to spatial heterogeneity will depend on the interaction between their life histories and environment. Spatial arrangement of habitat patches in a landscape can aVect various butterXy activities such as foraging, mate-location, predator avoidance, and Wnding oviposition sites (Rabasa et al. 2005). Therefore, a butterXies’ ability to locate suitable resource patches and perform these activities in a complex mosaic of vegetation may be an important determinant of their abundance and distribu- tion across the landscape. Species movement can be facilitated or impeded where one 1 C 742 Biodivers Conserv (2009) 18:739–763 part of the landscape is relatively homogeneous with low degree of structural contrast and another part heterogeneous with high contrast (Forman and Godron 1986). Spatial heterogeneity can also inXuence butterXy population structure (Van Dyck and Matthysen 1999; Thomas et al. 2000). For example, increasing patch isolation can alter the exchange rate (emigration and immigration) of individuals between local populations (Thomas et al. 2000). A change in composition and conWguration of the landscape due to natural and anthropogenic disturbances can inXuence butterXy species richness by chang- ing patch size, patch shape, distance between the habitat patches, edge density, spatial variation of host and larval plant species, variation in prevailing winds, temperature and moisture across the landscape. Therefore, it is important to understand how butterXies respond to spatial heterogeneity. Moreover, knowledge about the factors which inXuence butterXy species richness across the landscape is a prerequisite for their conservation, monitoring, and management. In addition, land managers also need information about the spatial distribution of diVerent species in surveyed as well as unsurveyed areas to focus their conservation or restoration activities; therefore, predictive models are required because we can only survey a very small portion of the landscape (generally <1%; Stohlgren 2006). Only a few community level studies on butterXies quantiWed spatial heterogeneity at multiple spatial scales (e.g., Fleishman et al. 2002; Collinge et al. 2003; Krauss et al. 2003; Davis et al. 2007) and none of them fully integrated spatial heterogeneity in landscape patterns (i.e., composition and conWguration) with other types of spatial heterogeneity. Moreover, the studies that considered the eVects of landscape heterogeneity included only a few of its measures such as composition, patch size or isolation (Collinge et al. 2003; Bergman et al. 2004; Rabasa et al. 2005; Strathmann 2005). In this study we considered all Wve components of landscape heterogeneity: (1) number of patch types; (2) proportion of each type; (3) spatial arrangement of patches; (4) patch shape; and (5) contrast between neighboring patches (Li and Reynolds 1994), because one or two indices of landscape pattern can not capture all the aspects of landscape heterogeneity (Gustafson 1998). In addition, we also included plot level variables (e.g., plant species richness, vegetation height and range in Normalized DiVerence Vegetation Index (NDVI)) and heterogeneity in topography (e.g., elevation, slope, aspect, and distance from stream/river). We hypothesized that spatial heterogeneity may aVect butterXy species richness and that this eVect may vary at diVerent spatial extents. Therefore, the purpose of this study was to investigate the eVects of spatial heterogeneity on butterXy species richness at multiple spatial extents. Our speciWc objectives were to: (1) Quantify spatial heterogeneity at multi- ple spatial extents, and relate it to butterXy species richness; (2) investigate whether butter- Xies have strongest response to a particular spatial extent; (3) identify environmental variables that aVect butterXy species richness; (4) develop predictive models for butterXy species richness in the study area; and (5) Wnd out whether consideration of Wve compo- nents of landscape heterogeneity with traditionally used environmental variables improve the predictive power of the models.

Methods

Study area

The study was conducted in central portion of Rocky Mountain National Park, CO, USA, located between 40°10Ј–40°34ЈN latitude and 105°30Ј–105°55ЈW longitude, covers 1 C Biodivers Conserv (2009) 18:739–763 743

N

Rocky Mountain Legend National Park Sample plots

Study area

Estes Park Colorado

Denver

Fig. 1 Map of study area with sample plot locations

approximately 107,500 ha. The elevation in the area varies from 2,300 to over 4,300 m above mean sea level and lies on the Front Range of the southern Rocky Mountains in Colorado (Fig. 1). The average annual temperatures vary from ¡1.5 to 14.0°C. The mean annual precipitation is 36 cm. The temperature and precipitation have been suggested as strong determinants of latitudinal and elevation arrangements of species distributions, as typically inXuenced by elevation and topographic position (Peet 1981, 1988; Allen et al. 1991). A number of vegetation communities are found in the area varying from montane dry meadows to subalpine forests and alpine turnda. Dominant vegetation types and plant spe- cies include: dry meadow vegetation dominated by short prairie grasses (Bouteloua gracilis Vasey in Rothr., Buchloe dactyloides Engelm.) and sage brush (Artemisia tridentate Nutt.); ponderosa pine (Pinus ponderosa Douglas ex. C. Lawson; 2,320–3,170 m); douglas-Wr (Pseudotsuga menziesii (Mirb.) Franco; 2,370–3,213 m); lodgepole pine (Pinus contorta Doug. ex. Loud; 2,380–3,480 m); aspen (Populus tremuloides Mich.; 2,350–3,500 m); lim- ber pine (Pinus Xexilis James; 2,620–3,560 m); and spruce (Picea engelmannnii Perry ex. Engel.; 2,530–3,710 m) and subalpine Wr (Abies lasiocarpa (Hook.) Nutt.; 2,530–3,710 m; Peet 1988). We used a stratiWed-random sampling design to measure plant and butterXy species rich- ness patterns in both dominant and rare but important vegetation types. Montane wet meadow, riparian shrubland, and open rock outcrop areas are examples of less common types that were included as strata due to resource management concerns. Aspen areas are another example of a vegetation type that seems to support a disproportionate contribution to local species diversity relative to its areal extent, providing critical habitat for many spe- cies of plants, butterXies, and birds (Simonson et al. 2001). 1 C 744 Biodivers Conserv (2009) 18:739–763

ButterXy survey

We used a stratiWed-random sampling design to measure butterXy species richness in the study area. We recorded the presence of butterXy species from seventy-six 20 £ 50 m plots that were overlaid directly on the 0.1-ha modiWed-Whittaker vegetation plots (Stohlgren et al. 1998). We followed established guidelines for butterXy monitoring during the survey to minimize observer bias and any eVects of environmental variation (Pollard and Yates 1993). Observations were made during systematic walking surveys (44 min per plot visit) only in sunny conditions (temperature >17°C) and under calm to light winds. We restricted the surveys to the dates and times when weather conditions were favorable for butterXy Xights. Surveys were conducted between 09.30 and 16.00 h and repeated on two visits to each plot location during June, July, and August, with additional plots sampled each year in 1996, 1997, and 1998 (Simonson et al. 2001). We recorded a total of 68 butterXy species from the seventy-six plots, representing approximately 51% of the total species found in the study area (Appendix 1). Although the butterXy plot sampling occurred across three seasons, the surveys were carefully standard- ized with respect to sample eVort. We are conWdent that the butterXy species richness dataset represents a useful initial estimate of the relative butterXy species richness that occurs at these sample locations. Although additional butterXy species would almost certainly be documented on subsequent visits to these plot locations, we recorded 68 of the 134 butterXy species known from Rocky Mountain National Park in just two visits to each plot location.

Quantifying plot-level characteristics

Plot-level characteristics were represented by plant species richness, average vegetation height, and range in NDVI (Normalized DiVerence Vegetation Index). Plant species rich- ness data were collected from 20 £ 50 m (0.1-ha) modiWed-Whittaker, multi-scale plots with the long axis parallel to the environmental gradient (Stohlgren et al. 1998). These plots were located based on stratiWed random sampling in diVerent vegetation types. Each site was sampled as close to peak plant phenology and biomass as possible. Plant species that could not be identiWed in the Weld were collected and identiWed at the herbarium at Colorado State University (Department of Biology, Fort Collins, CO). Some of the speci- mens could not be identiWed to species due to phenological stage or missing Xower parts. In these cases, plants were identiWed to genus and treated as individual species. Plot locations were recorded using the global positioning system (GPS; Trimble Navigation Ltd., Westminster, CO, USA), and coordinates were taken using the universal transverse merca- tor (UTM) system, which provides x, y coordinates in meters from a regional reference point. The average vegetation height (cm) by species was recorded in the ten 1-m2 subplots in each modiWed-Whittaker plot (Stohlgren et al. 1998) using a measuring rod for grasses and shrubs and ocular estimates for trees. The range in NDVI, 250 m spatial resolution, around the center of the sample plot was extracted using a discrete Fourier transform (Moody and Johnson 2001) from the National Aeronautics and Space Administration’s (NASA) moderate resolution imaging spectroradi- ometer (MODIS) instrument data, and were provided by researchers at NASA’s Goddard Space Flight Center, Greenbelt, MD, USA. The MODIS data were acquired from February 2000 through February 2004; see Morisette et al. (2006) for more details. The range in NDVI contains information on vegetation as a measure of photosynthetic activity and 1 C Biodivers Conserv (2009) 18:739–763 745 vegetation productivity. It was used as a surrogate for heterogeneity in vegetation produc- tivity because higher vegetational heterogeneity may provide more resources for some but- terXy species (Bailey et al. 2004; Seto et al. 2004).

Quantifying spatial heterogeneity

Spatial heterogeneity was quantiWed in terms of elevation, slope, aspect, topographic expo- sure, and distance from stream/river; and landscape metrics (representing both composition and conWguration; McGarigal and Marks 1995). Selection of spatial scale(s) at which to quantify spatial heterogeneity is a daunting task because spatial heterogeneity is a highly scale-dependent phenomenon (Kolasa and Rollo 1991; Wagner and Fortin 2005). The spa- tial scale may refer to both ‘grain’ and ‘extent’ (Turner et al. 1989). In this study, we did not change the grain size (Wxed it to 30 £ 30 m); however, we varied the spatial extents of analyses. Most ecologists, who study single species, use home range size or territory size or area of an organism’s activity to select spatial extent(s) at which they quantify spatial heteroge- neity. However, for community level studies, such as this, the scale for quantifying spatial heterogeneity at which species/community will show strongest aVects is diYcult to know a priori because diVerent species may have diVerential responses to spatial heterogeneity depending on their dispersal abilities, life history characteristics, and habitat requirements. One of the ways to deal with this problem is to select multiple arbitrary spatial extents to quantify spatial heterogeneity in diVerent environmental variables. Next, identify the domi- nant scale of organism’s/community’s response by comparing the variation explained (coeYcient of determination; R2) at each spatial extent. This approach has been used in a number of studies on diVerent taxa such as birds (Warren et al. 2005), (Kie et al. 2002), plants (Kumar et al. 2006), and insects including butterXies (SteVan-Dewenter et al. 2002; Thies et al. 2003; Krauss et al. 2003; Bergman et al. 2004; Davis et al. 2007). In general, information about the dispersal or home ranges of most butterXy species is limited. With the exception of a few long-distance migrants such as monarchs (Danaus plexippus Linnaeus), most butterXies move small distances (up to a few hundred meters; Wahlberg et al. 2002), while a few species move much larger distances between 1.0 and 2.25 km (Wahlberg et al. 2002; see review by Schneider 2003). However, there is an uncer- tainty in these estimates because these movement distances are based on mark-release- recapture studies, which may be biased due to the extent of the study area (Schneider 2003). Moreover, the movement distances of diVerent butterXies may also vary spatially and temporally (Wahlberg et al. 2002). Based on this information we quantiWed spatial heterogeneity at eight radii of 300, 600, 900, 1,200, 1,500, 1,800, 2,100, and 2,400 m, rep- resenting a nested set of spatial extents. Corresponding areas were 28, 113, 255, 453, 707, 1,018, 1,386, and 1,810 ha, respectively. The topographic variables included elevation, slope, aspect, topographic exposure and distance from a stream or river. Digital elevation model (DEM) data for the area were downloaded from the National Elevation Dataset, United States Geological Survey (USGS) website, to sample elevation (m) around the center of each 0.1-ha plot. Subsequently, the DEM grid (30 £ 30 m) was used to generate slope (in degrees), and aspect (in degrees) using Environmental Systems Research Institute’s (ESRI) ARC GIS, version 9.0, surface analysis functions (ESRI 2004). For statistical analyses, the circular variable aspect was transformed into a linear north–south gradient (northness) and an east–west gradient (east- ness) by performing cosine and sine transformations, respectively (Kumar et al. 2006). A stream or river network shapeWle was acquired from the Colorado Department of Water 1 C 746 Biodivers Conserv (2009) 18:739–763

Resources. After converting it into a raster layer of 30 £ 30 m cell size, Euclidean distance around the center of each 0.1-ha plot from the nearest stream or river was calculated using Map Calculator in Arc Map of ARC GIS version 9.0 (ESRI 2004). All of these topographic variables were quantiWed at eight spatial extents (radii of 300, 600, 900 to 2,400 m) around the center of the sample plots, and their summary statistics (mean, minimum, maximum and standard deviation) were calculated (Appendix 2). In all analyses, an extended area, 3 km outside the park boundary, was included to account for the edge eVects. Landscape metrics representing both landscape composition and conWguration, and Wve components of landscape heterogeneity (Li and Reynolds 1994), were quantiWed using the raster version of the FRAGSTATS (version 3.3) landscape pattern analysis program (McGarigal and Marks 1995). National Land Cover Dataset (NLCD; Vogelmann et al. 1998) for the study area was obtained from US Geological Survey, and was used to deWne habitat patches. DeWning habitat patches was a fairly daunting task because the perception of the landscape may be very diVerent for various species and for entire assemblage of butterXies, and may even change through time for individual organisms. The butterXy species have diVerent mobility and food resources, and so their perception of patches could range from the scale of individual plant parts to broad landscapes. As larva they are associated with host plants and plant patches and as adults they may move great distances in search of mates and nectar resources. Therefore we chose to group diVerent vegetation types to deWne habitat patches, based on our observations of butterXies in the Weld and the structural characteristics of vegetation that occur in our area. For example, we combined broad categories of conifer spe- cies from the NLCD to a single ‘conifer’ type category based on similar vegetation structure. Thus, the original land cover map was reclassiWed into Wve broad land cover types: coni- fer, deciduous, grasslands, shrubland, and non-vegetated (bare rocks, water, and ice/snow). The land cover map for the study area was clipped with an extended area, 3 km outside the park boundary, to account for the edge eVects in landscape metrics calculations (McGarigal and Marks 1995). This map in raster format (ESRI GRID; projection: UTM, Zone 13, datum: NAD 1983, and cell size: 30 m) was used as a basic input data layer for calculating landscape metrics (e.g., proportion of each land cover type, mean patch size, patch richness density, edge density, mean edge contrast index, cohesion, interspersion and juxtaposition index, Euclidean nearest neighbor distance, and Simpson diversity index; Appendix 2) at eight spatial extents (radii of 300, 600, 900–2,400 m). We also calculated “area weighted mean patch size,” at both landscape level and class level (Appendix 2), because it contains information about the number of patches and patch size, and is considered more robust (Li and Archer 1997) than “mean patch size” alone. These landscape metrics represented Wve components of landscape heterogeneity (Li and Reynolds 1994); and were chosen based on their potential biological relevance to butterXy species richness and their use in a number of previous studies on butterXies and other taxa (e.g., Mazerolle and Villard 1999; Kie et al. 2002; Collinge et al. 2003; Thies et al. 2003; Krauss et al. 2003; Kumar et al. 2006; Davis et al. 2007). The detailed information (deWnitions and formulas) about these metrics is given elsewhere (McGarigal and Marks 1995).

Statistical analyses

Multiple regression analyses were conducted using butterXy species richness as response variable and the plot-level characteristics and landscape context variables as predictors. Since eight nested spatial extents were not independent of each other, separate models were developed at each spatial extent. All the variables were tested for normality and the strongly skewed variables were transformed prior to analyses. For example, butterXy 1 C Biodivers Conserv (2009) 18:739–763 747 species richness and plant species richness data were square root transformed. Pearson correlation coeYcient (r) was used to investigate associations between butterXy species richness and measures of spatial heterogeneity at eight spatial extents. Prior to regression analyses, all the predictors were tested for multicollinearity (Neter et al. 1996) by examining cross-correlations among them (e.g., see Appendix 3). Only one variable from a set of highly cross-correlated (correlation coeYcient > § 0.75) variables was included in the regression models. The biological relevance of each predictor variable to butterXy species richness, and the ease of interpretation were used as criteria to select or drop the highly cross-correlated predictors. For example, at 2,100 m spatial extent, cohe- sion, edge density, mean patch size, Shannon-diversity index and Simpson-diversity index were highly cross-correlated; we excluded others and included mean patch size because it is easily understood and interpreted. We conducted stepwise forward multiple regressions to eliminate insigniWcant predictors (P > 0.05). In all analyses, residuals were examined for deviations from normality and homogeneity of variances. Regression analyses were conducted using the PROC REG procedure in SAS software (SAS Institute 2004) and alpha = 0.05 was used to determine signiWcance level in all cases. We used Akaike’s Information Criteria adjusted for small sample size (AICc) and the information-theoretic approach (Burnham and Anderson 2002) to evaluate multiple regres- sion models and select the “best” model for butterXy species richness from a set of candi- V date models developed at eight spatial extents. The models with a di erence of AICc <2 were considered competing models (Burnham and Anderson 2002). Residuals from the best regression model were tested for spatial autocorrelation using Moran’s I coeYcient of spatial autocorrelation (Moran 1948). All spatial statistical analyses were performed using S-Plus (version 7.0) statistical software (Insightful Corp., Seattle, Washington).

Results

ButterXies’ responses to spatial heterogeneity

There was a positive relationship between butterXy species richness and plant species rich- ness (r =0.42, P < 0.0001; Fig. 2a) in our study area. We found highly signiWcant negative relationships between butterXy species richness and average vegetation height (r = ¡0.36, P < 0.0001; Fig. 2b); elevation (plot-level) (r = ¡0.42, P < 0.0001; Fig. 2c); and range in NDVI (r = ¡0.36, P = 0.001; Fig. 2d). ButterXy species richness was signiWcantly nega- tively correlated with mean patch size at all spatial extents (Table 1; Fig. 3a). However, there were strong positive relationships between butterXy species richness and mean patch size of shrubland, and the proportion of shrubland at all spatial extents (Fig. 3b; Table 1). Also, there were positive correlations between butterXy species richness and proportion of grasslands, and proportion of deciduous land cover type (Table 1). In contrast, butterXy species richness was mostly negatively correlated with mean patch size of conifer type at all spatial extents (Table 1). No signiWcant relationships were detected between butterXy species richness and mean patch area of non-vegetated and proportion of non-vegetated land cover type at any of the spatial extents. Overall, across eight spatial extents, diVerent measures of spatial heterogeneity showed complex patterns of relationships with butterXy species richness (Table 1). Some of them were consistently positively correlated with butterXy species richness (e.g., pro- portion of shrubland, mean patch size of shrubland, and edge density; Table 1) and some

1 C 748 Biodivers Conserv (2009) 18:739–763

5.0 5.0 y = -0.82x + 3.38 y = 0.20x + 1.52 (b) (a) r = -0.36, P = 0.002 r = 0.42, P < 0.0001 4.0 4.0

3.0 3.0

2.0 2.0

1.0 1.0

0.0 0.0 2 4 6 8 10 0.0 0.5 1.0 1.5 SQRT (Total plant species richness) Log (Average vegetation height, m) 10

5.0 y = -0.002x + 7.89 5.0 y = -0.0004x + 3.46 tterfly species richness) (c) u r = -0.42, P < 0.0001 (d) r = -0.36, P = 0.001 4.0 4.0

3.0 3.0 SQRT (B

2.0 2.0

1.0 1.0

0.0 0.0 2300 2500 2700 2900 3100 3300 0 1000 2000 3000 4000 Elevation (m) NDVI range

Fig. 2 Relationship between butterXy species richness (SQRT, square root transformed) and plot-level (0.1- ha). a plant species richness (SQRT); b average vegetation height (m; (log10+1) transformed); c elevation (m); and d range in NDVI (250 £ 250 m)

Table 1 Scale-dependent eVects of spatial heterogeneity in topography and landscape patterns on butterXy species richness

Predictor variable (units) Spatial extent (m)

300 m 600 m 900 m 1,200 m 1,500 m 1,800 m 2,100 m 2,400 m

Mean patch size (ha) ¡0.47 ¡0.34 ¡0.35 ¡0.31 ¡0.32 ¡0.35 ¡0.35 ¡0.37 Mean patch size of shrubland (ha) 0.34 0.27 0.27 0.27 0.28 0.28 0.23 0.14 Mean patch size of conifer (ha) ¡0.45 ¡0.29 ¡0.27 ¡0.17 ¡0.18 ¡0.24 ¡0.22 ¡0.21 Mean patch size of deciduous (ha) 0.10ns ¡0.04ns ¡0.24 ¡0.26 ¡0.28 ¡0.36 ¡0.30 ¡0.29 Edge density (m/ha) 0.40 0.33 0.35 0.33 0.31 0.31 0.32 0.31 Proportion of shrubland (%) 0.40 0.38 0.30 0.33 0.34 0.33 0.31 0.29 Proportion of grasslands (%) 0.26 0.23 0.19 0.17 0.17 0.24 0.18 0.25 Proportion of deciduous (%) 0.33 0.23 0.08ns ¡0.06ns ¡0.05ns 0.01ns 0.07ns 0.12 Mean elevation (m) ¡0.39 ¡0.36 ¡0.34 ¡0.32 ¡0.30 ¡0.28 ¡0.28 ¡0.28 Mean topographic exposure (m) ¡0.24 ¡0.29 ¡0.30 ¡0.30 ¡0.30 ¡0.30 ¡0.25 ¡0.28

Pearson correlation coeYcients (r) between butterXy species richness (square root transformed) and select predictor variables at eight spatial extents are shown Note: Correlations were signiWcant at P = 0.05, except where noted; “ns” indicates nonsigniWcance at P >0.05; the P values were adjusted using progressive Bonferroni correction (Legendre and Legendre 1998). Transformed data were used where it was appropriate

1 C Biodivers Conserv (2009) 18:739–763 749

Fig. 3 Relationship between 5.0 y = -0.83x + 3.30 butterXy species richness (SQRT, (a) square root transformed) and (a) r = -0.47, P < 0.0001 mean patch size (ha; log10 trans- 4.0 formed); b proportion of shrub- land (log10 transformed); and (c) average distance from stream/ 3.0 river (m)

2.0

1.0 300 m extent 0.0 0.0 0.5 1.0 1.5 Mean patch size (ha)

5.0 (b) y = 0.44x + 2.45 r = 0.40, P < 0.0001 4.0

3.0

2.0 tterfly species richness) u 1.0 300 m extent SQRT (B 0.0 0.0 0.5 1.0 1.5 2.0 Shrubland (%)

5.0 (c) y = 0.002x + 1.80 r = 0.41, P < 0.0001 4.0

3.0

2.0

1.0 2400 m extent 0.0 200 300 400 500 600 700 800 900 Average distance from stream/river (m)

1 C 750 Biodivers Conserv (2009) 18:739–763 were consistently negatively correlated (e.g., mean patch size, mean elevation, and mean exposure; Table 1).

Predictive models of butterXy species richness

The best model explained 62% of the variation in butterXy species richness at 2,100 m spatial extent and included plot-level characteristics and variables representing spatial heterogeneity in topography and landscape patterns (Table 2, and model 1; Table 3). This model was highly signiWcant, and average vegetation height (partial R2 = 0.20), mean patch size (partial R2 = 0.17), and plant species richness (partial R2 = 0.15) were the three best predictors of butterXy species richness (Table 2). Patch richness density, interspersion and juxtaposition index, and standard deviation of elevation were three other predictors of butterXy species richness (Tables 2, 3). Six other models explained 29–59% of the variation in butterXy species richness (models 2–7; Table 3). The models that included only landscape (model 7; Table 3) or only topographic (model 6; Table 3) heterogeneity variables as predictors performed worse  ( AICc = 43.21 and 43.85, respectively) than the model that included only plot-level vege- tation characteristics (model 4; Table 3). However, the models based on the combinations of predictor variables representing plot-level characteristics, topography and landscape pat- terns performed better than the models based only on predictors from one type of variable (models 2–3, except model 5; Table 3). For example, the model based on the combination of vegetation and landscape variables was the second best model (model 2; Table 3) and explained 59% of the variation in butterXy species richness (Table 3). The third best model (model 3; Table 3) explained 54% of the variation in butterXy species richness and included vegetation and topographic variables. It is clear from the model evaluation exer- cise (Table 3) that considering landscape metrics in addition to other types of variables improved the models signiWcantly (Table 3). Spatial autocorrelation analyses using Moran’s I for both butterXy species richness (raw data) and the best model residuals indicated that there was a signiWcant positive spatial autocorrelation (I = 0.121, P < 0.0001) in the raw data, but no signiWcant spatial autocorre- lation in the model residuals (I = ¡0.016, P = 0.997), which indicates that the best regres- sion model (model 1; Table 3) was eVective in modeling spatial autocorrelation in the raw data. This may be due to the fact that spatial autocorrelation in the raw data was explained

Table 2 The best regression model linking patterns of butterXy species richness (n = 76, 0.1 ha) with plot- level characteristics and landscape context (2,100 m extent) variables

Dependent variable Independent variable Parameter Factor P Partial R2 Model adj. (spatial extent) estimate R2, F, P

(ButterXy SR)0.5 Average vegetation height ¡0.047 <0.0001 0.197 R2 =0.62 Mean patch size ¡0.243 <0.0001 0.169 F6, 69 = 21.57 Plant species richness 0.161 <0.0001 0.147 P < 0.0001 Patch richness density ¡10.873 <0.0001 0.055 Interspersion and ¡0.019 0.0020 0.047 juxtaposition index Elevation (standard deviation) 0.004 0.0085 0.037

Note: Average vegetation height (m) and plant species richness data were collected at plot-level (0.1 ha). Other predictors in the models were quantiWed at 2,100 m spatial extents. The variables were transformed where it was appropriate. SR is species richness 1 C Biodivers Conserv (2009) 18:739–763 751

Table 3 Regression models evaluated for butterXy species richness (n = 76, 2,100 m extent) X 2  Model No. Type of variable Predictors of butter y species Adj. R K AICC AICC richness (sign of regression coeYcient)

1 Landscape MPS (¡), AvgHt (¡), 0.62 7 ¡128.97 0.00 + topographic PRD (¡), Plant SR (+), + vegetation IJI (¡), ElevSD (+) 2 Landscape MPS (¡), AvgHt (¡), 0.59 6 ¡123.93 5.25 + vegetation PRD (¡), Plant SR (+), IJI (¡), 3 Topographic AvgHt (¡), Plant SR (+), 0.54 6 ¡115.88 13.09 + vegetation RngNDVI (¡), ExpSD (-), StrmDist (+) 4 Vegetation RngNDVI (¡), AvgHt (¡), 0.46 4 ¡105.33 23.64 Plant SR (+) 5 Landscape ExpSD (¡), ElevMax (+), 0.38 6 ¡92.81 36.15 + topographic EdgeCont (¡), StrmDist (+), ENN (¡) 6 Topographic ExpMax (¡), ElevMax (¡), 0.30 4 ¡85.76 43.21 StrmDist (+) 7 Landscape MPS (¡), PRD (¡), IJI (¡)0.294¡85.12 43.85

Notes: MPS mean patch size; AvgHt average vegetation height; PRD patch richness density; SR species rich- ness; IJI interspersion/ juxtaposition index; ElevSD elevation standard deviation; RngNDVI range in Normal- ized DiVerence Vegetation Index (NDVI); ExpSD exposure standard deviation; StrmDist average distance from the stream or river; ElevMax maximum elevation; EdgeCont mean edge contrast index; ENN nearest neighbor distance (m); ExpMax maximum exposure; AICc Akaike’s Information criterion adjusted for small sample size; K is the number of estimable parameters in the model. The variables were transformed where it was appropriate. The predictor variables are arranged in order of their decreasing importance (based on stan- dardized coeYcient) by the signiWcant spatial autocorrelation observed in some of the independent variables in the best model (Legendre and Legendre 1998).

Responses of butterXies to spatial heterogeneity at multiple scales

The consideration of landscape context with plot-level characteristics always improved the variation explained in butterXy species richness (Fig. 4; notice the change in adjusted R2 values on the left and right sides of the vertical dotted line). The response of butterXy species richness was strongest (adj. R2 = 0.62) at 2,100 m spatial extent (Fig. 4). However, across the scales, no continuous trend in the amount of variation explained in butterXy species richness was observed (Fig. 4). It may be because some of the butterXy species in the area are responding to small scale spatial heterogeneity (adj. R2 = 0.60; 300 m), while others respond to intermediate (adj. R2 = 0.60; 1,200 m) or larger scale (adj. R2 = 0.62; 2,100 m) spatial heterogeneity.

Discussion

In this comparative model evaluation study we used Weld data on plant and butterXy species to show that our understanding of patterns in butterXy species richness can be improved by considering spatial heterogeneity, and that the importance of diVerent land- scape characteristics varies at diVerent spatial extents. We also showed that including

1 C 752 Biodivers Conserv (2009) 18:739–763

0.65 Plot Plot Landscape + level level context 0.60 2 0.55 sted R u 0.50 Adj

0.45

0.40 300 600 900 1200 1500 1800 2100 2400 Spatial extent (radius, m)

Fig. 4 Variation explained in butterXy species richness by plot-level characteristics, and plot-level charac- teristics plus landscape context variables (quantiWed at diVerent spatial extents)

landscape context with plot-level characteristics advances our understanding of the vari- ation in butterXy species richness across the landscape. This is consistent with several studies on butterXies that have related diVerent types of spatial heterogeneity to butterXy species diversity (e.g., Kremen 1992; Wettstein and Schmid 1999; Fleishman et al. 2002; Summerville et al. 2002; Collinge et al. 2003; Krauss et al. 2003; Stefanescu et al. 2004; Strathmann 2005; Luoto et al. 2006; Davis et al. 2007). However, our models contained much more information on the diVerent aspects of spatial heterogeneity, and thus, were more explanatory than previous studies. Our best model explained 62% of the variation in butterXy species richness at 2,100 m spatial extent (Tables 2, 3) and included predic- tors representing all three categories of spatial heterogeneity. Addition of landscape metrics with other measures of spatial heterogeneity always helped improve the predic- tive models (Table 3).

EVects of plot-level characteristics on butterXy species richness

We found a strong positive relationship between butterXy species richness and plant species richness (Fig. 2a). These results are consistent with previous studies on butterXies (e.g., Thomas and Mallorie 1985; Simonson et al. 2001; Fleishman et al. 2005). This posi- tive association may be because plant species serve as important food plants for larvae, Xowers provide nectar resources for adult butterXies, and the physical structure of plants can create microhabitats, providing shelter to butterXies (Wood and Samways 1991). High plant species diversity may be an indicator of better habitat quality for butterXy species, because the presence of a variety of plants would be more likely to fulWll the host plant and habitat requirements of both specialist and generalist butterXy species. ButterXy species richness was negatively associated with average vegetation height (Fig. 2b). This suggests that taller vegetation can act as a barrier to butterXy move- ments. In our study area, the majority of the sample plots that showed higher butterXy species richness had an average vegetation height less than 5 m (Fig. 2b). Shading by tall trees may limit solar radiation needed for thermoregulation, and shady forested

1 C Biodivers Conserv (2009) 18:739–763 753 areas may not support required host plant food resources. This pattern is consistent with the results of earlier studies (e.g., Thomas and Mallorie 1985; Wettstein and Schmid 1999) that reported negative relationships between butterXy diversity and vegetation height. However, positive relationships between butterXy diversity and average plant height have also been reported (e.g., Dennis 2004; Strathmann 2005). This discrepancy in relationships may be explained by the range of variation in vegetation height in the study area. For example, our study area included more variation in vegetation height (range 0.29–30.41 m) than the study conducted by Strathmann (2005) (range 0.10–0.37 m) and Dennis (2004). ButterXies are often observed perching on or Xying near moderately high vegetation (e.g., grasslands and shrubland land cover types), and it has been suggested that they may avoid completely Xat or bare grounds (Wood and Samways 1991; Dennis 2004) or higher vegetation (e.g., conifer and deciduous cover types). This could be related to the energetic costs and the high risk of predation that may be associated with higher Xights. However, it should be noted that the highly mobile adult butterXies are also more likely to be detected by an observer when they are patrolling in Xight or perched on vegetation at (human) eye level. We found a negative relationship between range in NDVI and butterXy species richness (Fig. 2d). This could be because our study area is dominated by conifer land cover type which is relatively more productive than the other land cover types. This pattern suggests that the conifer type may act as a matrix for butterXies’ movements, with the relatively tall vegetation eVectively isolating the habitat patches that may be more suitable for butterXies (Ricketts 2001).

EVects of landscape context on butterXy species richness

ButterXy species richness was negatively correlated with elevation (Fig. 2c). Insect spe- cies richness patterns may decrease, increase, or show no trends relative to elevational gradients, depending on the biology of the insect species and the range of variation in the altitude (Hodkinson 2005). Some researchers have reported positive relationships between elevation and butterXy diversity (e.g., Wettstein and Schmid 1999; Stefanescu et al. 2004; Strathmann 2005), while others have reported negative relationships (e.g., Fleishman et al. 1998). These contrasting relationships may be because of the range in elevation at the study location and regional climate. For example, the elevation at the sample plots in our study varied from 2,400 to 3,200 m, whereas the elevation in other studies was less than 2,400 m (e.g., 800–1,400; Wettstein and Schmid 1999, and 850–2,300 m; Strathmann 2005). The negative relationship observed in our study may be because of the decrease in temperature with an increase in elevation in the area that in turn may be aVecting butterXy species richness, as butterXies are ectothermic organisms that exhibit behavioral thermoregulation, such as sun-basking, to maintain Xight activity in lower temperatures (Clench 1966). It may also be due to a decrease in the amount of time that conditions are suitable for Xight at higher elevations (e.g., studies of Colias philodice Godart by Kingsolver 1983a, b). It may be that butterXy species richness is both positively correlated with elevation at lower elevations and negatively associated with elevation at higher elevations. For exam- ple, in areas at lower elevations adjacent to our study area (<2,300 m) the complex topography provides an impressive diversity of butterXy habitats. In contrast, on the peaks and ridges at the upper elevations of the study area, harsh temperatures and short growing 1 C 754 Biodivers Conserv (2009) 18:739–763 seasons may limit distributions of all but the most specialized alpine butterXy species. ButterXies showed a positive response to distance from stream/river at higher spatial extents (e.g., Fig. 3c), which may be because at larger distances from streams or rivers, there is a high probability of Wnding ridges or hill tops in mountainous areas, which are frequently used by the butterXy species for diVerent activities such as basking and mate locating (Scott 1975). Mean patch size was negatively correlated with butterXy species richness (Table 1) and was one of the best predictors of butterXy species richness (Tables 2, 3) which indi- cates that butterXies in the area may tend to use relatively smaller patches. We suggest the following explanations for this pattern: (1) butterXies may prefer edge habitats, which can provide suitable microhabitats and resources for thermoregulation (Dennis et al. 2006), and many small patches will have more edges than a large patch of the same area; (2) several small patches may support high densities of host plants (Nieminen et al. 2004), and may be more dissimilar in plant species composition than a large patch with same area and may provide diverse resources for butterXies; and (3) this could also be related to the dominance of the conifer land cover type in the area that has relatively larger patches and taller vegetation structure which might be acting as a barrier to but- terXy movements (Ricketts 2001). ButterXy species richness was strongly positively correlated with the proportion of shrubland and mean patch size of shrubland at all spatial extents (Table 1). This could be because shrublands may provide better quality habitats and contain relatively higher plant species diversity or a greater number of important functional groups; and larger patches of shrublands may have more heterogeneous resources (e.g., host plants, nectar plants, ovipo- sition sites and microhabitats) than the smaller patches (Thomas and Hanski 1997; Golden and Crist 1999). This pattern is supported by previous studies that reported associations of butterXies with shrub vegetation (e.g., Bergman et al. 2004; Dennis 2004). Positive associ- ations between butterXy species richness and the proportion of deciduous land cover type at smaller spatial extents (Table 1) are consistent with our observations of the high numbers of butterXy species recorded in habitats dominated by deciduous plant species such as wil- lows (Salix spp. L.), and aspen, a dominant tree in deciduous land cover type (Simonson et al. 2001). Aspen is a suitable larval host plant for several butterXy species recorded in the study area (Scott 1986; Opler 1999), including Papilio rutulus Lucas, Erynnis icelus (Scudder and Burgess), Limenitis weidemeyerii (W.H. Edwards), and Nymphalis antiopa (Linnaeus). ButterXy species richness was positively correlated with edge density (Table 1). This may be because edges may support a greater abundance and diversity of host and nectar plants. For example, Kumar et al. (2006) found a positive relationship between plant spe- cies richness and the edge density in the same study area. Edges may also have diVerent microclimatic conditions than the interior of a patch and may provide warmth and shelter for butterXies (Wood and Samways 1991). Higher butterXy species richness on the edges may also be attributed to the tendency of some female butterXies to oviposit on the plants protruding from the surrounding vegetation patches with relatively taller vegetation (Warren 1984; Dennis 2004).

Multi-scale responses of butterXies to spatial heterogeneity

We did not Wnd a single dominant spatial extent at which butterXy species richness strongly responded to spatial heterogeneity; no consistent trend was observed (Fig. 4). This lack of a consistent trend in the community-level responses to spatial heterogeneity may be due to 1 C Biodivers Conserv (2009) 18:739–763 755 combined result of individual-species responses at multiple spatial extents that could be attributed to (1) their varying dispersal abilities or foraging ranges (Ehrlich 1961; Wahlberg et al. 2002; Schneider 2003); (2) spatial distribution of their host and larval plants; (3) their ability to use the matrix (non-vegetated or conifer areas in our study area); (4) specialist versus generalist species (Krauss et al. 2003); (5) their ability to cross structural barriers created by vegetation height [e.g., low-Xying versus high-Xying species; Summerville et al. (2002)]; (6) composition and conWguration of the landscape; and (7) range in elevation and coarseness of the environmental variation in the study area. However, butterXies’ responses to spatial heterogeneity at multiple spatial extents in relatively homogenous landscapes with lower topographic and landscape heterogeneity (e.g., grasslands or grasslands mixed with agricultural matrix) may be more consistent than we found (e.g., Krauss et al. 2003; Bergman et al. 2004; Davis et al. 2007).

Limitations and important caveats

Our model results highlight the importance of considering landscape spatial heterogeneity in understanding species diversity patterns, but further research is needed to attribute cau- sation to the observed occurrences of plants and butterXies. It may not be possible to directly use the predictive models presented in this study for species that occur in other landscapes because they may have varying degrees of spatial heterogeneity in topography and landscape patterns. For example, the models may not apply in areas of lower topo- graphic variability than Rocky Mountain National Park, areas with varying degrees of patchiness in landscape (e.g., grasslands mixed with shrubland versus grasslands mixed with forested areas), or diVerent disturbance regimes (natural vs. anthropogenic). However, a similar modeling approach could be used for other smaller or larger landscapes. The unexplained variation in butterXy species richness suggests that additional variables may need to be considered in other predictive models. For example, including local distribution data on larval and host plants, or irradiance may have improved the models. Also, the smaller resource patches (<900 m2) were not considered in this study because of the limita- tion of the spatial resolution (30 £ 30 m) of the vegetation map used in landscape analyses. Therefore, consideration of small resource patches by including vegetation maps generated using Wner resolution remotely sensed data in landscape analyses may improve the predic- tive power of the models. Average vegetation height in this study was measured at sample plot level (20 £ 50 m), however, landscape level vegetation height maps that can be cre- ated using the Shuttle Radar Topography Mission (SRTM) elevation data and lidar (light detection and ranging) remote sensing technique (e.g., Simard et al. 2006), have the poten- tial to improve butterXy species richness predictive models. For practical reasons, we developed butterXy species richness predictive models at vary- ing spatial extents and Wxed the grain size to 30 m. However, this range may not include the actual scale of perception of the individual butterXies through time. Measures of landscape heterogeneity (i.e., landscape metrics) are also sensitive to spatial (grain size) and thematic (number of categories) resolution and accuracy of the input land use map (Benson and MacKenzie 1995; O’ Neill et al. 1996; Baldwin et al. 2004; Shao and Wu 2008). The scale at which butterXy species richness data were collected can also aVect the predictive models. For example, the models would vary if the data were collected using a transect method rather than the 20 £ 50 plots we used. Therefore, further research is required to investigate how these models change with changing grain size, thematic resolution of the input map, and sampling methods.

1 C 756 Biodivers Conserv (2009) 18:739–763

Management and conservation implications

Our results suggest that spatial heterogeneity can greatly inXuence patterns in butterXy species richness across the landscape and highlights the need for understanding the eVects of spatial heterogeneity on biodiversity. The study also illustrates that land man- agers should maintain and consider not only the spatial heterogeneity in the management unit but also wider landscape context—what is happening around the area of interest is also important (Collinge et al. 2003; Krauss et al. 2003; Thies et al. 2003; Kumar et al. 2006; Davis et al. 2007). It becomes even more important when the area is prone to fre- quent disturbances. For example, a change in spatial heterogeneity due to management practices or disturbances such as Wre, grazing, or logging that can create large homoge- neous areas which may result in a decline in butterXy diversity. Knowledge about the spatial scale of the responses of diVerent species or communities to spatial heterogeneity may make biodiversity management more eVective because a mismatch in the spatial scale of management (Cumming et al. 2006) and the spatial scale of an organism’s per- ception may result in failure of a management strategy. This study, including several oth- ers (e.g., Mazerolle and Villard 1999; Fleishman et al. 2002; Collinge et al. 2003; Kumar et al. 2006; Davis et al. 2007), suggest that including measures of landscape heterogene- ity (i.e. landscape metrics) with traditionally used environmental covariates to predict butterXy species distributions across the landscape can provide ecologists with an addi- tional explanatory power. Therefore, we suggest that in addition to spatial heterogeneity in topography, climate, and soil, ecologists should explicitly consider measures of landscape heterogeneity in biodiversity studies; if possible all Wve components (Li and Reynolds 1994). Our results also suggest that conservation and management actions targeted toward biodiversity conservation eVorts should explicitly consider composition and conWguration of the landscape.

Acknowledgments We thank two anonymous reviewers whose comments greatly improved the manu- script. The data used in this study were collected by a number of Weld technicians assisted by many excellent taxonomists from the Natural Resource Ecology Laboratory at Colorado State University, the Department of Bioagricultural Sciences and Pest Management at Colorado State University, and the US Geological Survey Fort Collins Science Center. We thank staV at Rocky Mountain National Park for providing administrative and logistical support during the study. We also thank Paul Opler, Geneva Chong, Greg Newman, Richard Bray, and Catherine Jarnevich, and others for their help during the study. T. Stohlgren acknowledges funding for data analysis from NASA grant (NRA-03-OES-03). S. Kumar acknowledges the support of the Ford Foundation International Fellowships Program for Ph.D. studies at Colorado State University (Aug. 2002– Aug. 2005). To all we are grateful.

Appendix 1

List of butterXy species recorded on two visits to seventy-six 20 £ 50 m plots in Rocky Mountain National Park, CO. This Weld survey list includes 68 of the 134 butterXy species that are known to occur in the park (available via DIALOG. http://www.npwrc.usgs.gov/ resource/insects/insect/rmnp.htm. Cited 5th November 2008). This list follows the taxonomic order of the “ButterXies of North America: ScientiWc Names List for ButterXy Species of North America, north of Mexico,” by Opler and Warren (2003).

1 C Biodivers Conserv (2009) 18:739–763 757

Hesperiidae: Skipper family

(Spreadwing Skippers, Skipperlings, and Grass Skippers) Dreamy Duskywing Erynnis icelus (Scudder and Burgess) Persius Duskywing Erynnis persius (Scudder) Common Checkered-Skipper Pyrgus communis (Grote) Russet Skipperling Piruna pirus (W.H. Edwards) Garita Skipperling Oarisma garita (Reakirt) Western Branded Skipper Hesperia colorado (Scudder) Nevada Skipper Hesperia nevada (Scudder) Draco Skipper Polites draco (W.H. Edwards) Woodland Skipper Ochlodes sylvanoides (Boisduval) Snow’s Skipper Paratrytone snowi (W.H. Edwards) Common Roadside-Skipper Amblyscirtes vialis (W.H. Edwards)

Papilionidae: Swallowtail family

(Parnassians and Swallowtails) Rocky Mountain Parnassian Parnassius smintheus Doubleday Anise Swallowtail Papilio zelicaon Lucas Black Swallowtail Papilio polyxenes Fabricus Western Tiger Swallowtail Papilio rutulus Lucas Pale Swallowtail Papilio eurymedon Lucas Two-tailed Swallowtail Papilio multicaudata W.F. Kirby

Pieridae: White and Sulfur family

(Whites, Marbles, and Sulfurs) Pine White Neophasia menapia (C. Felder and R. Felder) Checkered White Pontia protodice (Boisduval and Leconte) Western White Pontia occidentalis (Reakirt) Spring White Pontia sisymbrii (Boisduval) Cabbage White Pieris rapae (Linnaeus) Margined White Pieris marginalis Scudder Large Marble Euchloe ausonides (Lucas) Clouded Sulfur Colias philodice Godart Orange Sulfur Colias eurytheme Boisduval Queen Alexandra’s Sulfur Colias alexandra W.H. Edwards Scudder’s Sulfur Colias scudderi Reakirt

Lycaenidae: Gossamer wing family

(Coppers, Hairstreaks, ElWns, and Blues) Ruddy Copper Lycaena rubidus (Behr) Blue Copper Lycaena heteronea Boisduval Purplish Copper Lycaena helloides (Boisduval) Thicket Hairstreak spinetorum (Hewitson) Western Pine ElWn Callophrys eryphon (Boisduval) Western Tailed-Blue Cupido amyntula (Boisduval) 1 C 758 Biodivers Conserv (2009) 18:739–763

Spring Azure Celastrina sidara (Clench) Silvery Blue Glaucopsyche lygdamus (Doubleday) Rocky Mountain Dotted-Blue Euphilotes ancilla (Barnes and McDunnough) Reakirt’s Blue Echinargus isola (Reakirt) Melissa Blue Plebejus melissa (W.H. Edwards) Greenish Blue Plebejus saepiolus (Boisduval) Boisduval’s Mountain Blue Plebejus icarioides (Boisduval) Lupine Blue Plebejus “lupini” near lutzi (Boisduval) Arctic Blue Plebejus glandon (de Prunner)

Nymphalidae: Brushfoot family

(Fritillaries, Checkerspots, Anglewings, Admirals, Wood Nymphs, and Arctics) Variegated Fritillary Euptoieta claudia (Cramer) Aphrodite Fritillary Speyeria aphrodite (Fabricius) Edwards’ Fritillary Speyeria edwardsii (Reakirt) Atlantis Fritillary Speyeria atlantis (W.H. Edwards) Northwestern Fritillary Speyeria hesperis (W.H. Edwards) Mormon Fritillary Speyeria mormonia (Boisduval) Silver-bordered Fritillary Clossiana selene (Denis and SchiVermüller) Arctic Fritillary Clossiana chariclea (Schneider) Dotted Checkerspot Poladryas minuta (W.H. Edwards) Northern Checkerspot Chlosyne palla (Boisduval) Gorgone Checkerspot Chlosyne gorgone (Hubner) Silvery Checkerspot Chlosyne nycteis (Doubleday and Hewitson) Northern Crescent Phyciodes cocyta (Cramer) Field Crescent Phyciodes pulchella (Boisduval) Green Comma Polygonia faunus (W.H. Edwards) Hoary Comma Polygonia gracilis (Grote and Robinson) Gray Comma Polygonia progne (Cramer) Milbert’s Tortoiseshell Aglais milberti (Godart) Mourning Cloak Nymphalis antiopa (Linnaeus) Red Admiral Vanessa atalanta (Linnaeus) Painted Lady Vanessa cardui (Linnaeus) Weidemeyer’s Admiral Limenitis weidemeyerii (W.H. Edwards) Common Ringlet Coenonympha tullia (Müller) Small Wood-Nymph Cercyonis oetus (Boisduval) Chryxus Arctic Oeneis chryxus (Doubleday)

Appendix 2

See Table 4.

1 C Biodivers Conserv (2009) 18:739–763 759

Table 4 Environmental variables considered for developing predictive models of total butterXy species rich- ness

Type of variable Variable (units) (spatial extent)

Vegetation Total plant species richness (plot level—0.1-ha) Average vegetation height (m) (plot level—0.1-ha) Range in NDVI (MODIS) (250 £ 250 m—6.25-ha); see Morisette et al. (2006) Topographic (All variables were calculated at eight spatial extents: 300, 600, 900, 1,200, 1,500, 1,800, 2,100, and 2,400 m radius around the center of the sample plot) Elevation (m)—mean, minimum, maximum and standard deviation Slope (degrees)—mean, minimum, maximum and standard deviation Eastness (calculated as: 100*sine (aspect in degrees); See Fleishman et al. 2002)— mean, minimum, maximum and standard deviation Northness (calculated as: 100*cosine (aspect in degrees); see Fleishman et al. 2002)—mean, minimum, maximum and standard deviation Mean topographic exposure (calculated as: elevation of the centroid cell—mean ele- vation of all cell with in a speciWed radius; see Fleishman et al. 2002)—mean, minimum, maximum and standard deviation Mean distance from stream/river (m) Landscape Metrics (All metrics were calculated at eight spatial extents: 300, 600, 900, 1,200, 1,500, 1,800, 2,100, and 2,400 m radius around the center of the sample plot); See McGarigal and Marks (1995) for deW- nitions and formulas Composition Proportion of conifer (%) Proportion of deciduous (%) Proportion of grasslands (%) Proportion of shrubland (%) Proportion of non-vegetated (%) ConWguration Mean patch size (ha) Mean patch size of each of the Wve land cover types (ha) Area weighted mean patch size (ha) Area weighted mean patch size of each of the Wve land cover types (ha) Patch richness density (no./100 ha) Interspersion and juxtaposition index (%) Cohesion Mean edge contrast index (%) Edge density (m/ha) Mean nearest neighbor distance (m) Shannon-diversity index Simpson diversity index Mean fractal dimension Mean shape index

Variables representing landscape composition and conWguration were calculated with FRAGSTATS 3.3 according to McGarigal and Marks (1995). See also (Fleishman et al. 2002; Morisette et al. 2006)

Appendix 3

See Table 5.

1 C 760 Biodivers Conserv (2009) 18:739–763

Table 5 Cross-correlations (Pearson correlation coeYcients) among predictor variables that were a part of the best model of total butterXy species richness at 2,100 m spatial extent (n = 76, 0.1-ha plots)

Variable Total Total Average Mean Patch IJI Elevation butterXy plant vegetation patch richness (SD) SR SR height size density

Total butterXy SR 1.0 Total plant SR 0.42 1.0 Average vegetation height ¡0.36 ¡0.25 1.0 Mean patch size ¡0.35 0.12ns ¡0.06ns 1.0 Patch richness density ¡0.11ns ¡0.17ns 0.07ns ¡0.37 1.0 IJI ¡0.06ns ¡0.18ns 0.05ns ¡0.41 ¡0.10ns 1.0 Elevation (SD) ¡0.01ns 0.02ns ¡0.06ns 0.50 ¡0.16ns ¡0.09ns 1.0

Notes: SR is species richness; IJI is interspersion and juxtaposition index; SD is standard deviation. Trans- formed data were used where it was appropriate. Correlations are signiWcant at alpha = 0.05, except where noted; “ns” is the nonsigniWcance at alpha >0.05

References

Allen RB, Peet RK, Baker WL (1991) Gradient analysis of latitudinal variation in southern Rocky-Mountain forests. J Biogeogr 18:123–139. doi:10.2307/2845287 Bailey SA, Horner-Devine MC, Luck G et al (2004) Primary productivity and species richness: relationships among functional guilds, residency groups and vagility classes at multiple spatial scales. Ecography 27:207– 217. doi:10.1111/j.0906-7590.2004.03631.x Baldwin DJB, Weaver K, Schnekenburger F et al (2004) Sensitivity of landscape pattern indices to input data characteristics on real landscapes: implications for their use in natural disturbance emulation. Landsc Ecol 19:255–271. doi:10.1023/B:LAND.0000030442.96122.ef Benson BJ, Mackenzie MD (1995) EVects of sensor spatial-resolution on landscape structure parameters. Landsc Ecol 10:113–120. doi:10.1007/BF00153828 Bergman KO, Askling J, Ekberg O et al (2004) Landscape eVects on butterXy assemblages in an agricultural region. Ecography 27:619–628. doi:10.1111/j.0906-7590.2004.03906.x Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theo- retic approach. Springer, New York Clench HK (1966) Behavioral thermoregulation in butterXies. Ecology 47:1021–1034. doi:10.2307/1935649 Collinge SK, Prudic KL, Oliver JC (2003) EVects of local habitat characteristics and landscape context on grassland butterXy diversity. Conserv Biol 17:178–187. doi:10.1046/j.1523-1739.2003.01315.x Cumming GS, Cumming DHM, Redman CL (2006) Scale mismatches in social-ecological systems: causes, consequences, and solutions. Ecol Soc 11:14. Available via DIALOG. http:www.ecologyandsoci- ety.org/vol11/iss1/art14/. Accessed 5 November 2008 Davis JD, Debinski DM, Danielson BJ (2007) Local and landscape eVects on the butterXy community in frag- mented Midwest USA prairie habitats. Landsc Ecol 22:1341–1354. doi:10.1007/s10980-007-9111-9 Dennis RLH (2004) ButterXy habitats, broad-scale biotope aYliations, and structural exploitation of vegetation at Wner scales: the matrix revisited. Ecol Entomol 29:744–752. doi:10.1111/j.0307-6946.2004.00646.x Dennis RLH, Shreeve TG, Van Dyck H (2006) Habitats and resources: the need for a resource-based deWni- tion to conserve butterXies. Biodivers Conserv 15:1943–1966. doi:10.1007/s10531-005-4314-3 Ehrlich PR (1961) Intrinsic barriers to dispersal in Checkerspot butterXy. Science 134:108–109. doi:10.1126/ science.134.3472.108 Ehrlich PR, Raven PH (1964) ButterXies and plants: a study in coevolution. Evol Int J Org Evol 18:586–608. doi:10.2307/2406212 ESRI (2004) Environmental Systems Research Institute, Inc., 380 New York St., Redlands, CA 97373, USA Fleishman E, Austin GT, Weiss AD (1998) An empirical test of Rapoport’s rule: elevational gradients in montane butterXy communities. Ecology 79:2482–2493 Fleishman E, Murphy DD, Sjogren-Gulve P (2002) Modeling species richness and habitat suitability for taxa of conservation interest. In: Scott JM, Heglund PJ, Morrison ML, HauXer JB, Raphael MG, Wall WA, Samson FB (eds) Predicting species occurrences: issues of scale and accuracy. Island Press, Covelo, pp 507–517

1 C Biodivers Conserv (2009) 18:739–763 761

Fleishman E, Mac Nally R, Murphy DD (2005) Relationships among non-native plants, diversity of plants and butterXies, and adequacy of spatial sampling. Biol J Linn Soc Lond 85:157–166. doi:10.1111/ j.1095-8312.2005.00479.x Forman RTT, Godron M (1986) Landscape ecology. Wiley, New York Fortin M-J, Agrawal AA (2005) Landscape ecology comes of age. Ecology 86:1965–1967. doi:10.1890/05- 0003 Gilbert LE, Singer MC (1975) ButterXy ecology. Annu Rev Ecol Syst 6:365–397. doi:10.1146/annurev.es.06. 110175.002053 Golden DM, Crist TO (1999) Experimental eVects of habitat fragmentation on old-Weld canopy insects: com- munity, guild and species responses. Oecologia 118:371–380. doi:10.1007/s004420050738 Gustafson EJ (1998) Quantifying landscape spatial pattern: what is the state of the art? Ecosystems (N Y, Print) 1:143–156. doi:10.1007/s100219900011 Hodkinson ID (2005) Terrestrial insects along elevation gradients: species and community responses to alti- tude. Biol Rev Camb Philos Soc 80:489–513. doi:10.1017/S1464793105006767 Hogsden KL, Hutchinson TC (2004) ButterXy assemblages along a human disturbance gradient in Ontario, Canada. Can J Zool 82:739–748. doi:10.1139/z04-048 Huston MA (1994) Biological diversity: the coexistence of species in changing landscapes. Cambridge Uni- versity Press, Cambridge Kareiva P (1983) InXuence of vegetation texture on herbivore populations: resource concentration and herbi- vore movement. In: Denno RF, McClure MS (eds) Variable plants and herbivore in natural and managed systems. Academic Press, New York, pp 259–289 Kie JG, Bowyer RT, Nicholson MC et al (2002) Landscape heterogeneity at diVering scales: eVects on spatial distribution of mule deer. Ecology 83:530–544 Kingsolver JG (1983a) Ecological signiWcance of Xight activity in Colias butterXies—implications for repro- ductive strategy and population structure. Ecology 64:546–551. doi:10.2307/1939974 Kingsolver JG (1983b) Thermoregulation and Xight in Colias butterXies—elevational patterns and mechanis- tic limitations. Ecology 64:534–545. doi:10.2307/1939973 Kolasa J, Rollo CD (1991) The heterogeneity of heterogeneity: a glossary. In: Kolasa J, Pickett STA (eds) Ecological heterogeneity. Springer, New York, pp 1–23 Krauss J, SteVan-Dewenter I, Tscharntke T (2003) How does landscape context contribute to eVects of habitat fragmentation on diversity and population density of butterXies? J Biogeogr 30:889–900. doi:10.1046/ j.1365-2699.2003.00878.x Kremen C (1992) Assessing the indicator properties of species assemblages for natural areas monitoring. Ecol Appl 2:203–217. doi:10.2307/1941776 Kumar S, Stohlgren TJ, Chong GW (2006) Spatial heterogeneity inXuences native and nonnative plant spe- cies richness. Ecology 87:3186–3199. doi:10.1890/0012-9658(2006)87[3186:SHINAN]2.0.CO;2 Legendre P, Legendre L (1998) Numerical ecology, 2nd edn. Elsevier Science, Amsterdam Levins R (1968) Evolution in changing environments. Princeton University Press, Princeton Li BL, Archer S (1997) Weighted mean patch size: a robust index for quantifying landscape structure. Ecol Modell 102:353–361. doi:10.1016/S0304-3800(97)00071-9 Li HB, Reynolds JF (1994) A simulation experiment to quantify spatial heterogeneity in categorical maps. Ecology 75:2446–2455. doi:10.2307/1940898 Luoto M, Heikkinen RK, Poyry J et al (2006) Determinants of the biogeographical distribution of butterXies in boreal regions. J Biogeogr 33:1764–1778. doi:10.1111/j.1365-2699.2005.01395.x Mazerolle MJ, Villard MA (1999) Patch characteristics and landscape context as predictors of species pres- ence and abundance: a review. Ecoscience 6:117–124 McGarigal K, Marks BJ (1995) FRAGSTATS. Spatial analysis program for quantifying landscape structure. USDA Forest Service General Technical Report PNWGTR-351. PaciWc Northwest Research Station, Portland Milne BT (1991) Heterogeneity as a multiscale characteristic of landscape. In: Kolasa J, Pickett STA (eds) Ecological heterogeneity. Springer, New York, pp 69–84 Moody A, Johnson DM (2001) Land-surface phenologies from AVHRR using the discrete Fourier transform. Remote Sens Environ 75:305–323. doi:10.1016/S0034-4257(00)00175-9 Moran PAP (1948) The interpretation of statistical maps. J R Stat Soc B 10:243–251 Morisette JT, Jarnevich CS, Ullah A et al (2006) A tamarisk habitat suitability map for the continental United States. Front Ecol Environ 4:11–17. doi:10.1890/1540-9295(2006)004[0012:ATHSMF]2.0.CO;2 Murphy DD, Weiss SB (1992) EVects of climate change on biological diversity in Western North America: species losses and mechanisms. In: Peters RL, Lovejoy TE (eds) Global warming and biological diver- sity. Yale University Press, London, pp 355–368

1 C 762 Biodivers Conserv (2009) 18:739–763

Neter J, Kutner MH, Nachtsheim CJ et al (1996) Applied linear statistical models: regression, analysis of var- iance, and experimental designs, 4th edn. Irwin, Chicago Nieminen M, Siljander M, Hanski I (2004) Structure and dynamics of Melitaea cinxia populations. In: Ehrlich PR, Hanski I (eds) On the wings of Checkerspots. A model system for population biology. Oxford Uni- versity Press, Oxford, pp 63–91 O’ Neill RV, Hunsaker CT, Timmins SP et al (1996) Scale problems in reporting landscape pattern at the regional scale. Landsc Ecol 11:169–180 Opler PA (1999) A Weld guide to western butterXies. Houghton MiZin, New York Opler PA, Krizek GO (1984) ButterXies east of the Great Plains. John Hopkins University Press, Baltimore Opler PA, Warren AD (2003) ButterXies of North America: scientiWc names list for butterXy species of North America, north of Mexico, April 30, 2003. Contributions of the C.P. Gillette Museum of Diversity, Department of Bioagricultural Sciences and Pest Management, Colorado State University, Fort Collins, CO, USA Parmesan C (1996) Climate and species’ range. Nature 382:765–766 Peet RK (1981) Forest vegetation of the Colorado Front Range-composition and dynamics. Vegetatio 45:3–75 Peet RK (1988) Forests of the Rocky Mountains. In: Barbour MG, Billings WD (eds) North American terres- trial vegetation. Cambridge University Press, New York, pp 64–101 Pickett STA, Cadenasso ML (1995) Landscape ecology—spatial heterogeneity in ecological systems. Science 269:331–334 Pollard E, Yates TJ (1993) Monitoring butterXies for ecology and conservation. Chapman and Hall, London Rabasa SG, Gutierrez D, Escudero A (2005) Egg laying by a butterXy on a fragmented host plant: a multi- level approach. Ecography 28:629–639 Ricketts TH (2001) The matrix matters: eVective isolation in fragmented landscapes. Am Nat 158:87–99 Risser PG, Karr JP, Forman RTT (1984) Landscape ecology: directions and approaches. Illinois Natural His- tory Survey Special Publication Number 2, Champaign SAS Institute (2004) SAS for windows, version 9.1. SAS Institute, Cary Schneider C (2003) The inXuence of spatial scale on quantifying insect dispersal: an analysis of butterXy data. Ecol Entomol 28:252–256 Scoble MJ (1992) The : form, function, and diversity. Oxford University Press, New York Scott JA (1975) Movement of Precis-Coenia, a pseudoterritorial submigrant (Lepidoptera-Nymphalidae). J Anim Ecol 44:843–850 Scott JA (1986) The butterXies of North America. Stanford University Press, Stanford Seto KC, Fleishman E, Fay JP, Betrus CJ (2004) Linking spatial patterns of bird and butterXy species richness with Landsat TM derived NDVI. Int J Remote Sens 25:4309–4324 Shao G, Wu J (2008) On the accuracy of landscape pattern analysis using remote sensing data. Landsc Ecol 23:505–511 Simard M, Zhang KQ, Rivera-Monroy VH et al (2006) Mapping height and biomass of mangrove forests in Everglades National Park with SRTM elevation data. Photogramm Eng Remote Sens 72:299–311 Simonson SE, Opler PA, Stohlgren TJ et al (2001) Rapid assessment of butterXy diversity in a montane land- scape. Biodivers Conserv 10:1369–1386 Stefanescu C, Herrando S, Paramo F (2004) ButterXy species richness in the north-west Mediterranean Basin: the role of natural and human-induced factors. J Biogeogr 31:905–915 SteVan-Dewenter I, Munzenberg U, Burger C et al (2002) Scale-dependent eVects of landscape context on three pollinator guilds. Ecology 83:1421–1432 Stohlgren TJ (2006) Measuring plant diversity: lessons from the Weld. Oxford University Press, New York Stohlgren TJ, Bull KA, Otsuki Y (1998) Comparison of rangeland vegetation sampling techniques in the cen- tral grasslands. J Range Manag 51:164–172 Strathmann KS (2005) Local and landscape inXuences on meadow butterXy communities in northeastern Sierra Nevada. MS Thesis, San Francisco State University, San Francisco, CA, USA. Available via DIALOG. http://sagehen.ucnrs.org/pubs/Theses/Strathmann.pdf. Accessed 5 November 2008 Summerville KS, Veech JA, Crist TO (2002) Does variation in patch use among butterXy species contribute to nestedness at Wne spatial scales? Oikos 97:195–204 Thies C, SteVan-Dewenter I, Tscharntke T (2003) EVects of landscape context on herbivory and parasitism at diVerent spatial scales. Oikos 101:18–25 Thomas JA (2005) Monitoring change in the abundance and distribution of insects using butterXies and other indicator groups. Philos Trans R Soc B-Biol Sci 360:339–357 Thomas CD, Hanski I (1997) ButterXy metapopulations. In: Hanski I, Gilpin M (eds) Metapopulation biol- ogy. Academic Press, London, pp 359–384 Thomas CD, Mallorie HC (1985) Rarity, species richness and conservation: butterXies of the Atlas Mountains in Morocco. Biol Conserv 33:95–117 1 C Biodivers Conserv (2009) 18:739–763 763

Thomas CD, Baguette M, Lewis OT (2000) ButterXy movement and conservation in patchy landscapes. In: Gosling LM, Sutherland WJ (eds) Behavior and conservation. Cambridge University Press, Cambridge, pp 85–104 Turner MG (2005) Landscape ecology: what is the state of the science. Ann Rev Ecol Evol Syst 36:319–344 Turner MG, Gardner RH, Dale VH et al (1989) Predicting the spread of disturbance across heterogeneous landscapes. Oikos 55:121–129 Turner MG, Gardner RH, O’Neill RV (2001) Landscape ecology in theory and practice-pattern and process. Springer, New York Van Dyck H, Matthysen E (1999) Habitat fragmentation and insect Xight: a changing ‘design’ in a changing landscape? Trends Ecol Evol 14:172–174 Vogelmann JE, Sohl TL, Campbell PV et al (1998) Regional land cover characterization using Landsat the- matic mapper data and ancillary data sources. Environ Monit Assess 51:415–428 Wagner HH, Fortin MJ (2005) Spatial analysis of landscapes: concepts and statistics. Ecology 86:1975–1987 Wahlberg N, Klemetti T, Selonen V et al (2002) Metapopulation structure and movements in Wve species of checkerspot butterXies. Oecologia 130:33–43 Warren MS (1984) The biology and status of the wood white butterXy Leptidea sinapis. (Lepidoptera: Pieri- dae) in the British Isles. Entomol Gaz 35:207–223 Warren TL, Betts MG, Diamond AW et al (2005) The inXuence of local habitat and landscape composition on cavity-nesting birds in a forested mosaic. For Ecol Manag 214:331–343 Wettstein W, Schmid B (1999) Conservation of arthropod diversity in montane wetlands: eVect of altitude, habitat quality and habitat fragmentation on butterXies and grasshoppers. J Appl Ecol 36:363–373 Wood PA, Samways MJ (1991) Landscape element pattern and continuity of butterXy Xight paths in an eco- logically landscaped botanical garden, Natal, South-Africa. Biol Conserv 58:149–166

1 C

View publication stats