Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
146
Vol. 13, Special Issue 5
Modeling Louisiana Pine Snake (Pituophis ruthveni) Habitat
Use in Relation to Soils
Robert O. Wagner1,*, Josh B. Pierce2, D. Craig Rudolph2, Richard R. Schaefer2,
and Dwayne A. Hightower1
Abstract - Ongoing surveys suggest that Pituophis ruthveni (Louisiana Pine Snake) has
declined range-wide and that known extant populations have continued to decline. Seven
known populations remain and occupy small, isolated blocks of habitat. Population sizes
are unknown, but all of them are believed to be critically small. Management for the species’
recovery requires an understanding of its habitat requirements and how resources used
by these snakes are distributed in space. Research suggests that the species’ primary prey,
Geomys breviceps (Baird’s Pocket Gopher), prefers sandy, well-drained soils. Thus soil
attributes may be used to identify potential Louisiana Pine Snake habitat. Using Natural
Resources Conservation Service Soil Survey Geographic Database (SSURGO) soil characteristics
and available historical and recent telemetered snake locations, we developed
resource selection functions describing Louisiana Pine Snake potential habitat at two spatial
scales. SSURGO variable hydgrp, a soil characteristic developed for modeling precipitation
runoff, incorporating soil permeability and depth to ground water, best predicted species
occurrence and was used to map potential habitat selection and identify areas of high conservation
value. Our model demonstrates that the distribution of Louisiana Pine Snakes on
the landscape is strongly influenced by edaphic factors, which are unlikely to be changed
at a landscape scale by human activities. Ample area of suitable soils remains available to
support the species throughout its historical range; however, we suspect that a miniscule
fraction of the potential habitat we identified has suitable vegetation communities on-site
to support Louisiana Pine Snakes.
Introduction
Pituophis ruthveni Stull (Louisiana Pine Snake) historically occupied a limited
range in eastern Texas and west-central Louisiana (Reichling 1995, Rudolph et al.
2006, Sweet and Parker 1991) coincident with the range of Pinus palustris Mill.
(Longleaf Pine) on the west Gulf Coastal Plain (Conant 1956, Reichling 1995,
Thomas et al. 1976). Presumed extirpated from much of its historical range (Rudolph
et al. 2006), the species is now restricted to seven extant populations (one of which
may be recently extirpated) which occupy a limited number of small and fragmented
localities (Reichling 1995; Rudolph et al. 2006; J.B. Pierce, unpubl. data) on federally
and privately owned lands. A limited number of studies have been published that
focus on the ecology of Louisiana Pine Snakes. As a result, it is difficult to develop
landscape-scale habitat models for this species, although these are essential to its
management and conservation.
1Quantitative Ecological Services, 3066 Skyward Way, Castle Rock, CO 80109. 2USDA
Forest Service, Southern Research Station, 506 Hayter Street, Nacogdoches, TX 75965.
*Corresponding author - rwagner@quanteco.com.
Manuscript Editor: Clifford Shackelford
Proceedings of the 5th Big Thicket Science Conference: Changing Landscapes and Changing Climate
2014 Southeastern Naturalist 13(Special Issue 5):146–158
Southeastern Naturalist
147
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014 Vol. 13, Special Issue 5
Louisiana Pine Snake habitat suitability may be largely influenced by the presence
of suitable soils. Sandy well drained soils, in particular, may dictate the
presence and abundance of Louisiana Pine Snake’s main prey Geomys breviceps
Baird (Baird’s Pocket Gopher) (Rudolph et al. 2012), the burrows of which also
provide snakes with shelter from unfavorable environmental conditions (Rudolph
and Burgdorf 1997; Rudolph et al. 1998, 2002). Attempts to model Louisiana Pine
Snake habitat suitability that relied on expert opinion to select soil types based on
perceived suitability yielded models of limited utility. Thus, robust models of Louisiana
Pine Snake habitat selection were not available.
Resource selection occurs when a resource is used disproportionate to its availability
and takes place in a hierarchical fashion at multiple spatial scales from the
species’ physical or geographic range (first-order), to selection of individual home
ranges within the geographic range (second-order), to the animal’s usage of features
within its home range (third-order), to selection of particular elements such
as food items (Johnson 1980). Resource selection functions (RSF) provide a tool
to rank areas by their relative probability of selection (Johnson et al. 2006, Manly
et al. 2002). Extrapolating those relative probabilities in a geographic information
system (GIS) can provide spatially explicit models that can be used for prioritizing
areas for conservation management (Aldridge and Boyce 2007).
Past efforts to locate extant Louisiana Pine Snake populations were based on
proximity to historical records, perceived suitable vegetative structure (i.e., pine
overstory with a sparse midstory and a well-developed herbaceous understory;
Himes et al. 2006, Rudolph and Burgdorf 1997) and soils (Conant 1956, Thomas
et al. 1976). Using those methods, extant populations might remain undiscovered.
Based on the published descriptions of Baird’s Pocket Gopher soil preferences
(Davis et al. 1938), we hypothesized that selection would increase with increasing
sand content and decreasing soil saturation. To examine this hypothesis, we used
edaphic factors (i.e., soil characteristics) to model potential habitat for the Louisiana
Pine Snake. To accomplish this, we analyzed existing location data for this
species to create RSF models for soils across the Louisiana Pine Snake’s historical
range at two spatial scales: second- and third-order, described below. We examined
habitat selection at multiple scales to test our RSF model with independently derived
data (i.e., recent telemetered locations of this species). This approach allowed
us to rigorously vet our model and insure that we made robust inferences about
habitat selection. Our objectives were to 1) develop a robust landscape-scale RSF
that described Louisiana Pine Snake potential habitat, 2) spatially depict potential
habitat selection to identify areas of high probability of selection for focused LPS
habitat management, restoration, potential conservation easement acquisition, and
reintroduction, and 3) provide data to other researchers to identify areas of potential
habitat not previously surveyed for Louisiana Pine Snakes.
Study Area
Our study area consisted of all counties in eastern Texas and parishes in west
central Louisiana that contained three or more Louisiana Pine Snake historical
locations. That area included Angelina, Jasper, Newton, and Sabine counties in
Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
148
Vol. 13, Special Issue 5
Texas, and Bienville, Natchitoches, Rapides, Sabine, and Vernon Parishes in Louisiana.
The study area included all known recently extant populations of Louisiana
Pine Snakes (Rudolph et al. 2006).
Methods
Location data development
Range-wide soils data from the Natural Resources Conservation Service (NRCS)
Soil Survey Geographic database (SSURGO, 2007) were readily available. These
data have been rigorously obtained through standardized methods by the NRCS.
Therefore, they are of sufficient detail, spatial resolution, extent, and attribution to
test hypotheses regarding Louisiana Pine Snake soil preferences.
To model habitat selection at multiple scales, we assembled two datasets of
Louisiana Pine Snake locations: historical data and validation data. Historical
data were the most complete set of range-wide location data available, with
location dates ranging from 1 December 1927 to 23 June 2009. Historical data
contained locations from literature records, museum collections, and our records
of trap-capture sites (established and monitored by the US Forest Service
(USFS), Southern Research Station, Wildlife Habitat and Silviculture Laboratory,
Nacogdoches, TX, and Fort Polk Conservation Branch, Fort Polk, LA),
road kills, and opportunistic sightings, along with similar records from other
researchers (Rudolph et al. 2006; Sweet and Parker 1991; D.C. Rudolph, unpubl.
data). Most locations collected after 1992 were estimated using GPS, which allowed
for acceptable detail. We estimated the specific locations described in
historical accounts from descriptive information contained in the sighting record.
We excluded records that (1) lacked sufficient detail to accurately estimate snake
locations, (2) were potentially misidentifications, and (3) were recapture locations
for the same individual snake. The final dataset consisted of one location for each
of 162 snakes and included records from all known extant populations.
Validation data consisted of 1094 unique telemetry relocations of 22 radio-tagged
Louisiana Pine Snakes made between 1993 and 1997 used in previous studies (Ealy
et al. 2004, Himes et al. 2006, Rudolph and Burgdorf 1997). Locations were distributed
among four Texas counties (Angelina [n = 75], Jasper [n = 50], Newton
[n = 136], and Sabine [n = 223]), and three Louisiana parishes (Bienville [n = 470],
Sabine [n = 14], and Vernon [n = 126]). The number of locations per snake ranged
from 8–130 (25th, 50th, and 75th percentile = 28, 42, and 73, respectively). Telemetry
locations were obtained by tracking snakes at various times throughout the day on
1–7 day intervals, with relocation-site coordinates obtained using post-processed,
differentially corrected GPS (Ealy et al. 2004; Himes et al. 2006; Rudolph and
Burgdorf 1997; J.B. Pierce, unpubl. data).
Second-order selection
Second-order selection contrasts available resources sampled throughout the
study area with a sample of used resources from all animals within the study area,
equivalent to the Manly et al. (2002) sampling protocol A (SP-A), design I. To account
for variable location precision and imprecision in the mapping of edaphic
Southeastern Naturalist
149
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014 Vol. 13, Special Issue 5
factors, we defined used resources as all resources within a 0.25-km radius of historical
snake locations. We considered all resources within a 3-km radius of historical
data snake locations as available. We chose the 3-km radius based on the distribution
of maximum distances among relocations for each snake in the validation dataset. We
accepted the 95th percentile of the maximum distance distribution among snakes as a
reasonable definition of available resources (median, 95th percentile, and maximum =
1.4, 2.9, and 3.6 km, respectively). We used ArcView (ESRI, Redlands, CA) to create
used and available buffer polygons.
We used SSURGO databases and GIS layers to create our initial list of candidate
predictor variables. We accomplished this by extracting from those data
the edaphic factor values that existed within used and available buffer polygons.
The edaphic factors extracted included values from the SSURGO component and
chorizon tables restricting the results to the dominant map unit components (e.g.,
soil series) and H1, or upper soil horizon, respectively. Each map unit (individual
polygons shown on the soil map) represents an area dominated by one to three
kinds of soils or components with individual properties, and each component
has data for each soil horizon in the chorizon table. We included edaphic factors
that appeared to influence Louisiana Pine Snake use, mostly complete within the
SSURGO dataset across the study area, variable among mapped soil units, and
relatively uncorrelated with other candidate factors considered. We evaluated collinearity
among candidate predictor variables using variable clustering (Harrell
2001)—a technique to choose among highly correlated variables and avoid multicollinearity
during model development.
Because few published studies of Louisiana Pine Snake habitat selection
exist, we used expert opinion to develop a competing set of a priori RSF models
(n = 26) from the set of candidate predictor variables (Appendix 1). When
developing models, we chose variables so that all competing models were biologically
supportable while avoiding inclusion of collinear variables and over-fitting
(maintaining ≥15 used locations per predictor degree of freedom; Harrell 2001).
To accommodate both continuous and categorical predictors, competing RSFs
were structured as weighted logistic regression models with weights proportional
to soil types within each location’s used and available buffer. We used Akaike’s
information criterion adjusted for small sample size (AICc; Burnham and Anderson
1998) to rank the set of a priori models and when ΔAICc among models was
<2, we selected the most parsimonious model (Arnold 2010, Burnham and Anderson
1998). We used the coefficients of the selected ΔAICc model to estimate the
relative probability of selection by Louisiana Pine Snakes (Johnson et al. 2006,
Manly et al. 2002) and used 95% confidence intervals to evaluate differences
among categories. Using the relative selection probabilities, we estimated more
readily interpreted selection indices (wi; Manly et al. 2002). Selection indices >1
indicate that a resource was used in greater proportion than available, wi not different
than 1 suggests use was in proportion to availability, and wi < 1 indicates
use was less than available (Manly et al. 2002). We classified wi > 1 as preferred,
not different from 1 as suitable, and <1 as avoided.
Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
150
Vol. 13, Special Issue 5
We conducted post-hoc analyses to test for additive and interaction effects between
the selected second-order RSF model variables and area. This was completed
to determine if the large number of locations from the Bienville population, the
largest extant population known, biased the results. We created a new categorical
variable, area, to distinguish used and available resources in Bienville Parish, LA,
from those within the balance of the study area. Two new models were created from
the selected second-order model by adding area as an additive and interaction term.
We then compared model parsimony using AICc.
Third-order selection and validation
Third-order selection contrasts used and available resources sampled for each
animal (SP-A, Design III; Manly et al. 2002). We used the validation data to investigate
third-order selection. Unlike the historical data, the validation data were
precise locations; thus, we accepted the point estimates as our definition of used
resources. We defined available resources for each snake as all resources within
composite 564-m radius buffers around the locations for each snake. A 564-m
radius encompasses the estimated 100-ha minimum multi-year home range for
Louisiana Pine Snakes (J.B. Pierce, unpubl. data). As in the second-order analysis,
we extracted the edaphic factor values for the used locations and available resource
buffers using GIS. However, because our goals were to contrast hierarchical selection,
examine selection variability among animals, and validate the second-order
model, we only extracted the variable included in the selected second-order model.
The selected second-order model included only a single categorical variable
simplifying third-order RSF modeling and second-order validation. We examined
log-likelihoods of used versus available proportions to estimate third-order RSFs
using methods described by Manly et al. (2002) for studies with resources defined
by several categories. We tested the null hypothesis that no selection across resource
unit categories occurred using likelihood-ratio tests, both within and across
animals. If selection occurred (at α < 0.05), then we used likelihood-ratio tests to
determine if selection indices (wi) were different from 1 (at Bonferoni adjusted
α < 0.05). We classified third-order selection indices as preferred, suitable, or
avoided, as we did with the second-order model results.
To validate the second-order model, we relied on used and available resource
estimates from the averaged third-order selection evaluation. We averaged used and
available resource estimates across all radio-tagged animals in the validation dataset
to approximate second-order selection, and compared these validation model
results to the selected second-order model results. Following model validation, we
mapped potential habitat throughout the species’ historical range by linking RSF
results with the SSURGO database and GIS layer for each county and parish.
Results
Second-order selection
Based on evaluations of completeness, variability, and collinearity, we identified
seven candidate edaphic factors from the SSURGO data that were suitable
Southeastern Naturalist
151
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014 Vol. 13, Special Issue 5
predictor variables for development of a priori RSF models (Table 1). Of 26 models
considered, the best-approximating model describing second-order Louisiana Pine
Snake potential habitat selection included only hydgrp (Appendix 1), a categorical
soil characteristic variable developed for modeling precipitation runoff, incorporating
soil permeability and depth to ground water. Because the best model included
only a single categorical variable and was developed from the proportions of used
and available resources at each location, wi was the proportion of used resources
(PUsed) divided by the proportion of available resources (PAvialable) within resource
unit i (Table 2).
Based on our wi estimates (Table 2), hydgrp category A was preferred and
differed statistically from all other categories (depth to ground water and soil permeability
decrease from hydgrp category A, to A/D, to C, to D). Categories B and C
were not statistically different than 1 and were thus classified as suitable, although
the confidence interval for Category C was large. Category D was avoided. These
results strongly support our hypothesis that selection increases with increasing
sand content and decreasing soil saturation. Data were insufficient to classify the
Table 2. Model results for second-order Louisiana Pine Snake potential habitat selection and percent
of available (PAvailable) and used (PUsed) habitat based on the historical data. βi is the model estimated
coefficients for the ith hydgrp, and SE(βi) its standard error. The selection index is wi, estimated as
exp(βi), and wi CI95 its 95% confidence interval.
Hydgrp
A A/D B C D
PAvailable 28.1% 0.2% 31.3% 12.7% 27.7%
PUsed 47.2% 0.2% 33.5% 7.9% 11.1%
βi 0.517 -0.011 0.069 -0.470 -0.911
SE(βi) 0.187 2.303 0.195 0.356 0.279
wi 1.68 0.99 1.07 0.62 0.40
wi CI95 1.16–2.42 0.01–90.46 0.73–1.57 0.31–1.26 0.23–0.69
Table 1. Candidate variables used to develop competing resource selection functions describing
second-order Louisiana Pine Snake potential habitat selection. Descriptions adapted from attribute
descriptions in SSURGO Metadata—Table Column Descriptions; SSURGO Metadata Version: 2.2.3
(obtained from http://soildatamart.nrcs.usda.gov/SSURGOMetadata.aspx on 5 December 2007).
SSURGO ID (table.field name) df Description
muaggatt.floodfreqdcd 4 Annual flooding probability class
component.drainagecl 7 Classes based on drainage, flood frequency and
duration
chorizon.sandtotal_r 1 % sand
chorizon.claytotal_r 1 % clay
chorizon.om_r 1 % organic matter
component.taxorder 6 Highest soil taxonomy level; e.g., Entisols, Ultisols
component hydgrp 5 Hydrologic group consisting of classes of soils
having similar runoff potential based on depth to a
seasonally high water table, and soil permeability
Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
152
Vol. 13, Special Issue 5
suitability of hydgrp category A/D. The A/D wi confidence interval was so wide that
the estimate was considered unreliable. The imprecision in the A/D selection likelihood
was a function of its rarity within the study area (<0.2% of the study area).
We found no evidence that the habitat selection results were influenced by the
large number of locations from the Bienville population or differed between Bienville
and the balance of the study area. Additive or interaction effects between
hydgrp and area were not supported by the data (ΔAICc hydgrp x area = 7.5;
ΔAICc hydgrp + area < 2). The additive model ΔAICc was <2, but differed from
the selected second-order model with hydgrp alone by 1 parameter and had essentially
the same deviance. Therefore, there was no support for the more complex
additive model and the more parsimonious selected second-order model was the
best model (Arnold 2009, Burnham and Anderson 1998).
Third-order selection and validation
Used and available resources varied among snakes resulting in differences in
resource selection among animals (Table 3). Most snakes (19 of 22; 86%) demonstrated
selection across hydgrp categories, but only hydgrp A was preferred. Of
those snakes demonstrating selection, hydgrp category A was available to 18 and
was preferred by 13 (72%); no snakes avoided hydgrp A. The remaining hydgrp categories
were avoided by the following percentages of snakes for which the category
was available: category B 15% (2 of 13), category C 38% (5 of 13), and category D
53% (10 of 19).
Based on the across-snake validation model estimates of wi, we obtained results
similar to those from the second-order model with hydgrp A preferred, B suitable,
and D avoided. However, unlike the second-order model, the validation model estimate
of wi for hydgrp C had sufficient power to determine that snakes avoided C.
Based on validation results, we classified hydgrp A as preferred, B as suitable, and
C and D as avoided (Fig. 1). Category A/D did not occur within the validation data
available area and thus, we could not evaluate the selected second-order suitable
classification. Because A/D represented an insignificant portion of three parishes
within the species’ historical range (Bienville, Vernon, and Winn), we used the
second-order suitable classification for mapping purposes, but the resulting map
should be interpreted with caution.
Discussion
The SSURGO database variable hydgrp, developed for modeling precipitation
runoff, incorporated the factors that we believed most influenced Louisiana
Pine Snake potential habitat selection: percent sand and depth to ground water.
Percent sand was highest in hydgrp category A and decreased in each subsequent
class A/D through D with corresponding increases in variability (Fig. 2). As we
hypothesized, the likelihood of Louisiana Pine Snake use increased with increasing
percent sand and depth to water table. Hydgrp, which incorporated both
components, explained a greater fraction of the variance among snake locations
than percent sand alone.
Southeastern Naturalist
153
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014 Vol. 13, Special Issue 5
Table 3. Number of telemetry locations per snake (n), and percent of used (PUsed) and available (PAvailable) area, selection index (wi), and selection classification
by snake and hydgrp for radio-tagged Louisiana Pine Snakes. Blank cells indicate hydgrp i was not available.
PUsed PAvailable wi SelectionB
hydgrp hydgrp hydgrp hydgrp
Snake n A B C D A B C D A B C D A B C D
1 45 1.00 0.00 0.00 0.00 0.61 0.26 0.05 0.08 1.63 0.00 0.00 0.00 + -
2 82 0.72 0.16 0.12 0.52 0.23 0.25 1.39 0.69 0.48 + -
3 72 0.47 0.00 0.51 0.01 0.26 0.04 0.59 0.11 1.79 0.00 0.88 0.12 + -
4 83 0.72 0.28 0.00 0.00 0.54 0.19 0.14 0.13 1.34 1.43 0.00 0.00 + - -
5 130 0.98 0.02 0.00 0.00 0.57 0.31 0.05 0.08 1.73 0.08 0.00 0.00 + - - -
6 14 0.50 0.07 0.43 0.15 0.20 0.65 3.26 0.36 0.66 +
7 27 0.89 0.11 0.00 0.00 0.51 0.25 0.13 0.11 1.75 0.45 0.00 0.00 +
8 73 0.92 0.04 0.04 0.52 0.25 0.23 1.77 0.16 0.18 + - -
9 52 0.23 0.77 0.00 0.13 0.75 0.12 1.81 1.02 0.00 -
10A 25 0.48 0.28 0.24 0.00 0.44 0.26 0.22 0.08 1.10 1.09 1.08 0.00
11 8 1.00 0.00 0.00 0.63 0.26 0.11 1.59 0.00 0.00
12 37 0.73 0.00 0.27 0.41 0.13 0.46 1.79 0.00 0.58 +
13A 9 0.78 0.00 0.22 0.62 0.13 0.25 1.26 0.00 0.88
14 88 0.40 0.22 0.35 0.03 0.46 0.15 0.29 0.10 0.86 1.49 1.21 0.34
15 41 0.90 0.00 0.05 0.05 0.52 0.01 0.28 0.19 1.73 0.00 0.18 0.25 + -
16 43 0.81 0.19 0.00 0.76 0.10 0.14 1.07 1.87 0.00 -
17 51 0.02 0.98 0.00 0.03 0.87 0.10 0.57 1.13 0.00
18 93 0.86 0.03 0.11 0.47 0.22 0.31 1.83 0.15 0.35 + - -
19 31 0.97 0.00 0.03 0.78 0.00 0.22 1.24 0.00 0.15 + -
20A 34 0.76 0.00 0.24 0.78 0.01 0.22 0.99 0.00 1.08
21 33 1.00 0.00 0.00 0.72 0.10 0.18 1.39 0.00 0.00 + -
22 23 1.00 0.00 0.92 0.08 1.09 0.00
Overall 1094 0.69 0.19 0.06 0.07 0.47 0.21 0.13 0.19 1.48 0.84 0.52 0.30 + - -
ALikelihood ratio test for resource selection was not significant (at α = 0.05).
BIf selection occurred and wi values were different than 1 (Bonferoni adjusted α < 0.05), a + indicates wi > 1 (which supports selection) and - indicates wi
< 1 (which supports avoidance).
Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
154
Vol. 13, Special Issue 5
Despite known differences in vegetation structure across snake locations used
in model development, our successful modeling of Louisiana Pine Snake secondand
third-order potential habitat selection suggested that edaphic factors may also
influence first-order selection. Although we did not have the data necessary to critically
examine first-order selection, as an exploratory analysis we estimated hydgrp
proportions within the counties and parishes throughout the snake's historical range
(A = 7.9%, B = 31.7%, C = 24.6%, and D = 35.8%). We replaced the estimated
available habitat in the selected second-order selection RSF with the range-wide
available estimates to estimate hypothetical first-order selection indexes by hydgrp.
Compared to the selected second-order selection results, preference for hydgrp A
increased (wi = 5.97), B and D remained approximately the same, and avoidance
of C increased (wi = 0.32). Although only an approximation of first-order potential
habitat selection, these results together with the second- and third-order results suggest
that soils strongly influenced the species’ historical range.
Encouraged by our modeling and validation results, we distributed the maps
of potential habitat (Fig. 1) to researchers and managers responsible for Louisiana
Pine Snake conservation. Those maps were used to delineate the boundary
of a proposed conservation area for the Bienville population, which provided a
Figure 1. Potential Louisiana Pine Snake habitat predicted, based on validation model estimates
for hydgrp categories.
Southeastern Naturalist
155
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014 Vol. 13, Special Issue 5
target area for the acquisition of protective easements by conservation organizations.
The maps were also used to focus habitat management of federal lands,
identify reintroduction sites outside of extant populations, and assess threats
to extant populations. We plan to use the results of this study to quantify the spatial
extent and location of potential habitat on federal lands and determine if there are
areas of suitable habitat that have not been adequately surveyed for Louisiana Pine
Snake occurrence. When available, these results will be provided to the US Fish
and Wildlife Service for consideration.
Our models focused on edaphic factors unlikely to be changed at a landscape
scale by human activities and thus were useful for identifying potential Louisiana
Pine Snake habitat. Suitable habitat consists of potential habitat with an appropriate
vegetative cover. Louisiana Pine Snake declines have been attributed to a loss of
suitable habitat associated with loss of the Longleaf Pine ecosystem, due largely to
conversion to short-rotation Pinus taeda L. (Loblolly Pine) plantations and exclusion
of frequent fire. We suspect that a miniscule fraction of the potential habitat
we identified has vegetation communities on-site suitable to support Louisiana
Pine Snakes. In addition to the presence of suitable soils, increasing the acreage
Figure 2. The percent sand (sandtotal.r) within hydrgp categories. The box represents the
inner quartile range (25th to 75th percentiles), and upper and lower whiskers extending
from the box represent the smallest and largest observations within one step (1.5 times inner
quartile range). The median (♦) is marked by a line through the box, and horizontal bars
(—) represent extreme values.
Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
156
Vol. 13, Special Issue 5
of Longleaf Pine communities on areas with these soils is likely needed for species
recovery. We believe that this can be achieved through the reestablishment of
appropriate timber stocking and fire regimes within and adjacent to extant populations.
A more detailed understanding of the vegetation communities required to
support this species is a topic worthy of future research.
Acknowledgments
We thank S. Shiveley, B. Gregory, R. Carrie, T. Trees, J. Helvey, J. Tull, R. Johnson, D.
Baggett, P. Taylor, T. Johnson, W. Ledbetter, K. Moore, K. Mundorf, E. Keith, C. Collins,
R. Maxey, J. Himes, M. Ealy, S. Burgdorf, M. Duran, C. Melder, and others for their assistance
in collecting and assembling the data used in this study. We thank NRCS Louisiana
Assistant Soil Scientist C. Guillory for assistance identifying candidate SSURGO variables.
Temple-Inland, Inc., International Paper Company, The Nature Conservancy, Department
of Defense, Mill Creek Ranch, Texas Parks and Wildlife Department, and Champion International
provided access to study sites. The Joint Readiness Training Center and Ft. Polk
Conservation Branch, Texas Parks and Wildlife Department, US Fish and Wildlife Service,
Temple-Inland, Inc., Louisiana Department of Wildlife and Fisheries, and US Forest Service
provided funding for this work. We also thank two anonymous reviewers for their
comments that improved this manuscript.
Literature Cited
Aldridge, C.L., and M.S. Boyce. 2007. Linking occurrence and fitness to persistence: A
habitat-based approach for endangered Greater Sage-grouse. Ecological Applications
17:508–526.
Arnold, T.W. 2010. Uninformative parameters and model selection using Akaike's information
criterion. Journal of Wildlife Management 74:1175–1178.
Burnham, K.P., and D.R. Anderson. 1998. Model Selection and Inference: A Practical
Information-Theoretic Approach. Springer-Verlag, New York, NY. 353 pp.
Conant, R. 1956. A review of two rare pine snakes from the Gulf Coastal Plain. American
Museum Novitates 1781:1–31.
Davis, W.B., R.R. Ramsey, and J.M. Arendale, Jr. 1938 Distribution of Pocket Gophers
(Geomys breviceps) in relation to soils. Journal of Mammalogy 19:412–418.
Ealy, M.J., R.R. Fleet, and D.C. Rudolph. 2004. Diel activity patterns of the Louisiana Pine
Snake (Pituophis ruthveni) in eastern Texas. Texas Journal of Science 56:383–394.
Harrell, F.E., Jr. 2001. Regression Modeling Strategies with Applications to Linear Models,
Logistic Regression, and Survival Analysis. Springer-Verlag, New York, NY. 568 pp.
Himes, J.G., L.M. Hardy, D.C. Rudolph, and S.J. Burgdorf. 2006. Movement patterns and
habitat selection by native and repatriated Louisiana Pine Snakes (Pituophis ruthveni):
Implications for conservation. Herpetological Natural History 9:103–116.
Johnson, D.H. 1980. The comparison of usage and availability measurements for evaluating
resource preference. Ecology 6:65–71.
Johnson, C.J., S.E. Nielsen, E.H. Merrill, T.L. McDonald, and M.S. Boyce. 2006. Resource
selection functions based on use-availability data: Theoretical motivation and evaluation
methods. Journal of Wildlife Management 70:347–357.
Manly, B.F.J, L.L. McDonald, D.L. Thomas, T.L. McDonald, and W.P. Erickson. 2002. Resource
Selection by Animals, Statistical Design, and Analysis for Field Studies, Second
Edition. Kluwer Academic Publishers, The Netherlands. 221 pp.
Southeastern Naturalist
157
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014 Vol. 13, Special Issue 5
Natural Resources Conservation Service (NRCS). 2007. Soil survey data from the Soil
Survey Geographic (SSURGO) Database. Available online at http://soildatamart.nrcs.
usda.gov. Accessed 5 December 2007.
Reichling, S.B. 1995. The taxonomic status of the Louisiana Pine Snake (Pituophis melanoleucus
ruthveni) and its relevance to the evolutionary species concept. Journal of
Herpetology 29:186–198.
Rudolph, D.C., and S.J. Burgdorf. 1997. Timber Rattlesnakes and Louisiana Pine Snakes:
Hypotheses of decline. Texas Journal of Science 49:111–122.
Rudolph, D.C., S.J. Burgdorf, J. Tull, M. Ealy, R.N. Conner, R.R. Schaefer, and R.R. Fleet.
1998. Avoidance of fire by Louisiana Pine Snakes, Pituophis melanoleucus ruthveni.
Herpetological Review 29:146–148.
Rudolph, D.C., S.J. Burgdorf, R.N. Conner, C.S. Collins, D. Saenz, R.R. Schaefer, T. Trees,
C.M. Duran, M. Ealy, and J.G. Himes. 2002. Prey handling and diet of Louisiana Pine
Snakes (Pituophis ruthveni) and Black Pine Snakes (P. melanoleuells lodingi), with
comparisons to other selected Colubrid snakes. Herpetological Natural History 9:57–62.
Rudolph, D.C., S.J. Burgdorf, R.R. Schaefer, R.N. Conner, and R.W. Maxey. 2006. Status
of the Louisiana Pine Snake, Pituophis ruthveni. Southeastern Naturalist 5:463–472.
Rudolph, D.C., C.A. Melder, J.B. Pierce, R.R. Schaefer, and B. Gregory. 2012. Diet of the
Louisiana Pine Snake (Pituophis ruthveni). Herpetological Review 43:243–245.
Sulentich, J.M., L.R. Williams, and G.N. Cameron. 1991. Geomys breviceps. Mammalian
Species 383:1–4.
Sweet, S.S., and W.S. Parker. 1991. Pituophis melanoleucus. Catalogue of American Amphibians
and Reptiles 474:1–8.
Thomas, R.A., B.J. Davis, and M.R. Culbertson. 1976. Notes on variation and range of
the Louisiana Pine Snake, Pituophis melanoleucus ruthveni Stull (Reptilia, Serpentes,
Colubridae). Journal of Herpetology 10:252–254.
Southeastern Naturalist
R.O. Wagner, J.B. Pierce, D.C. Rudolph, R.R. Schaefer, and D.A. Hightower
2014
158
Vol. 13, Special Issue 5
Appendix 1. A priori RSF candidate models for historical Louisiana Pine Snake locations.
fl = flodfreqded, dr = drainagecl, sa = sandtrotal.r, cl = claytotal.r, and om = om.r.
Variable
Model fl dr sa cl om taxorder hydgrp K ΔAICc wi
1 x x x x 7 12.29 0.001
2 x x x 6 12.79 0.001
3 x x x 9 17.06 0.000
4 x x x 9 6.66 0.017
5 x x x x 10 14.53 0.000
6 x x 8 11.03 0.002
7 x x 11 15.41 0.000
8 x x 11 11.40 0.002
9 x x x x 8 10.89 0.002
10 x x x x 8 5.62 0.029
11 x x x 7 10.47 0.003
12 x x x 7 11.20 0.002
13 x x x 7 3.56 0.081
14 x x x 7 3.53 0.082
15 x x x 10 8.14 0.008
16 x x 9 6.51 0.018
17 x x 6 10.93 0.002
18 x x 6 1.50 0.227
19 x x 5 11.72 0.001
20 x x x 4 7.08 0.014
21 x 4 12.66 0.001
22 x 7 9.12 0.005
23 x 2 6.58 0.018
24 x 5 10.03 0.003
25 x 5 0.00 0.479
Null 1 11.42 0.002