- Email: [email protected]
Grassland connectivity in fragmented agricultural landscapes of the north-central United States
Biological Conservation 217 (2018) 121–130
Contents lists available at ScienceDirect
Biological Conservation journal homepage: www.elsevier.com/locate/biocon
Grassland connectivity in fragmented agricultural landscapes of the northcentral United States
MARK
Michael C. Wimberlya,b,⁎, Diane M. Narema,b, Peter J. Baumanb, Benjamin T. Carlsonb, Marissa A. Ahleringc a b c
Geospatial Sciences Center of Excellence, South Dakota State University, Brookings, SD, USA Department of Natural Resource Management, South Dakota State University, Brookings, SD, USA The Nature Conservancy, Minneapolis, MN, USA
A R T I C L E I N F O
A B S T R A C T
Keywords: Connectivity Dispersal Graph theory Movement Network analysis Grassland Fragmentation
In the prairies of North America, remnant native grasslands are threatened by continuing agricultural extensification. Fragmentation of the remaining grassland isolates patches and limits the potential for dispersal of native species. We explored these impacts by analyzing the spatial pattern of native grassland habitats in the Prairie Coteau region of eastern South Dakota and western Minnesota, USA. Undisturbed grasslands were mapped using a GIS database of land use history combined with manual interpretation of high-resolution aerial photographs. Network analysis based on graph theory was used to examine how connectivity changed depending on the potential movement distances of organisms and to identify important patches that made large contributions to connectivity throughout the broader network. Interpatch movement was assessed using Euclidian distance as well as cost-weighted distance that assigned lower movement cost to grasslands than to humanmodified land cover types. Much of the undisturbed grassland was concentrated in a single large cluster, which was connected to other habitat concentrations via corridors of “stepping stone” patches. A small number of “keystone patches”, whose loss would have a disproportionately large effect on overall connectivity, were also identified. The locations of the major corridors were relatively consistent across different movement distances. Information about patch-level importance for overall network connectivity should be taken into account when prioritizing conservation and restoration. Future studies can build on this research by conducting more detailed assessments focused on particular species of concern and portions of the study area where connectivity is most limited.
1. Introduction The prairies of North America are perhaps the most threatened terrestrial ecosystem on the continent, and these vanishing grasslands are a priority for conservation. Nearly 98% of the native northern tallgrass prairie vegetation has been lost, mainly through conversion to annual row crops and planting of non-native grasses to support livestock production (Samson et al., 2004). Although much of this conversion took place in the late 19th and early 20th centuries, the trend of native grassland loss has continued to the present day. Beginning in the early 2000s, higher crop prices driven in part by increasing demands for production of ethanol and other biofuels have led to high rates of grassland conversion to croplands, particularly at the edges of the Midwestern Corn Belt (Lark et al., 2015; Wright et al., 2017; Wright and Wimberly, 2013). Most of the recently converted grassland has been on land enrolled in the Conservation Reserve Program (CRP), which ⁎
provides farmers with a yearly rental payment in exchange for removing environmentally sensitive land from agricultural production and planting perennial grasses. However, substantial conversion of native grasslands has also occurred in the eastern Dakotas (Wimberly et al., 2017). There are significant concerns about the negative impacts of agricultural expansion and grassland loss on wetland habitats, populations of native species, and carbon sequestration (Ahlering et al., 2016; Brennan and Kuvlesky Jr, 2005; Johnston, 2013; Mushet et al., 2014; Swengel et al., 2011). This paper contributes to our understanding of these environmental impacts by analyzing spatial patterns and connectivity in a network of undisturbed native grassland patches located at the western edge of the Corn Belt. The process of fragmentation, through which contiguous grasslands are broken apart into smaller and more isolated patches, is intrinsically coupled with grassland loss. Habitat fragmentation generally leads to a loss of biodiversity in the remnant patches (Haddad et al., 2015). In the
Corresponding author at: Geospatial Sciences Center of Excellence, South Dakota State University, Box 506B, Brookings, SD 57007, USA. E-mail address: [email protected] (M.C. Wimberly).
http://dx.doi.org/10.1016/j.biocon.2017.10.031 Received 27 June 2017; Received in revised form 12 October 2017; Accepted 29 October 2017 0006-3207/ © 2017 Elsevier Ltd. All rights reserved.
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 1. A) Major land cover types within the study area. Dashed black line represents the boundary of subfigure B. Solid black line represents the boundary of the Prairie Coteau ecoregion. B) Zoomed map of major land cover types with patch connections shown for a maximum cost-weighted movement distance of 1300 m. C) Study area location in the north-central United States.
potential connectivity (Calabrese and Fagan, 2004) by incorporating information about species dispersal distances as well as the suitability of dispersal habitat. This approach has been widely applied to assess connectivity of a variety of habitat types including forests (Minor and Urban, 2008), shrublands (Ferreras, 2001), and wetlands (BishopTaylor et al., 2015; McIntyre et al., 2014; Wright, 2010). Despite the burgeoning number of landscape connectivity studies, relatively few have been conducted in temperate grasslands (Correa Ayram et al., 2016). Given the high rate of recent grassland conversion in parts of the central United States (Lark et al., 2015; Wright and Wimberly, 2013), there is a particular need to understand the degree to which fragmentation limits the flow of organisms and their genes among prairie remnants. One reason for this dearth of studies is the difficulty of accurately mapping native grasslands. Although grasslands can be identified using medium-resolution satellite sensors such as the Landsat Thematic Mapper, Enhanced Thematic Mapper, and Operational Land Imager, it is difficult if not impossible to distinguish native prairie from tame pastures and other planted grasslands. The recent development of a pilot dataset characterizing the distribution of unplowed grasslands in eastern South Dakota using air photo interpretation and historical land use data provides a novel opportunity to study habitat connectivity in this rapidly changing region (Bauman et al., 2015; Bauman et al., 2016). The overarching goal of this study was to characterize the connectivity of the habitat network of undisturbed grassland patches in the Coteau des Prairies region of eastern South Dakota and western Minnesota. Specific objectives were to (1) identify the remnant grassland patches that are most important for facilitating movement through the habitat network and assess their current protection status, (2) determine whether these critical patches change with the maximum inter-patch movement distance, and (3) assess the sensitivity of these relationships to the effects of dispersal habitat quality in
fragmented grasslands of North America, patches of remnant prairie vegetation are often surrounded by croplands that provide low-quality habitat for most native species. Thus, many organisms including birds (Herkert, 1994; Johnson and Igl, 2001; Winter et al., 2006), insects (Davis et al., 2007; Swengel and Swengel, 2015), and plants (Wagenius et al., 2010; Wagenius et al., 2007) are sensitive to grassland patch size as well as the spatial pattern of the surrounding landscape. To a large degree, species respond to landscape patterns because they move within and between habitat patches to carry out activities necessary for survival and reproduction. These activities can include foraging, seeking shelter, finding a mate, seeking prey, avoiding predators, defending territory, dispersal, and migration. There has been considerable research on how fragmentation affects various species, and these studies have typically characterized landscape patterns using patch-based indices that measure the sizes and shapes of habitat patches, proximity to other habitat patches, and the contextual effects of the landscape matrix (Kupfer, 2012). A major drawback of these patch-based metrics is that they characterize the structural connectivity of the landscape itself, but do not necessarily capture important functional aspects of the landscape related to the behavior and movement of organisms (Kupfer, 2012). In response to this limitation, the development of graph-based methods focused on landscape connectivity has become a major research thrust in the field of landscape ecology (Galpern et al., 2011). An important goal of this approach is to quantify the degree to which landscape patterns facilitate or impede movement through the larger habitat network. Various connectivity metrics have been developed by combining elements of graph theory (Urban and Keitt, 2001), electrical circuit theory (McRae et al., 2008), and least-cost distance analysis (Adriaensen et al., 2003) to assess movement potential in heterogeneous landscapes. These techniques allow for the assessment of 122
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
2.3. Geospatial data processing
the intervening matrix.
A standard technique in connectivity analysis is to first identify a relatively small number of key source areas where there are large clusters of habitat, and then quantify the potential for movement among these clusters based on geographic distance and the conditions in the intervening matrix (Wade et al., 2015). This approach is justifiable in assessments focused on large mammals that require large concentrations of habitat and can potentially move over long distances. In the Prairie Coteau, however, many species of conservation concern are low-mobility organisms such as plants and insects. Previous research has suggested that connectivity assessments that are appropriate for long-range dispersers are often not relevant to other species with limited movement potential (Cushman and Landguth, 2012). Therefore, we decided to retain as many of the patches from the original undisturbed grassland dataset as possible and use graph theory to explore how the connectedness of the larger patch network changes with potential movement distance. The initial dataset consisted of 33,305 undisturbed grassland patches encompassing 519,556 ha (16% of the total study area), which included many small fragments. We conducted spatial analysis on a subset of these patches with a core area (based on a 30 m edge depth) of 5 ha or greater. Our rationale for this decision was that although small and narrow patches dominated by edge effects might provide temporary dispersal habitat for species moving between the larger patches, they would be less likely to support persistent populations of native species. We recognize that the minimum patch size required to support a viable population will vary depending on the species and habitat quality of the patch. Thus, our size threshold should be viewed as general rule intended to identify the subset of patches most likely to provide habitat for multiple species. Five hectares also represents a minimum mapping unit above which we are highly confident that all patches were accurately identified and digitized in the process of manually interpreting the aerial photographs. The total number of patches was thus reduced to 13,438 patches covering 481,670 ha, which accounted for 93% of the total area of undisturbed grasslands in the original dataset. For each patch, adjacent patches were identified and two types of interpatch distances were calculated using the LinkageMapper program in ArcGIS (McRae and Kavanagh, 2011). Euclidian distances were computed by identifying adjacent patches using Euclidian allocation, and then calculating interpatch distance as the shortest edge-to-edge distance. Cost-weighted distances were based on a movement cost surface defined using the 2012 Cropland Data Layer and the undisturbed grassland layer (Fig. 1). A generic weighting scheme was used to characterize variability in dispersal habitat for grassland-associated species (Table 1). Grasslands (including undisturbed grasslands as well as other grasslands identified in the 2012 CDL) were assigned a cost of one, such that movement distances would be equivalent to the Euclidian scenario. Forests, which are relatively rare in the study area, were assigned a cost of two. Croplands, primarily monocultures of corn and soybeans, provide relatively poor dispersal habitat for most grassland species and were assigned a cost of five. We assumed that most grassland organisms would move around rather than across water bodies and developed areas, and assigned these cover types costs of 15 and 20 accordingly. To reduce the computational burden of the costdistance analysis, the cost layer was rescaled from 30 m to 90 m resolution using a majority filter. Adjacent patches were identified using a cost-weighted allocation procedure, and inter-patch distances were measured as the least-cost distance between patch edges.
2. Methods 2.1. Study area The study area encompassed 3,347,874 ha in eastern South Dakota and western Minnesota (Fig. 1). This area included the Coteau des Prairies (hereafter referred to as the Prairie Coteau) region, which consisted of the following EPA Level IV ecoregions: Prairie Coteau, Big Sioux Basin, Northern Glaciated Plains, and Prairie Coteau Escarpment (Fig. 1a). The southeastern portion of the study area also included part of the Loess Prairies and Des Moines Plains ecoregions north of the Minnesota-Iowa border and west of the Des Moines River. The Prairie Coteau is a broad plateau that separated two glacial lobes during the Wisconsinan glaciation. This landform is composed primarily of glacial sediments and rises to 275 m above the surrounding plains. The terrain consists of low rolling hills with many seasonal wetlands, semi-permanent wetlands, and lakes, and is drained by several major river systems, including the Minnesota, Red, James, and Big Sioux River watersheds. Annual precipitation ranges from approximately 600 mm in the northwest portion of the study area to nearly 800 mm in the southeast corner, with most rainfall occurring during the growing season. Temperatures typically remain below 0 °C from December through February and frequently reach highs of 27 °C or more during the summer months. The dominant historical vegetation ranged from mixed-grass prairie on the west slope to northern tallgrass prairie in the east with wooded coulees occurring along the edges of the Coteau. These vegetation patterns were influenced by frequent fires, grazing by large herbivores such as elk and bison, and drought. The current land cover is a mosaic of cropland, remnant native grassland, hayfield and pasture planted to non-native grasses, and retired cropland enrolled in the Conservation Reserve Program (CRP), which may be planted to native or non-native grasses. The majority of the Prairie Coteau is under private ownership, with the rest divided between tribal, federal, and state agencies. The largest cities are Brookings (population of 23,225) and Watertown (population of 22,057) with Sioux Falls (population of 168,586) located just outside the southern border of the study area.
2.2. Undisturbed grassland dataset Undisturbed grasslands were defined as grasslands for which we could find no evidence of any previous tillage or mechanical soil disturbance history. They were mapped using the 2012 Common Land Unit (CLU) cropland data layer from the United States Department of Agriculture Farm Service Agency. The CLU dataset extends back to the 1950s and encompasses all crop fields that have been enrolled in a USDA program and assigned a farm number and contains spatial information on field boundaries along with attributes that characterize the agricultural use of each parcel on an annual basis. These historical data were used to identify all parcels for which there was any historical record of use as cropland. For example, CRP grasslands would have their historical cropland status identified in the CLU database and thus would be excluded from the area of potential undisturbed grasslands. After excluding historically cropped areas, 1-m resolution, true-color digital orthophotos from the National Agricultural Imagery Program were used to manually digitize grassland patches and exclude areas where there was other visual evidence of disturbance. More details on the methodology can be found in Bauman et al. (2015) and Bauman et al. (2016). The undisturbed grassland data are archived on South Dakota State University's Public Research Access Institutional Repository and Information Exchange (Open PRAIRIE, http://openprairie. sdstate.edu/).
2.4. Network analysis We used graph theory to characterize the landscape as a network of connected grassland patches, and to analyze the potential for organisms to move through the network. A set of networks was defined using 123
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Table 1 Land cover classes used to generate the cost surface used in the cost-weighted distance analysis. Class
Description
Cost weight
Other grassland
Grassland/pasture and other hay/non-alfalfa classes from the USDA cropland data layer (CDL) intersected with undisturbed grassland patches with less than 5 ha core area Deciduous, evergreen, and mixed forest classes from the CDL Corn, soybeans, wheat, and all crop types identified in the CDL Open water and herbaceous wetland classes from the CDL Developed/open space from the CDL. Roads and other linear features were filtered out and reclassified based on the majority of surrounding land cover types.
1
Forest Cropland Water/wetland Developed
2 5 15 20
We used two approaches to identify influential patches that were critical for maintaining connectivity throughout the grassland network. Stepping stones were defined for each combination of movement distance and distance metric as patches that were in the highest 1% of betweenness centrality values (Bishop-Taylor et al., 2015). These stepping stones were intersected by many paths linking other patches in the network, and were thus considered the most likely to be used for movement between other patches. Keystone patches were patches that provided critical connectivity that would be lost if the patch was converted to non-habitat. To identify keystone patches, we implemented a node-removal experiment (Galpern et al., 2011). Patches were removed from the network one at a time and the area-weighted mean component size was recomputed with each patch deleted from the network. Keystone patches were defined as stepping stones whose removal decreased the area-weighted mean component size by 10% or more, reflecting fragmentation of one of the larger components into two or more smaller components with a consequent reduction in the potential for movement through the network. The 10% threshold was defined based on a natural break in the patchremoval experiment results. Across all distances and for both distance types, more than 99.9% of patches reduced the area-weighted mean component size by less than 10% when removed from the network. There was a second, much smaller group of patches that reduced the area-weighted mean component size from 20 to 40%, and the 10% threshold was selected to classify this group of influential patches. All data analyses were carried out with the R language and environment for statistical computing, using the igraph package for network analysis (Csardi and Nepusz, 2006). We plotted the graph metrics as a function of maximum movement distance (100–2000 m at 100 m intervals) for each distance type (Euclidian and cost-weighted) to explore the sensitivity of network connectivity to movement distance. We also mapped stepping stones and keystone patches over a range of distances to identify patches that were most critical for maintaining connectivity and determine whether these patch locations changed with movement distance. An overall summary map was created by overlaying stepping stones and keystone patches from all movement distances and distance metrics. Important stepping stones identified were stepping stones in at least 50% of the 40 networks examined (20 distance thresholds × 2 distance types). Important keystone patches were defined as important stepping stones that were also keystone patches in at least one of the networks. The protection status of stepping stones and keystone patches was assessed by overlaying the network analysis results on a dataset of protected areas and comparing the numbers and areas of patches located inside and outside of protected areas. Protected areas were defined as lands either legally protected from conversion through fee title or permanent easement or those owned by a public entity and therefore assumed to have a very low probability of conversion risk. The protected lands dataset was developed by compiling data obtained from all organizations that held a fee title or permanent easement interest within the Prairie Coteau boundary, including the U.S. Fish & Wildlife Service, the U.S. Department of Agriculture Natural Resource Conservation Service, The Nature Conservancy, the South Dakota Department of Game Fish & Parks, and the Northern Prairies Land
maximum movement distances ranging from 100 to 2000 m at 100 m intervals for each of the distance types (Euclidian and cost weighted). Neighbors (also known as edges in graph theory terminology) were defined as connected pairs of patches (also known as nodes or vertices in graph theory terminology) that were closer to one another than the movement distance, and between which movement could therefore occur (Fig. 1). Paths were defined as routes through which two patches could be connected via consecutive movement between neighboring patches. Components were defined as groups of patches that were all connected to one another via movement along one or more paths. For each network, we calculated a set of descriptive graph metrics that quantified various aspects of network structure (Kolaczyk and Csardi, 2014). Relevant metrics were identified from a review of previous studies of landscape connectivity (Bishop-Taylor et al., 2015; Minor and Urban, 2008; Wright, 2010), and six representative metrics were selected to characterize a variety of ecologically-relevant network characteristics. 1) The number of components characterized the overall degree of fragmentation by measuring the degree to which habitat was subdivided into separate sub-networks. 2) The largest component size quantified movement potential throughout the network as the area of habitat in the largest connected component. This metric assumes that the largest connected component in the network will provide the greatest potential for sustaining organism flow among patches and is thus particularly important from a conservation standpoint. 3) The area-weighted mean component size measured whether the total habitat area in the network was made up of many small components or was dominated by few large components. Networks comprised of fewer, larger components are assumed to facilitate more movement of organisms among patches. 4) The mean number of neighbors (also known as degree centrality in graph theory terminology) was summarized across the entire network to measure the degree to which individual patches were connected to neighboring patches. The potential for local movement between patches is assumed to be greater when patches have more neighbors. 5) The clustering coefficient (also known as transitivity) quantified the proportion of node triplets (sets of three connected patches) that were closed (each patch in the triplet was connected to the other two patches). High clustering coefficient values indicate that neighbors of a patch were also likely to be neighbors of one another, and that movement between patches is thus facilitated by relatively short paths. 6) Betweenness centrality was calculated for each patch by first identifying the shortest paths between all other pairs of patches, and then determining how many of these paths passed through the patch of interest. At the patch level, betweenness centrality is an indicator of the relative influence of each patch in facilitating connectivity throughout the larger network. The mean betweenness centrality is a global indicator of the amount of connectivity throughout the entire network.
124
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 2. Change in global network indices as a function of movement distance and type of distance measurement. A) Number of components; B) maximum component size; C) areaweighted mean component size; D) mean number of neighbors; E) global clustering coefficient; F) mean betweenness centrality.
largest patch and other large patches highlighted along with the stepping stones and keystone patches (Figs. 3, 4). These distances were used for mapping because they represented critical points at which decreases in maximum movement distance resulted in large changes in the graph metrics. Maps highlighting the largest component across all movement distances confirmed this supposition (Figs. S1, S2). Comparisons of maps at the critical distances with those at the next lower movement distance showed visible reductions in the size of the largest connected component. For Euclidian distances, the network with a movement distance of 300 m had the largest component located along the northeastern edge of the Prairie Coteau, with stepping stones concentrated inside this component and only a single keystone patch (Fig. 3a). When the movement distance increased to 700 m, the largest component increased in size to include another concentration of patches along the northwestern edge of the study area (Fig. 3b). A corridor of stepping stones, including several keystone patches, connected the two major habitat concentrations. At movement distances of 1000 m and 1600 m the largest component increased in size and expanded to the south along with a corridor of stepping stones (Fig. 3c, d). There were relatively few keystone patches at the 1000 m movement distance and none at 1600 m. In the cost-weighted distance networks, there were generally similar changes with movement distance (Fig. 4). However, the movement
Trust. All of these data sources were current as of 2014. A geospatial dataset that identifies the areas of undisturbed grassland within the boundaries of these protected areas is available via OpenPRAIRIE (http://openprairie.sdstate.edu/). 3. Results As the movement distance decreased, the patch networks became more fragmented (Fig. 2) with a greater number of smaller components, fewer patch neighbors, increased distances required to move through the network (indicated by lower clustering coefficients), and fewer shortest pathways passing through each patch (indicated by higher mean betweenness centralities). At a given movement distance, networks defined based on cost-weighted distance were more fragmented than those defined based on Euclidian distance. Number of components, mean number of neighbors, and clustering coefficient all varied gradually with movement distance (Fig. 2). In contrast, largest component size, area-weighted mean component size, and mean betweenness centrality all varied irregularly with movement distance in a stair-step fashion. Sharper declines in these metrics occurred when the movement distance dropped below 1600 m, 1000 m, 700 m, and 300 m for Euclidian distance, and below 1700 m, 1300 m, 1000 m, and 200 m for cost-weighted distance. We generated maps of the patch network for each of these movement distances with the 125
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 3. Network structure at four movement distances based on Euclidian distance. Colored patches highlight large components and black and red points indicate the locations of stepping stones and keystone patches. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
and 7.4% of the total area (Table 2). Stepping stones and keystone patches tended to be relatively large compared to the overall patch size distribution. Stepping stones and keystone patches were also more likely to be located in protected areas (Fig. 5b) than were undisturbed grassland patches in general (Table 3).
distances at which the major shifts in connectivity occurred were longer. For example, in the cost-weighted distance networks, the northeastern and northwestern habitat concentrations became connected when movement distance increased to 1300 m (Fig. 4c) versus 700 m in the Euclidian distance networks (Fig. 3b). There were also more keystone patches in the cost-weighted distance networks than in the Euclidian distance networks. The map of stepping stone and keystone patches across all movement distances highlighted several major corridors that were important for maintaining undisturbed grassland connectivity in the Prairie Coteau (Fig. 5a). The most important patches, which were identified as stepping stones and keystone patches across multiple movement distances, were in the large concentration of undisturbed grassland in the northeastern part of the Prairie Coteau and in a narrow corridor connecting this patch to the western side of the Coteau. There were also numerous keystone patches along this east-west corridor, indicating that a loss of only one or a few critical patches has the potential to restrict east-west connectivity. Several other corridors of stepping stone patches extended into the southern part of the study area and were important for maintaining north-south linkages at longer movement distances. Stepping stones accounted for 6.1% of the number of undisturbed grassland patches and 26.2% of the total undisturbed grassland area, whereas keystone patches accounted for 0.8% of the number of patches
4. Discussion Remnant, undisturbed grasslands are highly fragmented on the Prairie Coteau. The largest concentration is located along the northeast edge of the Coteau, and consists of a cluster of patches that remain connected at movement distances as low as 100 m in the Euclidian distance networks and 200 m in the cost-weighted distance networks (Fig. S1, S2). For species not capable of traversing more than 700 m through the non-habitat matrix, the patch network is fragmented into many isolated components and the potential for movement is limited. These constraints are even more severe under the assumption that there is a higher cost associated with movement through non-grassland habitats, in which case the major northeastern habitat cluster concentration becomes isolated at cost-weighted movement distances shorter than 1300 m. Although the locations of stepping stones shifted with movement distance, there are also numerous stepping stones that were identified in at least 50% of the networks examined. These important stepping stones are mostly located in the large concentration of 126
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 4. Network structure at four movement distances based on cost-weighted distance. Colored patches highlight large components and black and red points indicate the locations of stepping stones and keystone patches. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 5. A) Locations of all patches that were identified as stepping stones or keystone patches for at least one movement distance/ distance metric combination. Larger dots highlight important patches that were identified as stepping stones in at least 50% of the movement distance/distance metric combinations. B) Distribution of protected areas through the Prairie Coteau, subdivided by undisturbed grassland versus other land cover types.
127
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
et al., 2003; DeKeyser et al., 2013; Greer et al., 2016). For a comprehensive habitat assessment, patch-level measurements of connectivity would need to be combined with information on patch size and vegetation condition (Hodgson et al., 2011). The cost-weighted distances that we used to assess the effects of matrix conditions on movement assume directed dispersal in which the organism has perfect knowledge of the shortest path. Alternative approaches based on circuit theory assume that movement occurs as a random walk and may be more appropriate for some organisms (McRae et al., 2008). In future studies, we suggest that circuit theory approaches might be fruitful for more precisely highlighting potential movement pathways in areas of the Prairie Coteau where there are many critical linkages among undisturbed grassland patches. Because the analysis was grounded in a set of basic movement rules and a generic resistance surface derived from broad land cover types, the results provide a broad assessment of potential connectivity rather than details about actual connectivity for particular species (Calabrese and Fagan, 2004). Despite this generality, we identified several important corridors comprised of stepping stone patches that were consistent over a range of movement distances and distance metrics. This finding suggests that it is feasible to apply a “coarse-filter” approach to quantify connectivity and identify patches that are important to support movement of multiple species with a range of movement distances and dispersal habitat associations (Beier et al., 2011; Koen et al., 2014). Our identification of clusters of stepping stones and keystone patches provides a starting point for further study of portions of the habitat network where native grassland loss may have disproportionately large effects on movement through the larger network. Reconstruction may be necessary to curb biodiversity loss given the already high level of habitat loss and fragmentation in this system (Haddad et al., 2015). However, we also acknowledge that more detailed evaluations will be needed to focus on particular species. These assessments can be conducted using similar methods, but will require a more precise characterization of the sizes and conditions of undisturbed grassland patches that provide core habitat areas for the species of interest, as well as additional information about organismal behavior and habitats that constrain dispersal movements through the matrix (Elliot et al., 2014; McClure et al., 2016). For species whose life history and habitat associations are well studied, spatial simulations could be used to evaluate optimal locations for high diversity grassland reconstructions to enhance or create new connectivity across the landscape (e.g., Lookingbill et al., 2010). Stepping stones and keystone patches were more likely to be associated with protected areas than were undisturbed grassland patches in general (Table 2). This finding highlights the importance of protected area networks in maintaining connectivity despite the high level of grassland fragmentation. Bishop-Taylor et al. (2015) similarly found that wetlands located in protected habitats were more likely to be important for maintaining connectivity than those in non-protected habitats. Many of the protected stepping stones and keystone patches are located in the linear cluster of protected areas located along the northeastern edge of the study area. This area encompasses the highest elevations of the Prairie Coteau, and is characterized by more rugged topography and poorer soils than the rest of the study area. In other portions of the Coteau, protected undisturbed grasslands are distributed along riparian corridors and chains of wetlands. Although these
Table 2 Patch statistics for all patches, stepping stones, and keystone patches (all areas in ha).
All patches Stepping stones Keystone patches
Number
Total area
Mean patch size
Minimum patch size
Median patch size
Maximum patch size
7982
437,627
54.8
8.1
25.1
4593.9
485
114,649
236.4
9.6
70.7
4593.9
60
32,660
544.3
11.1
99.4
4593.9
undisturbed grassland patches in the northeastern part of the study area and along the corridor that connects the east and west sides of the Prairie Coteau. There are also many keystone patches along this eastwest corridor, indicating that this connection is tenuous and could be broken by the loss of one or a few critical grassland patches. Corridors of stepping stone patches similarly extend into the southern portion of the Coteau and have several concentrations of keystone patches, but are only traversable at longer movement distances (> = 1000 m for Euclidian distance and > = 1700 m for cost-weighted distance). Species with relatively short movement distances are most likely to be sensitive to the fragmented nature of the undisturbed grassland network in the Prairie Coteau. For example, the threatened Dakota skipper (Hesperia dacotae) and the endangered Poweshiek skipperling (Oarisma poweshiek) are butterflies associated with unplowed, highquality native prairie. For the Dakota skipper, mean movement distance over a period of a week or less was found to be less than 300 m (Dana, 1991). Movement distances of up to 800 m have also been reported, but it is unlikely that Dakota skippers are capable of moving more than 1000 m between habitat patches (Cochrane and Delphey, 2002). Other types of insects such as native bees are also limited in their capability to move among habitat patches, with all but the largest bee species exhibiting maximum movement distances shorter than 1000 m (Greenleaf et al., 2007). Plant movements are also likely to be restricted. Seed falling directly from the plant may not travel further than 1 m (Rabinowitz and Rapp, 1981; Pleasants and Jurik, 1992), and even wind-dispersed seeds have maximum dispersal distances ranging from tens to hundreds of meters (Thomson et al., 2011). Many plant species also rely on pollinators to move genes and animals in general to disperse seed at greater distances, and are thus restricted by the same factors limiting animal movements in the landscape (McConkey et al., 2012; Pleasants and Jurik, 1992; Rabinowitz and Rapp, 1981). These movement limitations likely decrease gene flow among isolated populations (Britten and Glasford, 2002), reduce the stability of metapopulations (Hanski and Ovaskainen, 2000), and may also limit the potential for migration in response to climate change (Nuñez et al., 2013). When interpreting these results, several important limitations of this analysis should be noted. We assumed that all undisturbed grassland patches above a fixed core area were potential habitat for a variety of native species. However, the structure and composition of remnant grasslands varies considerably because of environmental context and land management history, resulting in variable habitat quality (Collinge
Table 3 Statistics from the overlays of all patches, stepping stones, and keystone patches with protected areas. Number of patches
All patches Stepping stones Keystone patches
Area of patches (ha)
Protected
Unprotected
Percent Protected
Protected
Unprotected
Percent Protected
1153 129 24
6829 356 36
14.4 26.6 40.0
122,890 55,586 20,980
329,304 66,367 13,828
27.1 45.6 60.3
128
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
References
patterns of protected areas likely reflect the distribution of lands with relatively low suitability for cropping rather than an intentional selection to facilitate connectivity, they also emphasize that similar patterns of protected areas could be intentionally designed to enhance movement potential through the habitat network. Identifying habitat patches that are important for connectivity but are currently unprotected can help to determine future priorities for land protection and restoration. Stepping stones and keystone patches tended to be relatively large compared to the overall distribution of patch sizes (Table 3). This relationship is understandable because larger patches have more edge distributed across a broader area, and thus have the potential to be adjacent to many other patches. Thus, larger patches will tend to have more shortest paths at a given movement distance because they are connected to more neighboring patches. Likewise, the loss of a large patch will remove many links between neighbors and is thus is more likely to have a strong impact on network-level connectivity than the loss of a small patch. However, at least half of the stepping stones and keystone patches were less than 100 ha in size. These relatively small but important patches, particularly when they are outside of protected areas, represent weak points in the habitat network where loss of a relatively small area of grassland could have substantial impacts on overall connectivity.
Adriaensen, F., Chardon, J.P., De Blust, G., Swinnen, E., Villalba, S., Gulinck, H., Matthysen, E., 2003. The application of ‘least-cost’ modelling as a functional landscape model. Landsc. Urban Plan. 64, 233–247. Ahlering, M., Fargione, J., Parton, W., 2016. Potential carbon dioxide emission reductions from avoided grassland conversion in the northern Great Plains. Ecosphere 7, e01625. Bauman, P., Carlson, B., Butler, T., 2015. Quantifying Undisturbed Land in Minnesota's Prairie Coteau and Lac Qui Parle Valley Regions. South Dakota State University Extension, Brookings, SD Available at: http://openprairie.sdstate.edu/data_landsouthwestMN/1/. Bauman, P., Carlson, B., Butler, T., 2016. Quantifying Undisturbed (Native) Lands in Eastern South Dakota: 2013. South Dakota State University Extension, Brookings, SD Available at: http://openprairie.sdstate.edu/data_land-easternSD/1/. Beier, P., Spencer, W., Baldwin, R.F., McRae, B., 2011. Toward best practices for developing regional connectivity maps. Conserv. Biol. 25, 879–892. Bishop-Taylor, R., Tulbure, M.G., Broich, M., 2015. Surface water network structure, landscape resistance to movement and flooding vital for maintaining ecological connectivity across Australia's largest river basin. Landsc. Ecol. 30, 2045–2065. Brennan, L.A., Kuvlesky Jr, W.P., 2005. Invited paper: North American grassland birds: an unfolding conservation crisis? J. Wildl. Manag. 69, 1–13. Britten, H.B., Glasford, J.W., 2002. Genetic population structure of the Dakota skipper (Lepidoptera: Hesperia dacotae): a North American native prairie obligate. Conserv. Genet. 3, 363–374. Calabrese, J.M., Fagan, W.F., 2004. A comparison-shopper's guide to connectivity metrics. Front. Ecol. Environ. 2, 529–536. Cochrane, J.F., Delphey, P., 2002. Status Assessment and Conservation Guidelines: Dakota Skipper. U.S. Fish and Wildlife Service, Twin Cities Field Office. Collinge, S.K., Prudic, K.L., Oliver, J.C., 2003. Effects of local habitat characteristics and landscape context on grassland butterfly diversity. Conserv. Biol. 17, 178–187. Correa Ayram, C.A., Mendoza, M.E., Etter, A., Salicrup, D.R.P., 2016. Habitat connectivity in biodiversity conservation: a review of recent studies and applications. Prog. Phys. Geogr. 40, 7–37. Csardi, G., Nepusz, T., 2006. The igraph software package for complex network research. InterJournal Complex Syst. 1695, 1–9. Cushman, S.A., Landguth, E.L., 2012. Multi-taxa population connectivity in the Northern Rocky Mountains. Ecol. Model. 231, 101–112. Dana, R.P., 1991. Conservation management of the prairie skippers Hesperia dacotae and Hesperia ottoe. In: Minnesota Agricultural Experiment Station, Station Bulletin 5941991 AD-SB-5511-S, St. Paul, MN. Davis, J.D., Debinski, D.M., Danielson, B.J., 2007. Local and landscape effects on the butterfly community in fragmented Midwest USA prairie habitats. Landsc. Ecol. 22, 1341–1354. DeKeyser, E.S., Meehan, M., Clambey, G., Krabbenhoft, K., 2013. Cool season invasive grasses in northern Great Plains natural areas. Nat. Areas J. 33, 81–90. Elliot, N.B., Cushman, S.A., Macdonald, D.W., Loveridge, A.J., 2014. The devil is in the dispersers: predictions of landscape connectivity change with demography. J. Appl. Ecol. 51, 1169–1178. Ferreras, P., 2001. Landscape structure and asymmetrical inter-patch connectivity in a metapopulation of the endangered Iberian lynx. Biol. Conserv. 100, 125–136. Galpern, P., Manseau, M., Fall, A., 2011. Patch-based graphs of landscape connectivity: a guide to construction, analysis and application for conservation. Biol. Conserv. 144, 44–55. Greenleaf, S.S., Williams, N.M., Winfree, R., Kremen, C., 2007. Bee foraging ranges and their relationship to body size. Oecologia 153, 589–596. Greer, M.J., Bakker, K.K., Dieter, C.D., 2016. Grassland bird response to recent loss and degradation of native prairie in central and western South Dakota. Wilson J. Ornithol. 128, 278–289. Haddad, N.M., Brudvig, L.A., Clobert, J., Davies, K.F., Gonzalez, A., Holt, R.D., Lovejoy, T.E., Sexton, J.O., Austin, M.P., Collins, C.D., 2015. Habitat fragmentation and its lasting impact on Earth's ecosystems. Sci. Adv. 1, e1500052. Hanski, I., Ovaskainen, O., 2000. The metapopulation capacity of a fragmented landscape. Nature 404, 755–758. Herkert, J.R., 1994. The effects of habitat fragmentation on midwestern grassland bird communities. Ecol. Appl. 4, 461–471. Hodgson, J.A., Moilanen, A., Wintle, B.A., Thomas, C.D., 2011. Habitat area, quality and connectivity: striking the balance for efficient conservation. J. Appl. Ecol. 48, 148–152. Johnson, D.H., Igl, L.D., 2001. Area requirements of grassland birds: a regional perspective. Auk 118, 24–34. Johnston, C.A., 2013. Wetland losses due to row crop expansion in the Dakota Prairie Pothole region. Wetlands 33, 175–182. Koen, E.L., Bowman, J., Sadowski, C., Walpole, A.A., 2014. Landscape connectivity for wildlife: development and validation of multispecies linkage maps. Methods Ecol. Evol. 5, 626–633. Kolaczyk, E.D., Csardi, G., 2014. Statistical Analysis of Network Data with R. Springer, New York. Kupfer, J.A., 2012. Landscape ecology and biogeography: rethinking landscape metrics in a post-FRAGSTATS landscape. Prog. Phys. Geogr. 36, 400–420. Lark, T.J., Salmon, J.M., Gibbs, H.K., 2015. Cropland expansion outpaces agricultural and biofuel policies in the United States. Environ. Res. Lett. 10, 044003. Lookingbill, T.R., Gardner, R.H., Ferrari, J.R., Keller, C.E., 2010. Combining a dispersal model with network theory to assess habitat connectivity. Ecol. Appl. 20, 427–441. McClure, M.L., Hansen, A.J., Inman, R.M., 2016. Connecting models to movements:
5. Conclusions The Prairie Coteau, like other areas at the edge of the Corn Belt, is undergoing agricultural expansion that is driving the continued loss of native grasslands (Lark et al., 2015; Wimberly et al., 2017; Wright and Wimberly, 2013). Conservation is a challenge in this region given that the majority of land is privately owned. Therefore, limited resources must be judiciously invested through activities such as incentive programs, easements, land purchases, and restorations. One way to maximize the environmental benefits gained from these investments is to target these activities to specific locations where they can be shown to have the greatest positive impact (Wünscher et al., 2008). Connectivity is an essential ecological function that must be sustained to support flows of organisms and maintain viable regional metapopulations. Incorporating native grassland connectivity into regional conservation assessment requires accurate data on the distribution of undisturbed grasslands and robust analytical techniques to provide straightforward assessments highlighting the most important habitat patches for maintaining connectivity. Our study represents a first step in this direction by conducting a generic analysis of habitat connectivity across a range of movement distances and identifying important stepping stones that facilitate connectivity as well as keystone patches whose loss would have a disproportionately large effect on overall connectivity. Future studies can build on this research by conducting more detailed connectivity assessments for potential reconstruction, particular species of concern and finer-grained assessment of movement potential in the portions of the study area where connectivity is most limited. Acknowledgements This research was supported by the NSF Macrosystems Biology Program (NSF-EF 1544083). The undisturbed lands dataset was developed with funding from The Nature Conservancy through a grant awarded by the National Fish and Wildlife Foundation (NFWF Grant 2009-0084-000). Tanner Butler, Cody Grewing, and Jim Madsen contributed to the development of the undisturbed lands dataset. We thank two anonymous reviewers for their constructive feedback on an earlier version of this manuscript. Appendix A. Supplementary figures Supplementary figures to this article can be found online at https:// doi.org/10.1016/j.biocon.2017.10.031. 129
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Thomson, F.J., Moles, A.T., Auld, T.D., Kingsford, R.T., 2011. Seed dispersal distance is more strongly correlated with plant height than with seed mass. J. Ecol. 99, 1299–1307. Urban, D., Keitt, T., 2001. Landscape connectivity: a graph-theoretic perspective. Ecology 82, 1205–1218. Wade, A.A., McKelvey, K.S., Schwartz, M.K., 2015. Resistance-surface-based wildlife conservation connectivity modeling: summary of efforts in the United States and guide for practitioners. In: General Technical Report RMRS-GTR-333, Fort Collins, CO. Wagenius, S., Lonsdorf, E., Neuhauser, C., 2007. Patch aging and the S-Allee effect: breeding system effects on the demographic response of plants to habitat fragmentation. Am. Nat. 169, 383–397. Wagenius, S., Hangelbroek, H.H., Ridley, C.E., Shaw, R.G., 2010. Biparental inbreeding and interremnant mating in a perennial prairie plant: fitness consequences for progeny in their first eight years. Evolution 64, 761–771. Wimberly, M.C., Janssen, L.L., Hennessy, D.A., Luri, M., Chowdhury, N.M., Feng, H., 2017. Cropland expansion and grassland loss in the eastern Dakotas: new insights from a farm-level survey. Land Use Policy 63, 160–173. Winter, M., Johnson, D.H., Shaffer, J.A., Donovan, T.M., Svedarsky, W.D., 2006. Patch size and landscape effects on density and nesting success of grassland birds. J. Wildl. Manag. 70, 158–172. Wright, C.K., 2010. Spatiotemporal dynamics of prairie wetland networks: power-law scaling and implications for conservation planning. Ecology 91, 1924–1930. Wright, C.K., Wimberly, M.C., 2013. Recent land use change in the Western Corn Belt threatens grasslands and wetlands. Proc. Natl. Acad. Sci. U. S. A. 110, 4134–4139. Wright, C.K., Larson, B., Lark, T.J., Gibbs, H.K., 2017. Recent grassland losses are concentrated around US ethanol refineries. Environ. Res. Lett. 12, 044001. Wünscher, T., Engel, S., Wunder, S., 2008. Spatial targeting of payments for environmental services: a tool for boosting conservation benefits. Ecol. Econ. 65, 822–833.
testing connectivity model predictions against empirical migration and dispersal data. Landsc. Ecol. 31, 1419–1432. McConkey, K.R., Prasad, S., Corlett, R.T., Campos-Arceiz, A., Brodie, J.F., Rogers, H., Santamaria, L., 2012. Seed dispersal in changing landscapes. Biol. Conserv. 146, 1–13. McIntyre, N.E., Wright, C.K., Swain, S., Hayhoe, K., Liu, G., Schwartz, F.W., Henebry, G.M., 2014. Climate forcing of wetland landscape connectivity in the Great Plains. Front. Ecol. Environ. 12, 59–64. McRae, B., Kavanagh, D., 2011. Linkage Mapper Connectivity Analysis Software. The Nature Conservancy, Seattle, WA. McRae, B.H., Dickson, B.G., Keitt, T.H., Shah, V.B., 2008. Using circuit theory to model connectivity in ecology, evolution, and conservation. Ecology 89, 2712–2724. Minor, E.S., Urban, D.L., 2008. A graph-theory framework for evaluating landscape connectivity and conservation planning. Conserv. Biol. 22, 297–307. Mushet, D.M., Neau, J.L., Euliss, N.H., 2014. Modeling effects of conservation grassland losses on amphibian habitat. Biol. Conserv. 174, 93–100. Nuñez, T.A., Lawler, J.J., McRae, B.H., Pierce, D.J., Krosby, M.B., Kavanagh, D.M., Singleton, P.H., Tewksbury, J.J., 2013. Connectivity planning to address climate change. Conserv. Biol. 27, 407–416. Pleasants, J.M., Jurik, T.W., 1992. Dispersion of seedlings of the prairie compass plant, Silphium laciniatum (Asteraceae). Am. J. Bot. 133–137. Rabinowitz, D., Rapp, J.K., 1981. Dispersal abilities of seven sparse and common grasses from a Missouri prairie. Am. J. Bot. 616–624. Samson, F.B., Knopf, F.L., Ostlie, W.R., 2004. Great Plains ecosystems: past, present, and future. Wildl. Soc. Bull. 32, 6–15. Swengel, A.B., Swengel, S.R., 2015. Grass-skipper (Hesperiinae) trends in midwestern USA grasslands during 1988–2013. J. Insect Conserv. 19, 279–292. Swengel, S.R., Schlicht, D., Olsen, F., Swengel, A.B., 2011. Declines of prairie butterflies in the midwestern USA. J. Insect Conserv. 15, 327–339.
130
Contents lists available at ScienceDirect
Biological Conservation journal homepage: www.elsevier.com/locate/biocon
Grassland connectivity in fragmented agricultural landscapes of the northcentral United States
MARK
Michael C. Wimberlya,b,⁎, Diane M. Narema,b, Peter J. Baumanb, Benjamin T. Carlsonb, Marissa A. Ahleringc a b c
Geospatial Sciences Center of Excellence, South Dakota State University, Brookings, SD, USA Department of Natural Resource Management, South Dakota State University, Brookings, SD, USA The Nature Conservancy, Minneapolis, MN, USA
A R T I C L E I N F O
A B S T R A C T
Keywords: Connectivity Dispersal Graph theory Movement Network analysis Grassland Fragmentation
In the prairies of North America, remnant native grasslands are threatened by continuing agricultural extensification. Fragmentation of the remaining grassland isolates patches and limits the potential for dispersal of native species. We explored these impacts by analyzing the spatial pattern of native grassland habitats in the Prairie Coteau region of eastern South Dakota and western Minnesota, USA. Undisturbed grasslands were mapped using a GIS database of land use history combined with manual interpretation of high-resolution aerial photographs. Network analysis based on graph theory was used to examine how connectivity changed depending on the potential movement distances of organisms and to identify important patches that made large contributions to connectivity throughout the broader network. Interpatch movement was assessed using Euclidian distance as well as cost-weighted distance that assigned lower movement cost to grasslands than to humanmodified land cover types. Much of the undisturbed grassland was concentrated in a single large cluster, which was connected to other habitat concentrations via corridors of “stepping stone” patches. A small number of “keystone patches”, whose loss would have a disproportionately large effect on overall connectivity, were also identified. The locations of the major corridors were relatively consistent across different movement distances. Information about patch-level importance for overall network connectivity should be taken into account when prioritizing conservation and restoration. Future studies can build on this research by conducting more detailed assessments focused on particular species of concern and portions of the study area where connectivity is most limited.
1. Introduction The prairies of North America are perhaps the most threatened terrestrial ecosystem on the continent, and these vanishing grasslands are a priority for conservation. Nearly 98% of the native northern tallgrass prairie vegetation has been lost, mainly through conversion to annual row crops and planting of non-native grasses to support livestock production (Samson et al., 2004). Although much of this conversion took place in the late 19th and early 20th centuries, the trend of native grassland loss has continued to the present day. Beginning in the early 2000s, higher crop prices driven in part by increasing demands for production of ethanol and other biofuels have led to high rates of grassland conversion to croplands, particularly at the edges of the Midwestern Corn Belt (Lark et al., 2015; Wright et al., 2017; Wright and Wimberly, 2013). Most of the recently converted grassland has been on land enrolled in the Conservation Reserve Program (CRP), which ⁎
provides farmers with a yearly rental payment in exchange for removing environmentally sensitive land from agricultural production and planting perennial grasses. However, substantial conversion of native grasslands has also occurred in the eastern Dakotas (Wimberly et al., 2017). There are significant concerns about the negative impacts of agricultural expansion and grassland loss on wetland habitats, populations of native species, and carbon sequestration (Ahlering et al., 2016; Brennan and Kuvlesky Jr, 2005; Johnston, 2013; Mushet et al., 2014; Swengel et al., 2011). This paper contributes to our understanding of these environmental impacts by analyzing spatial patterns and connectivity in a network of undisturbed native grassland patches located at the western edge of the Corn Belt. The process of fragmentation, through which contiguous grasslands are broken apart into smaller and more isolated patches, is intrinsically coupled with grassland loss. Habitat fragmentation generally leads to a loss of biodiversity in the remnant patches (Haddad et al., 2015). In the
Corresponding author at: Geospatial Sciences Center of Excellence, South Dakota State University, Box 506B, Brookings, SD 57007, USA. E-mail address: [email protected] (M.C. Wimberly).
http://dx.doi.org/10.1016/j.biocon.2017.10.031 Received 27 June 2017; Received in revised form 12 October 2017; Accepted 29 October 2017 0006-3207/ © 2017 Elsevier Ltd. All rights reserved.
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 1. A) Major land cover types within the study area. Dashed black line represents the boundary of subfigure B. Solid black line represents the boundary of the Prairie Coteau ecoregion. B) Zoomed map of major land cover types with patch connections shown for a maximum cost-weighted movement distance of 1300 m. C) Study area location in the north-central United States.
potential connectivity (Calabrese and Fagan, 2004) by incorporating information about species dispersal distances as well as the suitability of dispersal habitat. This approach has been widely applied to assess connectivity of a variety of habitat types including forests (Minor and Urban, 2008), shrublands (Ferreras, 2001), and wetlands (BishopTaylor et al., 2015; McIntyre et al., 2014; Wright, 2010). Despite the burgeoning number of landscape connectivity studies, relatively few have been conducted in temperate grasslands (Correa Ayram et al., 2016). Given the high rate of recent grassland conversion in parts of the central United States (Lark et al., 2015; Wright and Wimberly, 2013), there is a particular need to understand the degree to which fragmentation limits the flow of organisms and their genes among prairie remnants. One reason for this dearth of studies is the difficulty of accurately mapping native grasslands. Although grasslands can be identified using medium-resolution satellite sensors such as the Landsat Thematic Mapper, Enhanced Thematic Mapper, and Operational Land Imager, it is difficult if not impossible to distinguish native prairie from tame pastures and other planted grasslands. The recent development of a pilot dataset characterizing the distribution of unplowed grasslands in eastern South Dakota using air photo interpretation and historical land use data provides a novel opportunity to study habitat connectivity in this rapidly changing region (Bauman et al., 2015; Bauman et al., 2016). The overarching goal of this study was to characterize the connectivity of the habitat network of undisturbed grassland patches in the Coteau des Prairies region of eastern South Dakota and western Minnesota. Specific objectives were to (1) identify the remnant grassland patches that are most important for facilitating movement through the habitat network and assess their current protection status, (2) determine whether these critical patches change with the maximum inter-patch movement distance, and (3) assess the sensitivity of these relationships to the effects of dispersal habitat quality in
fragmented grasslands of North America, patches of remnant prairie vegetation are often surrounded by croplands that provide low-quality habitat for most native species. Thus, many organisms including birds (Herkert, 1994; Johnson and Igl, 2001; Winter et al., 2006), insects (Davis et al., 2007; Swengel and Swengel, 2015), and plants (Wagenius et al., 2010; Wagenius et al., 2007) are sensitive to grassland patch size as well as the spatial pattern of the surrounding landscape. To a large degree, species respond to landscape patterns because they move within and between habitat patches to carry out activities necessary for survival and reproduction. These activities can include foraging, seeking shelter, finding a mate, seeking prey, avoiding predators, defending territory, dispersal, and migration. There has been considerable research on how fragmentation affects various species, and these studies have typically characterized landscape patterns using patch-based indices that measure the sizes and shapes of habitat patches, proximity to other habitat patches, and the contextual effects of the landscape matrix (Kupfer, 2012). A major drawback of these patch-based metrics is that they characterize the structural connectivity of the landscape itself, but do not necessarily capture important functional aspects of the landscape related to the behavior and movement of organisms (Kupfer, 2012). In response to this limitation, the development of graph-based methods focused on landscape connectivity has become a major research thrust in the field of landscape ecology (Galpern et al., 2011). An important goal of this approach is to quantify the degree to which landscape patterns facilitate or impede movement through the larger habitat network. Various connectivity metrics have been developed by combining elements of graph theory (Urban and Keitt, 2001), electrical circuit theory (McRae et al., 2008), and least-cost distance analysis (Adriaensen et al., 2003) to assess movement potential in heterogeneous landscapes. These techniques allow for the assessment of 122
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
2.3. Geospatial data processing
the intervening matrix.
A standard technique in connectivity analysis is to first identify a relatively small number of key source areas where there are large clusters of habitat, and then quantify the potential for movement among these clusters based on geographic distance and the conditions in the intervening matrix (Wade et al., 2015). This approach is justifiable in assessments focused on large mammals that require large concentrations of habitat and can potentially move over long distances. In the Prairie Coteau, however, many species of conservation concern are low-mobility organisms such as plants and insects. Previous research has suggested that connectivity assessments that are appropriate for long-range dispersers are often not relevant to other species with limited movement potential (Cushman and Landguth, 2012). Therefore, we decided to retain as many of the patches from the original undisturbed grassland dataset as possible and use graph theory to explore how the connectedness of the larger patch network changes with potential movement distance. The initial dataset consisted of 33,305 undisturbed grassland patches encompassing 519,556 ha (16% of the total study area), which included many small fragments. We conducted spatial analysis on a subset of these patches with a core area (based on a 30 m edge depth) of 5 ha or greater. Our rationale for this decision was that although small and narrow patches dominated by edge effects might provide temporary dispersal habitat for species moving between the larger patches, they would be less likely to support persistent populations of native species. We recognize that the minimum patch size required to support a viable population will vary depending on the species and habitat quality of the patch. Thus, our size threshold should be viewed as general rule intended to identify the subset of patches most likely to provide habitat for multiple species. Five hectares also represents a minimum mapping unit above which we are highly confident that all patches were accurately identified and digitized in the process of manually interpreting the aerial photographs. The total number of patches was thus reduced to 13,438 patches covering 481,670 ha, which accounted for 93% of the total area of undisturbed grasslands in the original dataset. For each patch, adjacent patches were identified and two types of interpatch distances were calculated using the LinkageMapper program in ArcGIS (McRae and Kavanagh, 2011). Euclidian distances were computed by identifying adjacent patches using Euclidian allocation, and then calculating interpatch distance as the shortest edge-to-edge distance. Cost-weighted distances were based on a movement cost surface defined using the 2012 Cropland Data Layer and the undisturbed grassland layer (Fig. 1). A generic weighting scheme was used to characterize variability in dispersal habitat for grassland-associated species (Table 1). Grasslands (including undisturbed grasslands as well as other grasslands identified in the 2012 CDL) were assigned a cost of one, such that movement distances would be equivalent to the Euclidian scenario. Forests, which are relatively rare in the study area, were assigned a cost of two. Croplands, primarily monocultures of corn and soybeans, provide relatively poor dispersal habitat for most grassland species and were assigned a cost of five. We assumed that most grassland organisms would move around rather than across water bodies and developed areas, and assigned these cover types costs of 15 and 20 accordingly. To reduce the computational burden of the costdistance analysis, the cost layer was rescaled from 30 m to 90 m resolution using a majority filter. Adjacent patches were identified using a cost-weighted allocation procedure, and inter-patch distances were measured as the least-cost distance between patch edges.
2. Methods 2.1. Study area The study area encompassed 3,347,874 ha in eastern South Dakota and western Minnesota (Fig. 1). This area included the Coteau des Prairies (hereafter referred to as the Prairie Coteau) region, which consisted of the following EPA Level IV ecoregions: Prairie Coteau, Big Sioux Basin, Northern Glaciated Plains, and Prairie Coteau Escarpment (Fig. 1a). The southeastern portion of the study area also included part of the Loess Prairies and Des Moines Plains ecoregions north of the Minnesota-Iowa border and west of the Des Moines River. The Prairie Coteau is a broad plateau that separated two glacial lobes during the Wisconsinan glaciation. This landform is composed primarily of glacial sediments and rises to 275 m above the surrounding plains. The terrain consists of low rolling hills with many seasonal wetlands, semi-permanent wetlands, and lakes, and is drained by several major river systems, including the Minnesota, Red, James, and Big Sioux River watersheds. Annual precipitation ranges from approximately 600 mm in the northwest portion of the study area to nearly 800 mm in the southeast corner, with most rainfall occurring during the growing season. Temperatures typically remain below 0 °C from December through February and frequently reach highs of 27 °C or more during the summer months. The dominant historical vegetation ranged from mixed-grass prairie on the west slope to northern tallgrass prairie in the east with wooded coulees occurring along the edges of the Coteau. These vegetation patterns were influenced by frequent fires, grazing by large herbivores such as elk and bison, and drought. The current land cover is a mosaic of cropland, remnant native grassland, hayfield and pasture planted to non-native grasses, and retired cropland enrolled in the Conservation Reserve Program (CRP), which may be planted to native or non-native grasses. The majority of the Prairie Coteau is under private ownership, with the rest divided between tribal, federal, and state agencies. The largest cities are Brookings (population of 23,225) and Watertown (population of 22,057) with Sioux Falls (population of 168,586) located just outside the southern border of the study area.
2.2. Undisturbed grassland dataset Undisturbed grasslands were defined as grasslands for which we could find no evidence of any previous tillage or mechanical soil disturbance history. They were mapped using the 2012 Common Land Unit (CLU) cropland data layer from the United States Department of Agriculture Farm Service Agency. The CLU dataset extends back to the 1950s and encompasses all crop fields that have been enrolled in a USDA program and assigned a farm number and contains spatial information on field boundaries along with attributes that characterize the agricultural use of each parcel on an annual basis. These historical data were used to identify all parcels for which there was any historical record of use as cropland. For example, CRP grasslands would have their historical cropland status identified in the CLU database and thus would be excluded from the area of potential undisturbed grasslands. After excluding historically cropped areas, 1-m resolution, true-color digital orthophotos from the National Agricultural Imagery Program were used to manually digitize grassland patches and exclude areas where there was other visual evidence of disturbance. More details on the methodology can be found in Bauman et al. (2015) and Bauman et al. (2016). The undisturbed grassland data are archived on South Dakota State University's Public Research Access Institutional Repository and Information Exchange (Open PRAIRIE, http://openprairie. sdstate.edu/).
2.4. Network analysis We used graph theory to characterize the landscape as a network of connected grassland patches, and to analyze the potential for organisms to move through the network. A set of networks was defined using 123
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Table 1 Land cover classes used to generate the cost surface used in the cost-weighted distance analysis. Class
Description
Cost weight
Other grassland
Grassland/pasture and other hay/non-alfalfa classes from the USDA cropland data layer (CDL) intersected with undisturbed grassland patches with less than 5 ha core area Deciduous, evergreen, and mixed forest classes from the CDL Corn, soybeans, wheat, and all crop types identified in the CDL Open water and herbaceous wetland classes from the CDL Developed/open space from the CDL. Roads and other linear features were filtered out and reclassified based on the majority of surrounding land cover types.
1
Forest Cropland Water/wetland Developed
2 5 15 20
We used two approaches to identify influential patches that were critical for maintaining connectivity throughout the grassland network. Stepping stones were defined for each combination of movement distance and distance metric as patches that were in the highest 1% of betweenness centrality values (Bishop-Taylor et al., 2015). These stepping stones were intersected by many paths linking other patches in the network, and were thus considered the most likely to be used for movement between other patches. Keystone patches were patches that provided critical connectivity that would be lost if the patch was converted to non-habitat. To identify keystone patches, we implemented a node-removal experiment (Galpern et al., 2011). Patches were removed from the network one at a time and the area-weighted mean component size was recomputed with each patch deleted from the network. Keystone patches were defined as stepping stones whose removal decreased the area-weighted mean component size by 10% or more, reflecting fragmentation of one of the larger components into two or more smaller components with a consequent reduction in the potential for movement through the network. The 10% threshold was defined based on a natural break in the patchremoval experiment results. Across all distances and for both distance types, more than 99.9% of patches reduced the area-weighted mean component size by less than 10% when removed from the network. There was a second, much smaller group of patches that reduced the area-weighted mean component size from 20 to 40%, and the 10% threshold was selected to classify this group of influential patches. All data analyses were carried out with the R language and environment for statistical computing, using the igraph package for network analysis (Csardi and Nepusz, 2006). We plotted the graph metrics as a function of maximum movement distance (100–2000 m at 100 m intervals) for each distance type (Euclidian and cost-weighted) to explore the sensitivity of network connectivity to movement distance. We also mapped stepping stones and keystone patches over a range of distances to identify patches that were most critical for maintaining connectivity and determine whether these patch locations changed with movement distance. An overall summary map was created by overlaying stepping stones and keystone patches from all movement distances and distance metrics. Important stepping stones identified were stepping stones in at least 50% of the 40 networks examined (20 distance thresholds × 2 distance types). Important keystone patches were defined as important stepping stones that were also keystone patches in at least one of the networks. The protection status of stepping stones and keystone patches was assessed by overlaying the network analysis results on a dataset of protected areas and comparing the numbers and areas of patches located inside and outside of protected areas. Protected areas were defined as lands either legally protected from conversion through fee title or permanent easement or those owned by a public entity and therefore assumed to have a very low probability of conversion risk. The protected lands dataset was developed by compiling data obtained from all organizations that held a fee title or permanent easement interest within the Prairie Coteau boundary, including the U.S. Fish & Wildlife Service, the U.S. Department of Agriculture Natural Resource Conservation Service, The Nature Conservancy, the South Dakota Department of Game Fish & Parks, and the Northern Prairies Land
maximum movement distances ranging from 100 to 2000 m at 100 m intervals for each of the distance types (Euclidian and cost weighted). Neighbors (also known as edges in graph theory terminology) were defined as connected pairs of patches (also known as nodes or vertices in graph theory terminology) that were closer to one another than the movement distance, and between which movement could therefore occur (Fig. 1). Paths were defined as routes through which two patches could be connected via consecutive movement between neighboring patches. Components were defined as groups of patches that were all connected to one another via movement along one or more paths. For each network, we calculated a set of descriptive graph metrics that quantified various aspects of network structure (Kolaczyk and Csardi, 2014). Relevant metrics were identified from a review of previous studies of landscape connectivity (Bishop-Taylor et al., 2015; Minor and Urban, 2008; Wright, 2010), and six representative metrics were selected to characterize a variety of ecologically-relevant network characteristics. 1) The number of components characterized the overall degree of fragmentation by measuring the degree to which habitat was subdivided into separate sub-networks. 2) The largest component size quantified movement potential throughout the network as the area of habitat in the largest connected component. This metric assumes that the largest connected component in the network will provide the greatest potential for sustaining organism flow among patches and is thus particularly important from a conservation standpoint. 3) The area-weighted mean component size measured whether the total habitat area in the network was made up of many small components or was dominated by few large components. Networks comprised of fewer, larger components are assumed to facilitate more movement of organisms among patches. 4) The mean number of neighbors (also known as degree centrality in graph theory terminology) was summarized across the entire network to measure the degree to which individual patches were connected to neighboring patches. The potential for local movement between patches is assumed to be greater when patches have more neighbors. 5) The clustering coefficient (also known as transitivity) quantified the proportion of node triplets (sets of three connected patches) that were closed (each patch in the triplet was connected to the other two patches). High clustering coefficient values indicate that neighbors of a patch were also likely to be neighbors of one another, and that movement between patches is thus facilitated by relatively short paths. 6) Betweenness centrality was calculated for each patch by first identifying the shortest paths between all other pairs of patches, and then determining how many of these paths passed through the patch of interest. At the patch level, betweenness centrality is an indicator of the relative influence of each patch in facilitating connectivity throughout the larger network. The mean betweenness centrality is a global indicator of the amount of connectivity throughout the entire network.
124
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 2. Change in global network indices as a function of movement distance and type of distance measurement. A) Number of components; B) maximum component size; C) areaweighted mean component size; D) mean number of neighbors; E) global clustering coefficient; F) mean betweenness centrality.
largest patch and other large patches highlighted along with the stepping stones and keystone patches (Figs. 3, 4). These distances were used for mapping because they represented critical points at which decreases in maximum movement distance resulted in large changes in the graph metrics. Maps highlighting the largest component across all movement distances confirmed this supposition (Figs. S1, S2). Comparisons of maps at the critical distances with those at the next lower movement distance showed visible reductions in the size of the largest connected component. For Euclidian distances, the network with a movement distance of 300 m had the largest component located along the northeastern edge of the Prairie Coteau, with stepping stones concentrated inside this component and only a single keystone patch (Fig. 3a). When the movement distance increased to 700 m, the largest component increased in size to include another concentration of patches along the northwestern edge of the study area (Fig. 3b). A corridor of stepping stones, including several keystone patches, connected the two major habitat concentrations. At movement distances of 1000 m and 1600 m the largest component increased in size and expanded to the south along with a corridor of stepping stones (Fig. 3c, d). There were relatively few keystone patches at the 1000 m movement distance and none at 1600 m. In the cost-weighted distance networks, there were generally similar changes with movement distance (Fig. 4). However, the movement
Trust. All of these data sources were current as of 2014. A geospatial dataset that identifies the areas of undisturbed grassland within the boundaries of these protected areas is available via OpenPRAIRIE (http://openprairie.sdstate.edu/). 3. Results As the movement distance decreased, the patch networks became more fragmented (Fig. 2) with a greater number of smaller components, fewer patch neighbors, increased distances required to move through the network (indicated by lower clustering coefficients), and fewer shortest pathways passing through each patch (indicated by higher mean betweenness centralities). At a given movement distance, networks defined based on cost-weighted distance were more fragmented than those defined based on Euclidian distance. Number of components, mean number of neighbors, and clustering coefficient all varied gradually with movement distance (Fig. 2). In contrast, largest component size, area-weighted mean component size, and mean betweenness centrality all varied irregularly with movement distance in a stair-step fashion. Sharper declines in these metrics occurred when the movement distance dropped below 1600 m, 1000 m, 700 m, and 300 m for Euclidian distance, and below 1700 m, 1300 m, 1000 m, and 200 m for cost-weighted distance. We generated maps of the patch network for each of these movement distances with the 125
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 3. Network structure at four movement distances based on Euclidian distance. Colored patches highlight large components and black and red points indicate the locations of stepping stones and keystone patches. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
and 7.4% of the total area (Table 2). Stepping stones and keystone patches tended to be relatively large compared to the overall patch size distribution. Stepping stones and keystone patches were also more likely to be located in protected areas (Fig. 5b) than were undisturbed grassland patches in general (Table 3).
distances at which the major shifts in connectivity occurred were longer. For example, in the cost-weighted distance networks, the northeastern and northwestern habitat concentrations became connected when movement distance increased to 1300 m (Fig. 4c) versus 700 m in the Euclidian distance networks (Fig. 3b). There were also more keystone patches in the cost-weighted distance networks than in the Euclidian distance networks. The map of stepping stone and keystone patches across all movement distances highlighted several major corridors that were important for maintaining undisturbed grassland connectivity in the Prairie Coteau (Fig. 5a). The most important patches, which were identified as stepping stones and keystone patches across multiple movement distances, were in the large concentration of undisturbed grassland in the northeastern part of the Prairie Coteau and in a narrow corridor connecting this patch to the western side of the Coteau. There were also numerous keystone patches along this east-west corridor, indicating that a loss of only one or a few critical patches has the potential to restrict east-west connectivity. Several other corridors of stepping stone patches extended into the southern part of the study area and were important for maintaining north-south linkages at longer movement distances. Stepping stones accounted for 6.1% of the number of undisturbed grassland patches and 26.2% of the total undisturbed grassland area, whereas keystone patches accounted for 0.8% of the number of patches
4. Discussion Remnant, undisturbed grasslands are highly fragmented on the Prairie Coteau. The largest concentration is located along the northeast edge of the Coteau, and consists of a cluster of patches that remain connected at movement distances as low as 100 m in the Euclidian distance networks and 200 m in the cost-weighted distance networks (Fig. S1, S2). For species not capable of traversing more than 700 m through the non-habitat matrix, the patch network is fragmented into many isolated components and the potential for movement is limited. These constraints are even more severe under the assumption that there is a higher cost associated with movement through non-grassland habitats, in which case the major northeastern habitat cluster concentration becomes isolated at cost-weighted movement distances shorter than 1300 m. Although the locations of stepping stones shifted with movement distance, there are also numerous stepping stones that were identified in at least 50% of the networks examined. These important stepping stones are mostly located in the large concentration of 126
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Fig. 4. Network structure at four movement distances based on cost-weighted distance. Colored patches highlight large components and black and red points indicate the locations of stepping stones and keystone patches. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 5. A) Locations of all patches that were identified as stepping stones or keystone patches for at least one movement distance/ distance metric combination. Larger dots highlight important patches that were identified as stepping stones in at least 50% of the movement distance/distance metric combinations. B) Distribution of protected areas through the Prairie Coteau, subdivided by undisturbed grassland versus other land cover types.
127
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
et al., 2003; DeKeyser et al., 2013; Greer et al., 2016). For a comprehensive habitat assessment, patch-level measurements of connectivity would need to be combined with information on patch size and vegetation condition (Hodgson et al., 2011). The cost-weighted distances that we used to assess the effects of matrix conditions on movement assume directed dispersal in which the organism has perfect knowledge of the shortest path. Alternative approaches based on circuit theory assume that movement occurs as a random walk and may be more appropriate for some organisms (McRae et al., 2008). In future studies, we suggest that circuit theory approaches might be fruitful for more precisely highlighting potential movement pathways in areas of the Prairie Coteau where there are many critical linkages among undisturbed grassland patches. Because the analysis was grounded in a set of basic movement rules and a generic resistance surface derived from broad land cover types, the results provide a broad assessment of potential connectivity rather than details about actual connectivity for particular species (Calabrese and Fagan, 2004). Despite this generality, we identified several important corridors comprised of stepping stone patches that were consistent over a range of movement distances and distance metrics. This finding suggests that it is feasible to apply a “coarse-filter” approach to quantify connectivity and identify patches that are important to support movement of multiple species with a range of movement distances and dispersal habitat associations (Beier et al., 2011; Koen et al., 2014). Our identification of clusters of stepping stones and keystone patches provides a starting point for further study of portions of the habitat network where native grassland loss may have disproportionately large effects on movement through the larger network. Reconstruction may be necessary to curb biodiversity loss given the already high level of habitat loss and fragmentation in this system (Haddad et al., 2015). However, we also acknowledge that more detailed evaluations will be needed to focus on particular species. These assessments can be conducted using similar methods, but will require a more precise characterization of the sizes and conditions of undisturbed grassland patches that provide core habitat areas for the species of interest, as well as additional information about organismal behavior and habitats that constrain dispersal movements through the matrix (Elliot et al., 2014; McClure et al., 2016). For species whose life history and habitat associations are well studied, spatial simulations could be used to evaluate optimal locations for high diversity grassland reconstructions to enhance or create new connectivity across the landscape (e.g., Lookingbill et al., 2010). Stepping stones and keystone patches were more likely to be associated with protected areas than were undisturbed grassland patches in general (Table 2). This finding highlights the importance of protected area networks in maintaining connectivity despite the high level of grassland fragmentation. Bishop-Taylor et al. (2015) similarly found that wetlands located in protected habitats were more likely to be important for maintaining connectivity than those in non-protected habitats. Many of the protected stepping stones and keystone patches are located in the linear cluster of protected areas located along the northeastern edge of the study area. This area encompasses the highest elevations of the Prairie Coteau, and is characterized by more rugged topography and poorer soils than the rest of the study area. In other portions of the Coteau, protected undisturbed grasslands are distributed along riparian corridors and chains of wetlands. Although these
Table 2 Patch statistics for all patches, stepping stones, and keystone patches (all areas in ha).
All patches Stepping stones Keystone patches
Number
Total area
Mean patch size
Minimum patch size
Median patch size
Maximum patch size
7982
437,627
54.8
8.1
25.1
4593.9
485
114,649
236.4
9.6
70.7
4593.9
60
32,660
544.3
11.1
99.4
4593.9
undisturbed grassland patches in the northeastern part of the study area and along the corridor that connects the east and west sides of the Prairie Coteau. There are also many keystone patches along this eastwest corridor, indicating that this connection is tenuous and could be broken by the loss of one or a few critical grassland patches. Corridors of stepping stone patches similarly extend into the southern portion of the Coteau and have several concentrations of keystone patches, but are only traversable at longer movement distances (> = 1000 m for Euclidian distance and > = 1700 m for cost-weighted distance). Species with relatively short movement distances are most likely to be sensitive to the fragmented nature of the undisturbed grassland network in the Prairie Coteau. For example, the threatened Dakota skipper (Hesperia dacotae) and the endangered Poweshiek skipperling (Oarisma poweshiek) are butterflies associated with unplowed, highquality native prairie. For the Dakota skipper, mean movement distance over a period of a week or less was found to be less than 300 m (Dana, 1991). Movement distances of up to 800 m have also been reported, but it is unlikely that Dakota skippers are capable of moving more than 1000 m between habitat patches (Cochrane and Delphey, 2002). Other types of insects such as native bees are also limited in their capability to move among habitat patches, with all but the largest bee species exhibiting maximum movement distances shorter than 1000 m (Greenleaf et al., 2007). Plant movements are also likely to be restricted. Seed falling directly from the plant may not travel further than 1 m (Rabinowitz and Rapp, 1981; Pleasants and Jurik, 1992), and even wind-dispersed seeds have maximum dispersal distances ranging from tens to hundreds of meters (Thomson et al., 2011). Many plant species also rely on pollinators to move genes and animals in general to disperse seed at greater distances, and are thus restricted by the same factors limiting animal movements in the landscape (McConkey et al., 2012; Pleasants and Jurik, 1992; Rabinowitz and Rapp, 1981). These movement limitations likely decrease gene flow among isolated populations (Britten and Glasford, 2002), reduce the stability of metapopulations (Hanski and Ovaskainen, 2000), and may also limit the potential for migration in response to climate change (Nuñez et al., 2013). When interpreting these results, several important limitations of this analysis should be noted. We assumed that all undisturbed grassland patches above a fixed core area were potential habitat for a variety of native species. However, the structure and composition of remnant grasslands varies considerably because of environmental context and land management history, resulting in variable habitat quality (Collinge
Table 3 Statistics from the overlays of all patches, stepping stones, and keystone patches with protected areas. Number of patches
All patches Stepping stones Keystone patches
Area of patches (ha)
Protected
Unprotected
Percent Protected
Protected
Unprotected
Percent Protected
1153 129 24
6829 356 36
14.4 26.6 40.0
122,890 55,586 20,980
329,304 66,367 13,828
27.1 45.6 60.3
128
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
References
patterns of protected areas likely reflect the distribution of lands with relatively low suitability for cropping rather than an intentional selection to facilitate connectivity, they also emphasize that similar patterns of protected areas could be intentionally designed to enhance movement potential through the habitat network. Identifying habitat patches that are important for connectivity but are currently unprotected can help to determine future priorities for land protection and restoration. Stepping stones and keystone patches tended to be relatively large compared to the overall distribution of patch sizes (Table 3). This relationship is understandable because larger patches have more edge distributed across a broader area, and thus have the potential to be adjacent to many other patches. Thus, larger patches will tend to have more shortest paths at a given movement distance because they are connected to more neighboring patches. Likewise, the loss of a large patch will remove many links between neighbors and is thus is more likely to have a strong impact on network-level connectivity than the loss of a small patch. However, at least half of the stepping stones and keystone patches were less than 100 ha in size. These relatively small but important patches, particularly when they are outside of protected areas, represent weak points in the habitat network where loss of a relatively small area of grassland could have substantial impacts on overall connectivity.
Adriaensen, F., Chardon, J.P., De Blust, G., Swinnen, E., Villalba, S., Gulinck, H., Matthysen, E., 2003. The application of ‘least-cost’ modelling as a functional landscape model. Landsc. Urban Plan. 64, 233–247. Ahlering, M., Fargione, J., Parton, W., 2016. Potential carbon dioxide emission reductions from avoided grassland conversion in the northern Great Plains. Ecosphere 7, e01625. Bauman, P., Carlson, B., Butler, T., 2015. Quantifying Undisturbed Land in Minnesota's Prairie Coteau and Lac Qui Parle Valley Regions. South Dakota State University Extension, Brookings, SD Available at: http://openprairie.sdstate.edu/data_landsouthwestMN/1/. Bauman, P., Carlson, B., Butler, T., 2016. Quantifying Undisturbed (Native) Lands in Eastern South Dakota: 2013. South Dakota State University Extension, Brookings, SD Available at: http://openprairie.sdstate.edu/data_land-easternSD/1/. Beier, P., Spencer, W., Baldwin, R.F., McRae, B., 2011. Toward best practices for developing regional connectivity maps. Conserv. Biol. 25, 879–892. Bishop-Taylor, R., Tulbure, M.G., Broich, M., 2015. Surface water network structure, landscape resistance to movement and flooding vital for maintaining ecological connectivity across Australia's largest river basin. Landsc. Ecol. 30, 2045–2065. Brennan, L.A., Kuvlesky Jr, W.P., 2005. Invited paper: North American grassland birds: an unfolding conservation crisis? J. Wildl. Manag. 69, 1–13. Britten, H.B., Glasford, J.W., 2002. Genetic population structure of the Dakota skipper (Lepidoptera: Hesperia dacotae): a North American native prairie obligate. Conserv. Genet. 3, 363–374. Calabrese, J.M., Fagan, W.F., 2004. A comparison-shopper's guide to connectivity metrics. Front. Ecol. Environ. 2, 529–536. Cochrane, J.F., Delphey, P., 2002. Status Assessment and Conservation Guidelines: Dakota Skipper. U.S. Fish and Wildlife Service, Twin Cities Field Office. Collinge, S.K., Prudic, K.L., Oliver, J.C., 2003. Effects of local habitat characteristics and landscape context on grassland butterfly diversity. Conserv. Biol. 17, 178–187. Correa Ayram, C.A., Mendoza, M.E., Etter, A., Salicrup, D.R.P., 2016. Habitat connectivity in biodiversity conservation: a review of recent studies and applications. Prog. Phys. Geogr. 40, 7–37. Csardi, G., Nepusz, T., 2006. The igraph software package for complex network research. InterJournal Complex Syst. 1695, 1–9. Cushman, S.A., Landguth, E.L., 2012. Multi-taxa population connectivity in the Northern Rocky Mountains. Ecol. Model. 231, 101–112. Dana, R.P., 1991. Conservation management of the prairie skippers Hesperia dacotae and Hesperia ottoe. In: Minnesota Agricultural Experiment Station, Station Bulletin 5941991 AD-SB-5511-S, St. Paul, MN. Davis, J.D., Debinski, D.M., Danielson, B.J., 2007. Local and landscape effects on the butterfly community in fragmented Midwest USA prairie habitats. Landsc. Ecol. 22, 1341–1354. DeKeyser, E.S., Meehan, M., Clambey, G., Krabbenhoft, K., 2013. Cool season invasive grasses in northern Great Plains natural areas. Nat. Areas J. 33, 81–90. Elliot, N.B., Cushman, S.A., Macdonald, D.W., Loveridge, A.J., 2014. The devil is in the dispersers: predictions of landscape connectivity change with demography. J. Appl. Ecol. 51, 1169–1178. Ferreras, P., 2001. Landscape structure and asymmetrical inter-patch connectivity in a metapopulation of the endangered Iberian lynx. Biol. Conserv. 100, 125–136. Galpern, P., Manseau, M., Fall, A., 2011. Patch-based graphs of landscape connectivity: a guide to construction, analysis and application for conservation. Biol. Conserv. 144, 44–55. Greenleaf, S.S., Williams, N.M., Winfree, R., Kremen, C., 2007. Bee foraging ranges and their relationship to body size. Oecologia 153, 589–596. Greer, M.J., Bakker, K.K., Dieter, C.D., 2016. Grassland bird response to recent loss and degradation of native prairie in central and western South Dakota. Wilson J. Ornithol. 128, 278–289. Haddad, N.M., Brudvig, L.A., Clobert, J., Davies, K.F., Gonzalez, A., Holt, R.D., Lovejoy, T.E., Sexton, J.O., Austin, M.P., Collins, C.D., 2015. Habitat fragmentation and its lasting impact on Earth's ecosystems. Sci. Adv. 1, e1500052. Hanski, I., Ovaskainen, O., 2000. The metapopulation capacity of a fragmented landscape. Nature 404, 755–758. Herkert, J.R., 1994. The effects of habitat fragmentation on midwestern grassland bird communities. Ecol. Appl. 4, 461–471. Hodgson, J.A., Moilanen, A., Wintle, B.A., Thomas, C.D., 2011. Habitat area, quality and connectivity: striking the balance for efficient conservation. J. Appl. Ecol. 48, 148–152. Johnson, D.H., Igl, L.D., 2001. Area requirements of grassland birds: a regional perspective. Auk 118, 24–34. Johnston, C.A., 2013. Wetland losses due to row crop expansion in the Dakota Prairie Pothole region. Wetlands 33, 175–182. Koen, E.L., Bowman, J., Sadowski, C., Walpole, A.A., 2014. Landscape connectivity for wildlife: development and validation of multispecies linkage maps. Methods Ecol. Evol. 5, 626–633. Kolaczyk, E.D., Csardi, G., 2014. Statistical Analysis of Network Data with R. Springer, New York. Kupfer, J.A., 2012. Landscape ecology and biogeography: rethinking landscape metrics in a post-FRAGSTATS landscape. Prog. Phys. Geogr. 36, 400–420. Lark, T.J., Salmon, J.M., Gibbs, H.K., 2015. Cropland expansion outpaces agricultural and biofuel policies in the United States. Environ. Res. Lett. 10, 044003. Lookingbill, T.R., Gardner, R.H., Ferrari, J.R., Keller, C.E., 2010. Combining a dispersal model with network theory to assess habitat connectivity. Ecol. Appl. 20, 427–441. McClure, M.L., Hansen, A.J., Inman, R.M., 2016. Connecting models to movements:
5. Conclusions The Prairie Coteau, like other areas at the edge of the Corn Belt, is undergoing agricultural expansion that is driving the continued loss of native grasslands (Lark et al., 2015; Wimberly et al., 2017; Wright and Wimberly, 2013). Conservation is a challenge in this region given that the majority of land is privately owned. Therefore, limited resources must be judiciously invested through activities such as incentive programs, easements, land purchases, and restorations. One way to maximize the environmental benefits gained from these investments is to target these activities to specific locations where they can be shown to have the greatest positive impact (Wünscher et al., 2008). Connectivity is an essential ecological function that must be sustained to support flows of organisms and maintain viable regional metapopulations. Incorporating native grassland connectivity into regional conservation assessment requires accurate data on the distribution of undisturbed grasslands and robust analytical techniques to provide straightforward assessments highlighting the most important habitat patches for maintaining connectivity. Our study represents a first step in this direction by conducting a generic analysis of habitat connectivity across a range of movement distances and identifying important stepping stones that facilitate connectivity as well as keystone patches whose loss would have a disproportionately large effect on overall connectivity. Future studies can build on this research by conducting more detailed connectivity assessments for potential reconstruction, particular species of concern and finer-grained assessment of movement potential in the portions of the study area where connectivity is most limited. Acknowledgements This research was supported by the NSF Macrosystems Biology Program (NSF-EF 1544083). The undisturbed lands dataset was developed with funding from The Nature Conservancy through a grant awarded by the National Fish and Wildlife Foundation (NFWF Grant 2009-0084-000). Tanner Butler, Cody Grewing, and Jim Madsen contributed to the development of the undisturbed lands dataset. We thank two anonymous reviewers for their constructive feedback on an earlier version of this manuscript. Appendix A. Supplementary figures Supplementary figures to this article can be found online at https:// doi.org/10.1016/j.biocon.2017.10.031. 129
Biological Conservation 217 (2018) 121–130
M.C. Wimberly et al.
Thomson, F.J., Moles, A.T., Auld, T.D., Kingsford, R.T., 2011. Seed dispersal distance is more strongly correlated with plant height than with seed mass. J. Ecol. 99, 1299–1307. Urban, D., Keitt, T., 2001. Landscape connectivity: a graph-theoretic perspective. Ecology 82, 1205–1218. Wade, A.A., McKelvey, K.S., Schwartz, M.K., 2015. Resistance-surface-based wildlife conservation connectivity modeling: summary of efforts in the United States and guide for practitioners. In: General Technical Report RMRS-GTR-333, Fort Collins, CO. Wagenius, S., Lonsdorf, E., Neuhauser, C., 2007. Patch aging and the S-Allee effect: breeding system effects on the demographic response of plants to habitat fragmentation. Am. Nat. 169, 383–397. Wagenius, S., Hangelbroek, H.H., Ridley, C.E., Shaw, R.G., 2010. Biparental inbreeding and interremnant mating in a perennial prairie plant: fitness consequences for progeny in their first eight years. Evolution 64, 761–771. Wimberly, M.C., Janssen, L.L., Hennessy, D.A., Luri, M., Chowdhury, N.M., Feng, H., 2017. Cropland expansion and grassland loss in the eastern Dakotas: new insights from a farm-level survey. Land Use Policy 63, 160–173. Winter, M., Johnson, D.H., Shaffer, J.A., Donovan, T.M., Svedarsky, W.D., 2006. Patch size and landscape effects on density and nesting success of grassland birds. J. Wildl. Manag. 70, 158–172. Wright, C.K., 2010. Spatiotemporal dynamics of prairie wetland networks: power-law scaling and implications for conservation planning. Ecology 91, 1924–1930. Wright, C.K., Wimberly, M.C., 2013. Recent land use change in the Western Corn Belt threatens grasslands and wetlands. Proc. Natl. Acad. Sci. U. S. A. 110, 4134–4139. Wright, C.K., Larson, B., Lark, T.J., Gibbs, H.K., 2017. Recent grassland losses are concentrated around US ethanol refineries. Environ. Res. Lett. 12, 044001. Wünscher, T., Engel, S., Wunder, S., 2008. Spatial targeting of payments for environmental services: a tool for boosting conservation benefits. Ecol. Econ. 65, 822–833.
testing connectivity model predictions against empirical migration and dispersal data. Landsc. Ecol. 31, 1419–1432. McConkey, K.R., Prasad, S., Corlett, R.T., Campos-Arceiz, A., Brodie, J.F., Rogers, H., Santamaria, L., 2012. Seed dispersal in changing landscapes. Biol. Conserv. 146, 1–13. McIntyre, N.E., Wright, C.K., Swain, S., Hayhoe, K., Liu, G., Schwartz, F.W., Henebry, G.M., 2014. Climate forcing of wetland landscape connectivity in the Great Plains. Front. Ecol. Environ. 12, 59–64. McRae, B., Kavanagh, D., 2011. Linkage Mapper Connectivity Analysis Software. The Nature Conservancy, Seattle, WA. McRae, B.H., Dickson, B.G., Keitt, T.H., Shah, V.B., 2008. Using circuit theory to model connectivity in ecology, evolution, and conservation. Ecology 89, 2712–2724. Minor, E.S., Urban, D.L., 2008. A graph-theory framework for evaluating landscape connectivity and conservation planning. Conserv. Biol. 22, 297–307. Mushet, D.M., Neau, J.L., Euliss, N.H., 2014. Modeling effects of conservation grassland losses on amphibian habitat. Biol. Conserv. 174, 93–100. Nuñez, T.A., Lawler, J.J., McRae, B.H., Pierce, D.J., Krosby, M.B., Kavanagh, D.M., Singleton, P.H., Tewksbury, J.J., 2013. Connectivity planning to address climate change. Conserv. Biol. 27, 407–416. Pleasants, J.M., Jurik, T.W., 1992. Dispersion of seedlings of the prairie compass plant, Silphium laciniatum (Asteraceae). Am. J. Bot. 133–137. Rabinowitz, D., Rapp, J.K., 1981. Dispersal abilities of seven sparse and common grasses from a Missouri prairie. Am. J. Bot. 616–624. Samson, F.B., Knopf, F.L., Ostlie, W.R., 2004. Great Plains ecosystems: past, present, and future. Wildl. Soc. Bull. 32, 6–15. Swengel, A.B., Swengel, S.R., 2015. Grass-skipper (Hesperiinae) trends in midwestern USA grasslands during 1988–2013. J. Insect Conserv. 19, 279–292. Swengel, S.R., Schlicht, D., Olsen, F., Swengel, A.B., 2011. Declines of prairie butterflies in the midwestern USA. J. Insect Conserv. 15, 327–339.
130