Predicting Corridors of Dispersal Movement for Gopher Tortoise,
Gopherus polyphemus, Under Different Roadway Mortality Scenarios

William W. Hargrove, Forrest M. Hoffman, William D. Meyer, and James D. Westervelt

Introduction

Human development pressure has caused a dramatic loss of wildlife habitat. In Jackson Hole, Wyoming, declines in neo-tropical migrant avian species have been correlated with increasing residential development along the Snake River riparian corridor (Courtland et al. 2005). Amphibian populations have also been affected by habitat destruction and fragmentation and, moreover, are the group with the highest proportion of species threatened with extinction (Stuart et al. 2004; BeeBee and Griffiths 2005). Habitat destruction and fragmentation are often considered to be the most devastating threats to worldwide biological diversity. Land Managers are continually searching for methods to try to accommodate species within declining habitats, particularly where a threatened and endangered species (TES) is present.

Species in habitats isolated as the result of fragmentation often succumb to inbreeding pressures which erode genetic diversity and threaten the prospect of the species' continued survival (Joly et al. 2003). To assist species in fragmented habitats, some have called for wildlife connective corridors between fragmented habitats. Landscape corridors connecting isolated habitat patches would allow species to move freely from one patch to the next, where they may join depleted populations and potentially revive those populations (Brown and Kodric-Brown 1977). Corridors have also been associated with the theories of Island Biogeography (Macarthur and Wilson 1967) and Metapopulation Dynamics (Levins 1969, Hanski and Gilpin 1997). When individuals of a single species are located across many habitat patches with intervening connective corridors, they can disperse, forage, and seek mates in other habitat patches.

What constitutes a suitable connective corridor and how can we predict locations of the best corridors through a particular landscape? A simulation model developed by two of us (Hargrove and Hoffman 2005) attempts to provide the answers to these questions. The model launches virtual entities called "walkers" from identified habitat patches in a GIS data layer simulating their travel as they journey through land cover types in the intervening matrix, and finally arrive at a different habitat "island." Each walker is imbued with a set of user-specified habitat preferences that make its movement behavior resemble a particular animal species. Because the tool operates in parallel on a supercomputer, large numbers of walkers can be efficiently simulated. The Corridor Tool, called the PATH model, uses the concepts of "source" and "sink," a source being the home habitat patch and the sink being the destination habitat patch. The tool can be parameterized from publicly available digital geographic data sources and species knowledge contained in scientific literature.

In December of 2002, Congress created the Army Compatible Use Buffer (ACUB) program to assist installations in slowing urban encroachment on their boundaries which could pose a conflict to the Army training mission. The ACUB program expands the legislative authority of the Private Lands Initiative 10 USC§2684a which allows military departments to partner with government and private organizations to establish buffer areas around active training and testing areas (ACUB website). The ACUB program offers Army natural resource managers a mechanism for identifying lands suitable as connective corridors between on/off-installation TES habitats, which may then be suggested for inclusion in ACUB acquisitions. To identify the most suitable land, we used the PATH model to identify connective habitat corridors for the Gopher Tortoise (Gopherus polyphemus), which occur naturally on the Fort Benning military installation.

The gopher tortoise is historically found in parts of six southeastern states. It is a keystone species within its habitat. Over 330 vertebrate and invertebrate species have been documented as burrow commensuals (Burke 1989). Common associates in many parts of the gopher tortoise's range include other rare species such as eastern indigo snake (Drymarchon couperi), gopher frog (Rana capito) and the sandhill chaffhead (Carphephorus beliidifolius). The original range was associated with open pine forests, especially the longleaf pine (Pinus palustris), where friable soils allowed construction of the tortoise burrows. It is now restricted at the edges of its distribution in South Carolina and Louisiana to only one or two counties or parishes. Large populations are found in Mississippi, Alabama, Georgia and Florida.

Auffenberg and Franz (1982) estimated that gopher tortoise populations have declined by 80% in the last 100 years. This significant decline contributed to the species being listed by the U.S. Fish and Wildlife Service (FWS) as "Threatened" in the western portion of the range (Louisiana, Mississippi, and west of the Tombigbee and Mobile Rivers in Alabama) (Federal Register, July 7, 1987). However, populations are declining throughout the southeast because of habitat destruction, fragmentation, and lack of fire management. Because listing the population east of the Mobile and Tombigbee as threatened could cause training conflicts at locations within the currently non-listed (eastern) population, the gopher tortoise is being studied throughout its range as a part of the Army Threatened and Endangered Species (TES) research program.

Development of the Habitat Map

The study area, about 55 km wide, is located on the northeastern portion of Fort Benning, GA, extending eastward beyond Fort Benning's northeast boundary approximately 30 km (Figure 1). We selected this area to see how well-connected the Gopher Tortoise habitat patches within the installation were to systems of off-base habitat patches on public lands to the east. Thus, the analysis might help target lands for future acquisition via the ACUB process.

Figure 1. Gopher tortoise habitat landscape near Ft. Benning, GA.

The gopher tortoise habitat map was produced from the USGS National Land Cover Data GAP Analysis program data set for the state of Georgia. The NLCD is derived from Landsat TM imagery with a spatial resolution of 30×30m, and has an overall statewide accuracy of 75.46%. Definitions for the 44 classes of land cover are contained in Table 1.

Land Cover Type

Table 1. Land cover types found in the Ft. Benning, GA landscape.

Code	Description
Open sand, sandbars, dunes	7	Open sand, sandbars, mud, and some sand dunes - natural environments as well as exposed sand from dredging and other activities. Mainly in coastal areas, but also inland, especially along the banks of reservoirs.
Coastal Dune	9	Sand dunes and associated vegetation.
Open Water	11	Lakes, rivers, ponds, ocean, industrial water, aquaculture.
Transportation	18	Roads, railroads, airports, and runways.
Utility swaths	20	Open swaths maintained for transmission lines.
Low Intensity Urban - Nonforested	22	Low intensity urban areas with little or no tree canopy.
High Intensity Urban	24	Commercial/industrial and multi-family residential areas.
Clear-cut - Sparse Vegetation	31	Recent clear cuts, sparse vegetation, and other early successional areas.
Quarries, Strip Mines	33	Exposed rock and soil from industrial uses, gravel pits, landfills.
Rock Outcrop	34	Rock outcrops and mountain tops.
Parks, Recreation	72	Cemeteries, playing fields, campus-like institutions, parks, schools.
Golf Course	73	Golf courses.
Pasture, Hay	80	Pasture, non-tilled grasses.
Row Crop	83	Row crops, orchards, vineyards, groves, horticultural businesses.
Forested Urban - Deciduous	201	Low intensity urban areas containing mainly deciduous trees.
Forested Urban - Evergreen	202	Low intensity urban areas containing mainly evergreen trees.
Forested Urban - Mixed	203	Low intensity urban areas containing mixed deciduous and evergreen trees.
Mesic Hardwood	410	Mesic forests of lower elevations in the mountain regions (Blue Ridge, Cumberland Plateau, and Ridge and Valley) and upper Piedmont. Includes species such as yellow-poplar, sweetgum, white oak, northern red oak, and American beech.
Sub-mesic Hardwood	411	Moderately mesic forests of the mountain regions and upper Piedmont. Includes typical oak-hickory forests. The dominant natural cover class in most mountain areas.
Hardwood Forest	412	Mesic to moderately mesic forests of the lower Piedmont and Coastal Plain. Includes non-wetland floodplain forests of yellow-poplar and sweetgum, ravines of oaks and American beech, and many upland oak-hickory stands.
Xeric Hardwood	413	Dry hardwood forests found throughout the state, although most common in the mountain regions, and progressively more rare southward. Includes areas dominated by southern red oak, scarlet oak, post oak, and blackjack oak.
Deciduous Cove Hardwood	414	Mesic forests of sheltered valleys in the Blue Ridge and Cumberland Plateau at moderate to high elevations. Typically includes northern red oak, basswood, buckeye, and yellow-poplar.
Northern Hardwood	415	Restricted to the highest elevations of the Blue Ridge. Dominant tree species may include yellow birch, black cherry, and American beech.
Live Oak	420	Forests dominated by live oak. Most common in maritime strands along the Atlantic Coast. Also may occur in strip along southern border into southwest Georgia.
Open Loblolly-Shortleaf Pine	422	Only mapped in the Piedmont. Includes older, fairly open stands that may be almost savanna-like in appearance.
Xeric Pine	423	Very dry evergreen forests restricted to the mountain regions and upper Piedmont. Includes Virginia, shortleaf, pitch, and table mountain pines.
Hemlock-White Pine	424	Mesic evergreen forests frequently associated with riparian areas. Restricted to Blue Ridge and Cumberland Plateau.
White Pine	425	Moderately mesic evergreen forests of the Blue Ridge, usually dominated by white pine.
Montane Mixed Pine-Hardwood	431	Moderately mesic mixed forests of the Blue Ridge. Typical species include white pine, white oak, hickories, and yellow-poplar.
Xeric Mixed Pine-Hardwood	432	Dry mixed forests found throughout the state, although most common in the mountain regions, and progressively more rare southward. Includes areas dominated by a mix of pines (most frequently shortleaf or Virginia in the mountains, and shortleaf or longleaf elsewhere) and hardwood species such as southern red oak, scarlet oak, post oak, and blackjack oak.
Mixed Cove Forest	433	Mesic mixed forests of sheltered valleys and riparian areas in the Blue Ridge and Cumberland Plateau at moderate to high elevations. Typically includes eastern hemlock, yellow-poplar, and black birch.
Mixed Pine-Hardwood	434	Mesic to moderately dry forests of mixed deciduous and evergreen species found throughout the state at lower elevations. May include areas dominated by sweetgum, yellow-poplar, various oak species, and loblolly or shortleaf pine.
Loblolly-Shortleaf Pine	440	Found from the upper Coastal Plain northward (rare in the Blue Ridge except at the lowest elevations). Includes many stands heavily managed for silviculture as well as areas regenerating from old field conditions.
Loblolly-Slash Pine	441	Found on the lower Coastal Plain. Includes many heavily managed stands as well as a few natural areas.
Shrub Bald	511	Restricted to mountain tops at high elevations of the Blue Ridge. May be dominated by mountain laurel, rhododendron, or blueberry.
Sandhill	512	Areas of scrub vegetation on deep, sandy soils on the Coastal Plain, especially near the Fall Line and along larger streams. May be dominated by turkey oak, blackjack oak, live oak, holly, and longleaf pine.
Coastal Scrub	513	Thickets between coastal dunes, typically dominated by wax myrtle. Sometimes found adjacent to salt marsh areas.
Longleaf Pine	620	Open, savanna-type stands. Heavily managed plantations would likely be classed with 440 or 441. Most common on the lower Coastal Plain, although found up to the lower Piedmont and historically in the Ridge and Valley.
Cypress-Gum Swamp	890	Regularly flooded swamp forests mainly found on the Coastal Plain. May include either riparian or depressional wetlands. Usually dominated by pond or baldcypress and/or tupelo gum.
Bottomland Hardwood	900	Less frequently flooded wetland forests found throughout the state, but most common on the Coastal Plain. To the north, may be dominated by sweetgum, elms, and red maple. To the south, wetland oaks (water oak, willow oak, overcup oak, swamp chestnut oak), black gum, and even spruce pine become more common.
Salt marsh	920	Emergent brackish or saltwater wetlands dominated by Spartina or Juncus.
Freshwater Marsh	930	Emergent freshwater wetlands found throughout the state. May be dominated by grasses or sedges.
Shrub Wetland	980	Closed canopy, low stature woody wetland. Found throughout the state, although most common on the Coastal Plain. May be result of clear cutting of wetland forests. Frequently includes willows, alders, and red maple.
Evergreen Forested Wetland	990	Restricted to the Coastal Plain. Includes forests dominated by bay species, wet pine forests (typically slash or pond pine), or Atlantic white cedar.

Source: U.S. Geological Survey. 1999. Georgia Land Cover Data Set, from The National Land Cover Dataset, Sioux Falls, SD

The following is a list of constraints that were used to develop the habitat landscape for the PATH model. These constraints define species specific characteristics of sub-adult and adult gopher tortoises (assumed to be age 6 or older).

• Minimum Habitat needed to support 50 individuals with 1:1 sex ratio
  • 19 hectares (Eubanks et al. 2002)
• Mean Habitat Core (Eubanks et al. 2002)
  • Female 0.4 to 0.6 hectares
  • Male 1.1 to 1.2 hectares
• Core (Mean Foraging distance from burrow) for Males and Females
  • 30 meters (McRae et al., 1981)
• Home Range Maximum Habitat (Eubanks et al. 2002)
  • Female 3.4 hectares
  • Male 4.8 hectares
• Maximum Traversing Distance
  • Female 375 meters (Smith 1995)
  • Male 474 meters (Diemer 1992)
  • Atypical Long Range movements (0.6 to 2.6 km) recorded over many days (McRae et al. 1981)

We used a parallel version of the Hoshen-Kopelman algorithm to find habitat patches that are spatially contiguous within the Ft. Benning landscape (Figure 2).

Figure 2. Patches large enough to support sustainable gopher tortoise sub-populations near Ft. Benning, GA.

Patches that are spatially contiguous according to an 8-neighbor rule and are larger than the 19 ha minimum size are shown in the same random color. There are 211 patches larger than the minimum size required for sustainable Gopher Tortoise patches; these areas will be used as origins and destinations for potential corridors, and were re-assigned a new habitat category of 1. All smaller patch fragments are shown in red. These fragments remain in the map, retain the same land cover class, and utilize the same GT habitat preferences as the sustainable patches, but are not used as origins and destinations for potential corridors.

A distance analysis was performed based on the longest mean dispersal distance known to have been traversed by Gopher Tortoises (740m, Figure 3). Patches farther from the nearest neighbor patches than this distance are unreachable by Gopher Tortoises, and were not considered as origins or destinations for potential dispersal corridors. Three GT habitat patches (within the red areas) were found to be unreachable by Gopher Tortoises according to this rule, leaving 208 potentially connected patches in the potential corridor analysis.

Figure 3. Distance analysis of patches large enough to support gopher tortoise.

All roads were buffered so that at least 3 road cells must be traversed for each road crossing. This prevents "leakage" of walkers through the cracks formed between one-cell wide roads whose adjacent cells touch only at the diagonal corners (Rothley 2005, Theobald 2005).

Figure 4 shows the pre-processed Ft. Benning Gopher Tortoise habitat landscape. Patches shown in red represent origins and destinations for potential Gopher Tortoise dispersal corridors. Smaller patch fragments are shown in white. Darker areas within the landscape matrix are poorer quality GT habitat, and have lower preference values. Roads and urban areas are shown in black, and are the least preferred habitat categories. Gopher tortoise habitat is most dense on the installation (lower left corner), but substantial GT habitat is also available off-base to the east. The northern half of this landscape is generally less-preferred (darker) than the southern half.

Figure 4. Patchy landscape of gopher tortoise habitat near Ft. Benning, GA.

Parameterization of the PATH Model for Gopher Tortoise

In addition to a land cover map with spatially contiguous patches of each habitat category like Figure 4, the PATH model requires an input matrix of weights representing four types of habitat specific information: preferences for being in each type of habitat, energy costs of movement through each type of habitat category, likelihood of finding food in each type of habitat category, and likelihood of mortality (other than starvation) in each habitat category. Only two of the four possible types of PATH parameters were used here: preference for habitat type and risk of mortality in each habitat type.

Gopher tortoises prefer open-canopied habitats with ample herbaceous ground vegetation for forage, friable soils in which to burrow, and sunlight for egg incubation (Diemer 1986). The tortoise occupies a variety of habitat types but predominantly makes its home on well drained xeric sites with friable soil conditions. Vegetation and soil associations that the Gopher Tortoise has been reported to colonize include Sandhill, Oak Scrub, Sandpine Scrub (Pinus clausa), Southeastern Coastal Plain, Longleaf Pine-Oak (Pinus palustris and Quercus spp.), Xeric Hammock, pine flatwoods, dry prairies, mixed hardwood-pine communities and Ruderal (roadsides, fencerows, utility right-of-ways, pasture edges, clearings and fallow fields) (Auffenberg and Franz 1982, Diemer 1986, Burke 1989).

Using guidance from the published literature, a relative preference value was assigned to each of the 44 classes in Table 1 in terms of its relative suitability as gopher tortoise habitat. While the rank order of preference for the 44 cover classes was usually clear, the assignment of relative numerical values to each cover class was subjectively based on expert opinion. Such quantification is necessary, however, in order to correctly predict potential corridors through the landscape matrix. The classes selected as the most suitable GT habitat included class 620 (Longleaf pine) and 512 (Sandhill) with a preference value of 1 (100%). The remaining classes were given the habitat preference values shown in Table 2.

Land Cover
Class

Table 2. Initial Relative Gopher Tortoise Habitat Preference Parameters Assigned Subjectively using Expertise. No other NLCD categories were present in the Ft Benning, GA landscape.

Land Cover Name	Relative Gopher Tortoise Preference
512	Sandhill	1
620	Longleaf Pine	1
422	Open Loblolly-Shortleaf Pine	0.764
440	Loblolly-Shortleaf Pine	0.764
80	Pasture, Hay	0.593
20	Utility swaths	0.593
31	Clear-cut - Sparse Vegetation	0.593
432	Xeric Mixed Pine-Hardwood	0.537
434	Mixed Pine-Hardwood	0.494
7	Open sand, sandbars, dunes	0.344
900	Bottomland Hardwood	0.312
930	Freshwater Marsh	0.312
413	Xeric Hardwood	0.263
202	Forested Urban - Evergreen	0.18
203	Forested Urban - Mixed	0.18
412	Hardwood Forest	0.18
990	Evergreen Forested Wetland	0.18
980	Shrub Wetland	0.18
890	Cypress-Gum Swamp	0.18
11	Open Water	0.18
22	Low Intensity Urban - Nonforested	0.18
18	Transportation	0.18
24	High Intensity Urban	0.18
33	Quarries, Strip Mines	0.18
201	Forested Urban - Deciduous	0.18
72	Parks, Recreation	0.18
73	Golf Course	0.18
83	Row Crop	0.18

No other NLCD categories were present in the study area.

Gopher Tortoise Roadway Mortality Scenarios

There is scant information about mortality of adult gopher tortoises. The mortality value applied to each land use type was based solely on information derived from the study of adult tortoises. Data for adult mortality was collected from the report (Testudines, Testudinidae) West of the Tombigbee and Mobile Rivers by Lohoefener and Lohmeier (1984). Lohoefener and Lohomeir (1984) considered both natural and human-caused mortality but found that human predation and mortality due to road traffic were the most significant. It was estimated that 4.8% of mature tortoises are killed annually as a result of human predation and 0.8% are killed annually by vehicles in the state of Mississippi.

In order to explore the ramifications on connectance through the GT habitat landscape, we used the PATH model to predict the connectance through this landscape in the absence of roadway mortality, and with seven levels of mortality within the Transportation land cover category (class 18) that spanned five orders of roadway mortality magnitude.

Results

GT habitat patches used as origins and destinations of potential GT dispersal corridors are shown in Figure 5 (below). Each spatially contiguous patch is shown in the same random color.

Figure 5. Gopher tortoise sustainable habitat patches near Ft. Benning, GA

Without roadway mortality, most GT habitat patches in the Ft. Benning landscape are connected to some degree. Warmer colors in Figure 6 indicate greater numbers of successfully dispersing GT individuals passing through this area. GT habitat patches are shown in black.

Figure 6. Potential dispersal corridors predicted for gopher tortoise near Ft. Benning, GA in the absence of roadway mortality.

The strongest potential GT dispersal corridors connecting GT habitat patches on Ft. Benning to habitat patches east of the installation are via higher quality matrix patches at the bottom of the map. That this strongest potential connection is via a southern route, and does not directly involve the large GT habitat patches in the middle of the map is surprising, and might not have been predicted before the PATH simulation was run.

The spatial arrangement of the GT habitat patches in the center of the map look like "stepping stones" that could be used by dispersing GTs to reach from on-base habitats to off-base habitat patches to the east. However, simple examination of the spatial arrangement of these "midway" habitat patches does not consider the quality of the intervening landscape matrix, which consists of less-preferred habitat types between the base and these intervening habitats. Thus, the stronger and more preferred dispersal route is to the south of these ostensible bridging habitat patches.

Figure 7 (below) shows the importance of each habitat patch as the source, or origin of successful dispersals. Darker red shows more important source patches. Most GT habitat patches in this landscape have roughly equal importance as sources, except those at the periphery of the map.

Figure 7. Importance of gopher tortoise sustainable habitat patches as dispersal sources or originations.

The importance of each GT habitat patch can also be calculated as the destination, or endpoint of successful dispersals (Figure 8). Darker red patches represent better "sinks" for dispersing GTs. Habitat patches that are larger, and more central in the map tend to be better sinks (darker red), but such generalizations are tempered by the arrangement of the landscape matrix.

Figure 8. Importance of gopher tortoise sustainable habitat patches as dispersal sinks or destinations.

The cost of efforts made on the part of a land manager to preserve or improve particular GT habitat patches would likely be accrued on a per-unit-area basis. We can change the patch source and sink importance values to importance per unit area to see where management efforts might have the greatest payoff per unit area cost (Figure 9 and Figure 10, respectively). Smaller patches have higher benefits per unit area cost of effort than larger ones, reflecting their general value as bridging "stepping stones". Many of the same small GT habitat patches are identified as important conservation and management targets in both the source-area and sink-area maps.

Figure 9. Area-weighted importance of gopher tortoise sustainable habitat patches as dispersal sources or originations.

Figure 10. Area-weighted importance of gopher tortoise sustainable habitat patches as dispersal sinks or destinations.

Rothley, K. 2005. Finding and filling the "cracks" in resistance surfaces for least-cost modeling. Ecology and Society 10(1):4. http://www.ecologyandsociety.org/vol10/iss1/art4/

Smith, L.L., Nesting Ecology, Female Home Range and Activity, and Population Size-Class Structure of the Gopher Tortoise, Gopherus Polyphemus, on the Katherine Ordway Preserve, Putnam County, Florida, Bulletin, Florida Museum Natural History 37, pt I(4): 97-126.

Smith L.L., Survivorship of Hatchling Gopher Tortoises in North-Central Florida, Proceedings: Conservation, Restoration and Management of Tortoises and turtles, pp 100-103, 1997

Taylor R.W., Human Predation on the Gopher Tortoise (Gopherus polyphemus) In North-Central Florida, Bulletin Florida State Museum, Biological Science 28(4):79-102, 1982

Theobald, D.M. 2005. A note on creating robust resistance surfaces for computing functional landscape connectivity. Ecology and Society 10(2):r1 http://www.ecologyandsociety.org/vol10/iss2/resp1/

Wilson D.S., H. R. Mushinsky, and E. D. McCoy, Home Range, Activity, and Use of Burrows of Juvenile Gopher Tortoises in Central Florida, Fish and Wildlife Research 13:147-160

For additional information contact:

William W. Hargrove

Oak Ridge National Laboratory

Environmental Sciences Division

Building 1507, Room 211

Mail Stop 6407

Oak Ridge, TN 37831-6407

(865) 241-2748

(865) 574-4665 fax

hnw@fire.esd.ornl.gov