Parent et al. JWM 2015 - Distance Sampling Workshop

The Journal of Wildlife Management; DOI: 10.1002/jwmg.992
Research Article
Northern Bobwhite Abundance in Relation to
Precipitation and Landscape Structure
CHAD J. PARENT,1,2 Caesar Kleberg Wildlife Research Institute, Texas A&M University Kingsville, Kingsville, TX 78363, USA
FIDEL HERNANDEZ,
Caesar Kleberg Wildlife Research Institute, Texas A&M University Kingsville, Kingsville, TX 78363, USA
LEONARD A. BRENNAN, Caesar Kleberg Wildlife Research Institute, Texas A&M University Kingsville, Kingsville, TX 78363, USA
DAVID B. WESTER, Caesar Kleberg Wildlife Research Institute, Texas A&M University Kingsville, Kingsville, TX 78363, USA
FRED C. BRYANT, Caesar Kleberg Wildlife Research Institute, Texas A&M University Kingsville, Kingsville, TX 78363, USA
MATTHEW J. SCHNUPP, King Ranch, Inc., Kingsville, TX 78363, USA
ABSTRACT Northern bobwhite (Colinus virginianus) populations have declined across their range. The
decline is associated with broad-scale losses of their habitats. Additionally, the presence of essential,
structural features provided by vegetation in the remaining habitats is contingent on variable spatial and
temporal trends in precipitation. This complicates the management of the bobwhite’s habitats. We modeled
counts of bobwhite coveys as a function of landscape structure and precipitation covariates from arid
landscapes in southern Texas. Our results indicated that numbers of coveys in landscapes with greater
amounts of woody cover were predicted to be highly independent of precipitation. This has important
management implications because certain landscape structures associated with woody cover buffer bobwhite
populations from drought. To facilitate management based on our results, we mapped our model predictions
for covey counts. This allows managers to spatially prioritize where management interventions need to occur,
and evaluate the potential efficacy for these interventions to create positive bobwhite population responses.
Ó 2015 The Wildlife Society.
KEY WORDS bobwhite, brush, drought, ecology, FRAGSTATS, landscape, landscape structure, management,
northern bobwhite, precipitation, shrubland.
Managers of northern bobwhite (Colinus virginianus) are
beneficiaries of decades of prior research (Stoddard 1931,
Rosene 1969, Roseberry and Klimstra 1984, Guthery 2002,
Hernandez and Guthery 2012) and applied this knowledge
to stabilize numerous bobwhite populations at local, small
scales. Despite localized success, range-wide populations
of bobwhites have experienced a long-term decline since at
least the mid-1960s (Sauer et al. 2011). Broad-scale loss
and deterioration of habitats are primary causes for the
population declines of bobwhites in different ecosystems
(Brady et al. 1998, Veech 2006, Lohr et al. 2011, Blank
2012, Hernandez et al. 2012). These studies indicate the
factors limiting populations operate at large spatial scales.
Therefore, if conservation is to be effective throughout the
bobwhite’s range, then applications of some management
actions need to be applied at larger spatial scales. This
entails understanding how effects of weather and the
Received: 8 June 2014; Accepted: 17 August 2015
1
E-mail: [email protected]
Present address: Department of Fisheries and Wildlife, Michigan State
University, 480 Wilson Rd., East Lansing, MI 48824, USA.
2
Parent et al.
Mapping Northern Bobwhite
landscape structure of cover types, together, operate at
large scales.
The effect of weather on bobwhites in arid ecoregions is
well understood. For example, a variety of different
precipitation data have been used to draw correlative or
linear relationships with different bobwhite population
indices and estimates, collected at small and large scales
(Rice et al. 1993, Bridges et al. 2001, Guthery et al. 2002,
Lusk et al. 2002, Tri et al. 2013). These studies
demonstrated significant, positive links between the
quantity of precipitation and the direction of bobwhite
population metrics. These relationships are strongly tied to
the timing and amount of precipitation (Peterson 2001),
which varies regionally across space, and can vary
remarkably from year to year. Moreover, severe droughts
followed immediately by tropical storms that deliver aboveaverage precipitation to arid parts of Texas along the Gulf
Coast strongly influence bobwhite survival and reproduction (proportion nesting, nesting rate, and nesting season
length; Hernandez et al. 2002, 2005), and are increasingly
expected to be the norm as climate in this region changes
(Norwine et al. 2007). Presumably, periods with minimal
precipitation can reduce the amount of resources needed to
carry out nesting, foraging, and brood-rearing activities,
which can potentially create an ecological bottleneck
1
(Wiens 1977), a condition under which severe weather
limits resources, and by extension, species abundances
(Williams and Middleton 2008).
Understanding composition and configuration (Li and
Reynolds 1994) of land cover types on the landscape provides
a means to formally quantify the surface of a landscape and
permits comparisons in space and time. For example,
landscapes with diverse composition of seral stages required
by bobwhite have the capacity to influence estimates of
abundance (Schairer et al. 1999, Veech 2006, Howell et al.
2009, Blank 2012). The configuration of cover types with
these studies influences bobwhite populations less, but each
study notes variation depends on the spatial scale at which
these land cover configurations are quantified (i.e., scaledependence; Howell et al. 2009, Duren et al. 2011).
Understanding how landscape structure influences bobwhites on landscapes in arid ecoregions would provide
information about how bobwhite population abundances
are spatially structured and how to prioritize and
strategically place management interventions to habitats
in areas with the greatest efficacy to benefit bobwhites.
Additionally, if we know drought is a limiting factor to
bobwhite populations through the creation of ecological
bottlenecks, there are management implications associated
with identifying landscape structures that can buffer
bobwhite populations from drought. This requires a
more thorough investigation into the composition and
configuration of land covers that yield positive relationships
with bobwhite population abundance, and how such
relationships change with the quantity of precipitation.
Our objectives were 3-fold: describe the timing of
precipitation related to northern bobwhite abundance
during a prolonged drought, determine how landscape
structure influences northern bobwhite populations on arid
landscapes, and understand how northern bobwhite
populations respond to landscape structure with variable
trends in precipitation. Finally, we wished to demonstrate
the utility in mapping spatially explicit relationships.
The ability to prioritize where to apply habitat management on the landscape given recent trends in precipitation
would be an extremely valuable tool to bobwhite
managers in arid ecoregions. Thus, an additional objective
was to map predictions from our models for managers to
prioritize and initiate management of northern bobwhite
habitat.
STUDY AREA
The study area consisted of 333,866 ha and was comprised
by the King Ranch (King Ranch, Kingsville, TX; Fig. 1).
The ranch spanned 6 counties in Texas: Brooks, Jim Wells,
Kenedy, Kleberg, Nueces, and Willacy. The ranch was
approximately bound between the south and north latitudes
from 26.68 to 27.78 and approximately between the east and
west longitudes from 97.58 to 98.18, respectively.
Average elevation was 16 m (SD ¼ 12 m) with gentle slopes
that increased from <1 m along the Gulf Coast to
occasional peaks of 52 m on western parts of the ranch.
The ranch comprised 4 divisions (Encino, Laureles, Norias,
2
Santa Gertrudis). All divisions were located in the Rio
Grande Plains ecoregion (Gould 1975); however, the
eastern edges of the Laureles and Norias divisions were
located in the Gulf Prairies and Marshes ecoregion (Gould
1975). Divisions were further subdivided into 122
management units (i.e., fenced pastures), many of which
were managed for white-tailed deer (Odocoileus virginianus)
and bobwhites. Management units on the King Ranch were
sufficiently large to contain multiple populations of
bobwhite; the average size of each unit was 2,023 ha,
which could support 135 coveys/unit assuming the average
covey home range on our study area was 15 ha (Haines et al.
2004) and usable space was saturated in space-time
(Guthery 1997). Thus, we defined each management
unit as a landscape because they were ecologically capable of
supporting multiple populations of bobwhites. Climate was
semiarid and subtropical with prolonged periods of
drought. Average annual precipitation (1971–2000) was
44.7 cm, and average annual temperature was 21.78 C
(http://www.ncdc.noaa.gov/IPS/coop; Kingsville and Falfurrias weather stations). Average annual precipitation
during the duration of our study was 54.96 cm (SD
¼ 22.26), but much of this precipitation occurred during a
wet period in 2010. Aside from 2010, our study area
experienced prolonged, severe drought conditions based on
the Palmer Drought Severity Index, particularly in 2009,
2011, and 2012 (see Fig. S1, available online at www.
wildlifejournals.org). Major land uses on the ranch were
cattle grazing, petroleum extraction, and fee-lease hunting
(Fulbright and Bryant 2002). Major vegetation communities occurring on the study area included bluestem prairie
(Schizachyrium scoparium), mesquite-granjeno thornbrush
(Prosopis glandulosa- Celtis pallida), mesquite-bluestem
savannah, oak-bluestem (Quercus virginiana, Quercus
stellata) woodland, mesquite savanna, and cordgrassbluestem (Spartina foliosa; McLendon 1991, Fulbright
and Bryant 2002).
METHODS
Covey Counts
We performed bobwhite relative abundance surveys in
conjunction with a general wildlife survey that also counted
white-tailed deer, wild turkeys (Meleagris gallopavo), feral
pigs (Sus scrofa), and other wildlife on the ranch. We
performed these surveys by helicopter using a survey design
based on a protocol to estimate bobwhite density at our
study area via distance sampling used by Rusk et al. (2007)
and Schnupp et al. (2012). Importantly though, compared
to the general wildlife surveys, the altitude and speed of
the helicopter during distance sampling was usually slower
and lower to ensure detection of covey flushes, and so
variation in covey flushes may exist between general wildlife
surveys. We recorded the number of bobwhite coveys
flushed (i.e., individual flocks) and survey effort (transect
length; km).
Data from the general wildlife surveys were suitable for
indexing bobwhite population abundance as long as the
The Journal of Wildlife Management
9999
Figure 1. The study area is located in southern Texas, USA on 333,866 ha of the King Ranch in Kingsville, Texas. The ranch comprises 4 divisions and within
each division are 122 landscapes.
relationship between the numbers of coveys flushed per
survey effort changed proportionally with population
abundance. A critical assumption then is that detection is
constant over space and time. Violations of this assumption
would suggest differential detection probabilities along a
transect depending on vegetation. We do not believe
variation in vegetation across space was a significant source
of bias in our general wildlife surveys because our sampling
took place during early morning hours after bobwhites leave
their roosts to forage in open, herbaceous vegetation
(Stoddard 1931); in essence there is no spatial heterogeneity
in the cover they were using during the periods we performed
our surveys. However, it is possible that variation in land uses
through time influenced detection of bobwhite covey flushes.
Because changes to herbaceous vegetation due to grazing,
burning, or any other means that could change the
composition and configuration of such covers accumulate
over time, detection between these periods may not have
been constant. We speculate that because the timing and
amount of precipitation is an unpredictable process in arid
regions such as ours (Norwine et al. 2007), management
activities that could alter the structure of the vegetation
Parent et al.
Mapping Northern Bobwhite
(grazing, burning) would be tied to precipitation. Our study
area received extensive drought during these survey periods
and accordingly, very little grazing and burning activities
occurred. The structure of the vegetation did not change
appreciably through time to influence detection. Moreover,
previous unpublished research has indicated that bobwhite
population indices from general wildlife surveys change
proportionally with distance-based helicopter surveys (M. J.
Schnupp, King Ranch, unpublished data); thus, the general
survey is robust to potential violations in the constant
detection assumption.
Environmental Data
We acquired spatial dataset that interpolated monthly
precipitation at a 4-km resolution from the PRISM Climate
Group for each month and year of our study (http://prism.
oregonstate.edu, Accessed 18 Jan 2014). We lagged
covariates for precipitation between October and December
by 1 year (precipitation during Oct 2009 was associated
with covey counts in Sep 2010) because covey counts were
collected in September of time t and using covariates that
have not occurred yet (i.e., time t þ 1) to predict covey
3
counts in time t would be misleading. We selected the
precipitation covariates that would be used in our candidate
models according to corrected Akaike’s Information
Criterion (AICc) values from a set of univariate models
for each month.
We used the Texas Ecological Systems Classification
(TESC) dataset (Elliott et al. 2009), which is a land cover
classification map based on ecological system classifications
described by Comer et al. (2003). The TESC was derived
using remotely sensed Landsat Thematic Mapper satellite
data acquired in 2010 and abiotic variables (i.e., soil,
hydrology, elevation, ecoregion) to classify cover types at a
10-m resolution. Although the TESC dataset describe land
cover on our study area in 2010 only, it is unlikely that the
overall structure of vegetation (i.e., grass, shrub, brush)
varied considerably across a 5-year period (D. D. Diamond,
Missouri Resource Assessment Partnership, University of
Missouri, personal communication).
We reclassified the >50 land cover types in the TESC into
4 general types corresponding to the functional cover needs
of bobwhite during various parts of their life history, and
covers they do not use (Hernandez and Guthery 2012).
Specifically, these included 1) non-usable covers we do not
expect quail to occupy, 2) grasslands that provide quail
nesting cover via bunch grasses, 3) grasslands that provide
brood-rearing cover that offers concealment from above via
thick-canopied grasses, and 4) shrubland or brush covers that
offer escape, resting, and thermal covers under different
conditions (Table 1).
We further reclassified grassland cover based on soil type.
Vegetation and soils are intricately related on arid rangelands
because soil type can influence the distribution and
physiognomy of vegetation (Box 1961). We obtained soil
data from the Soil Survey Geographic Database (SSURGO)
available from the United States Department of Agriculture
(USDA) Natural Resource Conservation Service (NRCS;
http://soildatamart.nrcs.usda.gov, Accessed 27 Jan 2013).
We aggregated soil map units based on how they influence
vegetation (Table 1). We classified soil based on attributes in
the database that describe surface texture of soil: fine sand,
fine sandy loam, loamy fine sand, and non-sandy. This
created a polygon boundary of each soil class. We isolated
grassland cover from the reclassified TESC dataset and
overlaid it on the soil data to create 4 distinct classes of
grassland based on soil type (Table 1). We merged these and
the remaining cover types back into a single raster.
Landscape structure may be represented with the patchmosaic model (Forman 1995), which defines land cover on
the surface of the landscape as a mosaic of habitat patches
with specific compositions and configurations. Most land
cover dataset consist of cover or use information stored as
grid cells (i.e., a raster) making the patch-mosaic landscape
model simple to apply because landscape structure can
easily be accounted for in a geographic information system
(GIS; McGarigal 2014). We used FRAGSTATS (Version
4.2; University of Massachusetts, Amherst, MA) to
quantify aspects of composition and configuration. We
assessed composition and configuration at the level of land
cover type (i.e., class level in FRAGSTATS) and the
landscape level (i.e., landscape level in FRAGSTATS).
The landscape level can be interpreted as a measure of
landscape heterogeneity as it represents the aggregate of all
patches on the landscape, and is an overall measure of
landscape structure (McGarigal 2014). We used a
6,400-m2 rectangular moving window analysis. This
window size corresponds to the smallest area needed to
encompass typical non-dispersal movement by bobwhites
in a landscape (Fies et al. 2002, Townsend et al.
2003). The result is a raster (30-m) depicting landscape
structure across the ranch. These landscape structure
dataset provided raster outputs that were the basis for
covariates in our models. We aggregated environmental
data rasters by averaging all pixels that fell within the
boundary of a landscape in Geospatial Modelling
Environment (GME; Version 0.7.2.1, http://www.
spatialecology.com/gme).
Table 1. We reclassified the Texas Ecological Systems Classification land cover dataset to a more general classification scheme based on the functional cover
needs of bobwhite during various parts of their life history. Descriptions of cover are based on Elliott et al. (2009).
Name
Description
Life-history need
None
Loamy fine sand (LFSG)
Fine sandy loam (FSLG)
Covers such as marsh, open water, urban, or sand dunes that are not used by bobwhites
Covers dominated by herbaceous vegetation (both, pasture, and prairie) with <25% woody
cover
Grasslands occurring on loamy fine sand soils
Grasslands occurring on fine sandy loam soils
Fine sand (FSG)
Non-sandy (NSG)
Grasslands occurring on fine sand soils
Grasslands occurring on non-sandy soils such as loamy or clayey soils
Shrubland (SHRUB)
Cover consisting of shrubs (<4 m tall) that account for >25% of woody cover
Brushland/woodland
(BRUSH)
Cover consisting of trees (>4 m tall) that account for >75% of woody cover
Non-usable habitat (NU)
Grasslands
a
b
c
4
Nestinga
Brood rearing,
forageb
Nestinga
Brood rearing,
forageb
Escape, resting,
thermalc
Escape, resting,
thermalc
Hansmire et al. (1988).
Larson et al. (2010).
Hernandez and Guthery (2012).
The Journal of Wildlife Management
9999
Modeling
We used a 2-step approach to reduce the number of
landscape structure covariates that would be considered in
the development of our candidate models. First, we reviewed
Li and Wu (2004) in efforts to identify metrics with intuitive
interpretations. Additionally, we arbitrarily excluded covariates based on how well they could be translated to
management applications. Based on these criteria, we
identified 22 categories of metrics in FRAGSTATS, which
expanded to 139 landscape structure covariates because when
cover-level metrics are applied to each land cover type, the
number of covariates multiplies according to how many cover
types are used in data collection (see Table S1, available
online at www.wildlifejournals.org). Second, we further
reduced the 139 covariates by excluding collinear covariates
(r > |0.6|; retaining the covariate with the most intuitive
application to management) and modeling each covariate in a
univariate manner to identify important relationships.
Specifically, we used AICc to select the 2 best composition
and configuration covariates associated with each of the land
cover types required by bobwhite life-history needs (defined
above; Table 1). Additionally, we selected the single best
composition and configuration metric to define non-usable
covers and overall landscape heterogeneity. We intentionally
selected a single pair of composition–configuration covariates for non-usable covers and landscape heterogeneity.
Non-usable cover comprised a small amount of the overall
study area, and we wanted to avoid undue importance of
non-usable covariates in the model because they are not
variables that managers can augment (but are certainly of
management interest). Few landscape heterogeneity covariates within <2 AICc remained after our covariate reduction
criteria. We developed a candidate set of models to evaluate
how landscape structure of bobwhite land cover types and
precipitation, individually and jointly (i.e., 2 main effects),
affect covey counts (Appendix A). We explored precipitation–landscape structure interactions for covariates in our
best-fitting models to evaluate if interactions improved
model fit in a post hoc manner.
We modeled relationships between bobwhite counts and
covariates using mixed effects negative binomial regression
with landscapes being a random effect (Bolker et al. 2009,
Irwin et al. 2013). Because sampling effort (i ¼ km of
transect sample) varied among landscapes, we used the
natural logarithm of effort as an offset term in all models
(Maunder and Punt 2004). There are multiple ways to
parameterize the negative binomial distribution for regression analyses, and we used the common parameterization
whereby the variance increases quadratically with the mean
(i.e., NB2 parameterization of Hilbe 2011):
VarðY Þ ¼ m þ km2 :
where Y is the observed value for total coveys counted, m is
the mean, and k is the scale parameter (1/k ¼ overdispersion). Moreover, we used a log link to model the
expected value of coveys counted as a function of
hypothesized covariates:
Parent et al.
Mapping Northern Bobwhite
lnðmij Þ ¼ lnðEf f ortij Þ þ b0 þ ai þ
Q
X
bq X qij ;
q¼1
where Effortij is the km of transect sampled at landscape i in
year j, b0 is the intercept, ai is the random intercept for
landscape i, and bq is the effect of covariate Xq on coveys for
landscape i in year j, where Q is the total number of
covariates. The random intercept for each site is assumed to
be independent and identically distributed as Nð0; s2a Þ.
Covariate Xq were centered and standardized by subtracting
the mean and dividing by their standard deviation.
We fit all models using the glmmADMB package in R
(Version 3.0.2; R Foundation for Statistical Computing,
Vienna, Austria). This package uses the computational
power of the random effects module in AD Model Builder
(hereafter ADMB, http://admb-project.org, accessed 18
Nov 2013; Fournier et al. 2012) to integrate out the random
effects from the full likelihood and maximize the marginal
likelihood function of the data (Bolker et al. 2013). We
compared the set of candidate models using AICc and
AICc weights (wi) that we calculated using the bblme
package in R. We evaluated model fit of the top model by
calculating root mean-squared error (hereafter RMSE),
which effectively provides an estimate of average error
between the fitted and observed values for each model on the
scale of the observed data (i.e., measured in number of
coveys). Because we did not expect the residuals to be
normally distributed, we used Anscombe residuals to
construct standard regression diagnostic plots for our models
(e.g., residual-by-fitted value plots; Hilbe could predict our
data by performing leave-one-out cross validation
(LOOCV). Essentially, our top model was applied iteratively
to each n – 1 experimental units and the RMSE between
predicted and observed covey counts was estimated. We
compared error derived from our top model (RMSEmodel)
and derived from LOOCV (RMSECV) to determine if our
model suitably predicted our data.
Mapping Model Predictions
We mapped predictions of covey counts from our top model
to provide managers a means to prioritize and strategically
place management interventions. Because precipitation
varies, we mapped predictions under 3 precipitation
scenarios: drought, average, and wet.
Historical minimums and maximums in precipitation
occurred in 2009 (x ¼ 31.22 cm, SD ¼ 2.80) and 2010
(x ¼ 115.87 cm, SD ¼ 6.24); these years represented drought
and wet precipitation scenarios, respectively, and should
provide good contrast to visualize how covey predictions
change across space with different amounts of precipitation.
We derived the average precipitation scenario using the
average values based on PRISM data collected between 2008
and 2012 (54.96 cm, SD ¼ 22.26). This exceeds the 1971–
2000 average for precipitation (44.7 cm) by 10 cm (http://
www.ncdc.noaa.gov/IPS/coop; Kingsville and Falfurrias
weather stations). The PRISM dataset are collected at
different spatial scales, during different periods of time, and
in the presence of different trends in climate. Because the
5
Table 2. Covariates reflecting land cover composition and configuration of management units (landscapes) that we identified as covariates affecting
bobwhite abundance in southern Texas, 2008–2012, using univariate-negative binomial regressions from a larger suite of candidate covariates. We selected
the top pair of covariates for each vegetation cover according to corrected Akaike’s Information Criterion (AICc). We defined essential bobwhite covers by
combining composition and configuration covariates.
Heterogeneity
Habitat
Cover
Non-usable cover (NUC)
Nesting cover (NC)
Brood/forage cover (BFC)
Escape/resting/thermal cover (ERTC)
Landscape
Landscape heterogeneity (LH)
Compositiona
Configurationb
TE
PLAND_LFSG
TE_FSG
PLAND_FSLG
PLAND_NSG
CA_SHRUB
CA_BRUSH
TA
LSI
ED_LFSG
MPS_FSG
CLUMPY_FSLG
LSI_NSG
PD_SHRUB
IJI_BRUSH
NP
a
TE, total edge (m); PLAND_LFSG, percent loamy fine sand grasslands (%); TE_FSG, total edge of fine sand grassland (m); PLAND_FSLG, percent fine
sandy loam grasslands (%); PLAND_NSG, percent non-sandy grasslands (%); CA_SHRUB, area of shrubland (ha); CA_BRUSH, area of brush (ha); TA,
total area (ha).
b
LSI, landscape shape index; ED_LFSG, edge density of loamy fine sand grasslands (m/ha); MPS_FSG, mean patch size of fine sand grassland (ha);
CLUMPY_FSLG, clumpiness of fine sandy loam grasslands as a measure of fragmentation; LSI_NSG, landscape shape index for non-sandy grasslands;
PD_SHRUB, patch density of shrubland (no. patches/100 ha); IJI_BRUSH, interspersion and juxtaposition of brush (%); NP, number of patches.
difference is small, we suggest our 2008–2012 precipitation
averages provide reasonable surrogates for average precipitation at our study area. We used the original rasters for
landscape structure covariates created during moving
window analyses in FRAGSTATS. We extracted the spatial
random effects from our model and included the effort offset
in predictions for each landscape (each value was held at its 5year average). We extracted coefficient estimates for the fixed
effects from our model and applied these to the covariate
rasters to produce predictions of relative covey counts for
each landscape using the raster calculator within ArcMAP
(ArcGIS 10.1, ESRI, Redlands, CA).
RESULTS
We performed surveys of bobwhite coveys in September
during 2008–2012 with 27,184 km of helicopter surveys.
Average survey effort was 50 km/landscape (n ¼ 122 units;
range ¼ 3–242 km, SD ¼ 34 km/landscape). Counts averaged across divisions and years were greatest on the Norias
(n ¼ 38 landscapes; x ¼ 11.63 coveys/km; SD ¼ 12.34),
followed by Santa Gertrudis (n ¼ 41; x ¼ 11.40; SD
¼ 10.53), Laureles (n ¼ 23; x ¼ 9.00; SD ¼ 9.12), and
Encino (n ¼ 20; x ¼ 8.57; SD ¼ 7.50) divisions.
We identified 3 months that had consistently significant
bobwhite–precipitation relationships: October (previous
year), April (current year), and June (current year). Our
2-step covariate reduction approach yielded 16 landscape
structure covariates (Table 2). Composition covariates
included total area of all patches of a cover type or total
area of all patches in ha, proportion of a cover type with
respect to the overall landscape, and total length of edge in m.
Configuration covariates were represented by indices of
contagion and interspersion that included degree of patch
clumping (range ¼ 1 to 1); interspersion and juxtaposition
(range ¼ 0 to 100%); a landscape shape index (range ¼ 1 to
infinity); estimates of area and edge, which included mean
patch size and edge density measured in ha and m/ha,
respectively; and indices of fragmentation, which included
6
number of patches and patch density, denoted by n patches,
and n patches/ha, respectively.
Modeling
Our best-fitting model consisted of the 3 precipitation
covariates; 4 landscape structure covariates including area of
shrubland, area of brush, patch density of shrubland, and
interspersion and juxtaposition of brush; and 2 landscape
heterogeneity covariates: total area and number of patches.
Only 1 other model in our candidate set was of similar
parsimony but accounted for a negligible proportion of
Akaike weight (Appendix A). We explored precipitation–
landscape structure interactions, but these did not improve
model fit according to AICc. Cross-validated prediction error
Table 3. Parameter estimates for a negative binomial regression between
northern bobwhite covey counts (n ¼ 573) and covariates for precipitation,
landscape structure, and spatial heterogeneity. The model is based on data
collected from 2008 to 2012 on the King Ranchsouthern Texas. Covariates
were centered and standardized by subtracting their mean and dividing by
their standard deviation so they can be interpreted on the same scale.
Because the log of expected counts are modeled in negative binomial
models, we exponentiated our coefficient estimates and converted them to
percent changes.
Covariatea
(Intercept)
Apr
Jun
Oct
CA_SHRUB
CA_BRUSH
PD_SHRUB
IJI_BRUSH
TA
NP
a
Coefficient (%)
Z
P
81.75
0.88
0.34
1.26
0.01
0.02
5.31
0.20
0.01
0.01
9.630
6.700
2.710
6.280
1.260
2.810
2.820
0.770
1.840
0.090
<0.001
<0.001
0.007
<0.001
0.207
0.005
0.005
0.441
0.065
0.925
Apr, April precipitation (mm); Jun, June precipitation (mm); Oct,
October precipitation in previous year (mm); CA_SHRUB, area of
shrubland (ha); CA_BRUSH, area of brush (ha); PD_SHRUB, patch
density of shrubland (no. patches/ha); IJI_BRUSH, interspersion and
juxtaposition of brush (%); TA, total area (ha); NP, number of patches.
The Journal of Wildlife Management
9999
was similar between the top model and LOOCV analyses
(RMSEmodel ¼ 7.765, RMSECV ¼ 7.921). The model fit
well but over-predicted small, outlying counts. Finally,
covey counts indeed were overdispersed (1/k ¼ 4.63,
SE ¼ 0.48) providing support for use of the negative
binomial framework.
Our best-fitting model indicated April precipitation,
area of brush, and patch density of shrubland positively
influenced covey counts, whereas landscape heterogeneity
(total area) and notably, precipitation in June and October
negatively influenced counts (Table 3). Several covariates
in our supported model did not have significant relationships with covey counts (area of shrubland, interspersion
and juxtaposition of brush, and number of patches). To
facilitate interpretation of this model, we created effects
plots for all possible combinations of precipitation–
landscape structure covariates in which significant relationships existed with covey counts. These plots strictly
reflect contribution of 2 main effects when all other values
are held at their mean, and not an interaction. We present
the 3 plots that yielded the largest effect sizes in covey
counts for 1-unit changes in covariates when standardized
for the range of values that each covariate can take on. This
included April precipitation, area of brush, and patch
density of shrubland (Fig. 2). More specifically, precipita-
tion and landscape structure covariates for brush and
shrubland had the greatest positive effect on covey counts
at their maximum recorded values.
Under the best possible simulation conditions for
bobwhites (i.e., wet, high-precipitation scenarios), predicted
relative covey counts averaged 13.50 coveys (Fig. 3). During
drought scenarios, 17 landscapes (management units) on the
ranch were predicted to maintain moderate to high covey
counts >13.50 coveys (Fig. 3). The elevated covey counts in
these 17 units resulted from greater areas of brush and higher
densities of shrubland patches, despite receiving less
precipitation in 2008–2012 (Table 4). Our model also
predicted low-to-moderate covey counts at 71 of 122
landscapes in wet years apparently because they contained
markedly fewer shrubland patches and smaller quantities of
brush (Table 4).
DISCUSSION
Precipitation regulates the height, structure, and by extension,
the function of vegetation with respect to bobwhites (Rice et al.
1993, Bridges et al. 2001, Hernandez et al. 2002). If the timing
of precipitation is not synchronized with important bobwhite
life-history events (nesting, brood rearing), then vegetation may
not be functionally usable (Lusk et al. 2001). Accordingly,
understanding the quantity and timing of precipitation during
Figure 2. Covey count predictions (solid line) and 95% confidence intervals (dashed lines) in southern Texas between 2008 and 2012. Predictions were plotted
in relation to a range of effect sizes for precipitation in April, area of brush (ha), and shrubland patch density (no. shrubland patches/100 ha); all other covariates
were held at their mean.
Parent et al.
Mapping Northern Bobwhite
7
Figure 3. Relative predicted covey counts on the King Ranch in southern Texas between 2008 and 2012. Predictions were based on covariates for
precipitation (Apr and Jun of current year, and Oct of preceding year) and spatial heterogeneity (area of shrubland and brush, patch density of shrubland,
interspersion–juxtaposition of brush, total area and number of patches on the landscape). We made predictions under 3 precipitation scenarios. A
drought scenario was invoked by holding precipitation covariates at their 2009 values (a historically severe drought); similarly, a wet scenario was invoked
by holding precipitation covariates at their 2010 values (a historically wet year). We held covariates at their 2008–2012 average to simulate average
conditions.
periods of drought has important management implications.
Our study took place during a particularly dry period (4 of 5 years
were classified as drought; see Fig. S1, available online at www.
wildlifejournals.org) on arid landscapes in southern Texas and
allowed us to evaluate this effect.
Past research has demonstrated positive relationships
between precipitation and various aspects of bobwhite
population demographics (Bridges et al. 2001, Hernandez
et al. 2005, Lusk et al. 2009). Our results are inconsistent
with previous studies, however, because we detected both
positive and negative relationships with precipitation. For
example, covey counts were most sensitive to precipitation in
April at our study. This is not surprising because even during
wet years in Texas, spring precipitation initiates bobwhite
nesting (Lehmann 1946) and has been associated with
improved nesting success through herbaceous vegetation
Table 4. We modeled relative bobwhite covey counts on the King Ranch in southern Texas from 2008 to 2012 under simulated wet, average, and dry
(drought) conditions. Under the best possible simulation conditions for bobwhites (i.e., wet, high-precipitation scenarios), relative covey counts averaged
13.50 coveys/management unit. Yet, even during simulated drought conditions, predicted covey counts were >13.50 (moderate–high) at 17 management
units, and moderate–high at 51 management units under average precipitation scenarios. Average (SD) April precipitation (mm), patch density of shrubland
(PD_SHRUB; no. patches/100 ha), and area of brush (AREA_BRUSH; ha) at these sites under each precipitation scenarios were numerically greater than
sites where predicted covey counts were low–moderate (<13.50 coveys/management unit).
Precipitation scenario
Below average precipitation
Above average precipitation
8
n
Predicted coveys
17
105
51
71
Moderate–high
Low–moderate
Moderate–high
Low–moderate
Obs. coveys
25.09
7.78
16.43
5.39
(14.12)
(6.72)
(12.04)
(4.64)
Apr. precip
34.74
37.22
35.15
38.28
(37.02)
(37.70)
(37.79)
(37.39)
PD_SHRUB
5.89
4.30
5.37
3.84
(3.54)
(3.56)
(3.64)
(3.41)
AREA_BRUSH
594.56
302.06
491.15
224.84
(1,185.56)
(719.2)
(962.58)
(641.19)
The Journal of Wildlife Management
9999
growth (Parmalee 1955, Lehmann 1984). Additionally,
spring precipitation likely synchronizes bobwhite brood
rearing to coincide with peak invertebrate availability.
Invertebrates are essential to the survival of bobwhite chicks
(Campbell-Kissock et al. 1985, Giuliano et al. 1996), and
their availability in brood-rearing cover used by bobwhites is
contingent on the timing of precipitation (Wisdom 1991).
Accordingly, even small quantities of spring precipitation
during drought can elicit strong, positive population
responses in bobwhites.
We also demonstrated that bobwhite covey counts on our
landscapes were negatively related to summer and fall
precipitation, which runs counter to existing long-term,
broad-scale studies at other arid landscapes (Guthery et al.
2002, Tri et al. 2012). One hypothesis that may explain
negative bobwhite–precipitation relationships in our study
can be explained by the effect late-season precipitation has on
bobwhite nesting phenology. Amidst persistent drought
bobwhites attempted to nest in the presence of considerable
June and October precipitation (see Fig. S2, available online
at www.wildlifejournals.org). Nesting attempts in October
are rare, but do occur in southern Texas (Lehmann 1984)
likely because precipitation appears to be a key determinant
in the timing of breeding (Trewella 2014). Under these
circumstances, the bobwhite hatch occurs out of sync and
many habitat features critical to brood rearing are not
available or are insufficient. This effect on bobwhites
manifested as a negative bobwhite–precipitation relationship
in our models. This phenomenon is well documented in a
variety of avian taxa and systems and is a major influence on
avian population abundances (Williams and Middleton
2008).
Seventeen landscapes in our study area supported high
relative abundances of bobwhites despite persistent droughts.
This suggests another factor buffers bobwhites from the
effects of drought. The existence of positive relationships
between bobwhite population responses and landscape
structure covariates may provide this buffer. We show covey
counts of bobwhites were positively related to landscapes at
our study area containing higher densities of shrubland
patches and greater interspersion of brush cover. This is
consistent with previous work that has shown bobwhites
respond positively to greater composition of early seral
habitats that fulfill certain life-history requirements (Schairer
et al. 1999, White et al. 2005, Howell et al. 2009, Duren et al.
2011, Blank 2012). However, the features of early seral
habitat important to bobwhite usually do not contain brush
cover types. Brush covers are more important for loafing
(Johnson and Guthery 1988), escape, or concealment from
predators (Roseberry and Klimstra 1984), and thermal escape
to reduce exposure to excessive heat (Guthery et al. 2005).
Moreover, because long-term bobwhite population persistence can be accomplished on as little as 11% brush cover
(DeMaso et al. 2014), managers usually devote substantial
resources to reduce brush (Hernandez and Guthery 2012).
Thus, the presence of landscape structure covariates
describing brush cover in our models is curious. We
hypothesize that early seral habitat features provided by
Parent et al.
Mapping Northern Bobwhite
grassland cover types may be unavailable or unusable to
bobwhites during drought. Under these conditions, shrubland and brush cover may be the only vegetation available for
bobwhites to meet the cover needs dictated by their life
history. For example, it is reasonable to assume that shade
created by shrubland and brush cover increases soil moisture
(an important determinant in invertebrate abundance in arid
regions; Wallwork 1972), which could provide forage covers
during drought. A similar population response to soil
moisture gradients in logged and non-logged forest was
reported for capercaillie (Tetrao urogallus) in a forest
ecosystem (Stuen and Spids 1988) and migratory birds on
playa wetlands under various flooding scenarios (Anderson
and Smith 2000).
That predicted counts of coveys are greatest in the
presence of abundant spring precipitation is not surprising;
this is well established in the quail literature. However, that
covey counts in some landscapes are consistently high or low
irrespective of drought, presumably because of landscape
structure, has important management implications. This
indicates landscape structure and precipitation operate
jointly to affect bobwhite population responses. This
suggests some landscapes contain habitat features capable
of buffering bobwhite populations from drought. If
managers understand relationships between bobwhite
population response and the landscape structure associated
with these habitat features, then these drought-tolerant
landscape conditions can be reproduced on other arid
landscapes (or at the very least, landscapes can be managed
to be more drought-tolerant). We offer a simplistic example
from our study to demonstrate the application of these ideas
to other landscapes. At our study, the 17 landscapes with
consistently high counts of bobwhites regardless of drought
contained an average of 422 ha of shrubland and 594 ha of
brush. To simplify, a bobwhite likely views shrubland and
brush patches as woody cover, in which 1,016 ha (11% of
each landscape) exist on average at each of the 17
landscapes, distributed at an average density of 6 patches/
100 ha (Table 4), under intermediate interspersion (approx.
60%). Put differently, if managing a 100-ha landscape we
would strive for 6 patches of woody cover with
intermediate interspersion and an average patch size of
1.83 ha (i.e., a rectangular patch with 135-m sides, or
circular patch with a 76-m radius) that comprised 11% cover
of the landscape. This is roughly analogous to previous
recommendations of strip–motte patterns in the configuration of woody cover (Hernandez and Guthery 2012). More
generally, bobwhite counts during persistent drought were
highly sensitive to even small amounts of brush; thus,
regardless of configuration, simply having some minimum
area of suitable brush cover for bobwhites is of principal
importance.
The conversion of shrubland and brush covers to usable
space often is the primary aim of habitat management for
bobwhites (Hernandez et al. 2007). Unfortunately, this
management task also is costly in terms of economics and
logistics (Hernandez and Guthery 2012). Because management resources are finite, an obvious advantage of translating
9
model results into maps is the capability to visualize spatial
patterns of predictions, and spatially prioritize where
management should occur on the landscape. We suggest
performing the above exercise in conjunction with a map of
the model predictions to weigh the costs associated with
management interventions with the potential population
response it may elicit.
We acknowledge the inferences we used to draw these
conclusions are specific to our landscapes. Although the
extent to which these relationships are representative of
random samples at other landscapes is not clear, we
speculate the landscape structure of brush on managed or
unmanaged landscapes is relatively uniform across the arid
portions of the bobwhite range. This idea needs to be
challenged and represents an avenue for future research.
Another assumption about our model that needs to
be evaluated is the threshold effect of bobwhite–landscape
structure relationships. Our top model indicated positive
relationships with the landscape structure of shrubland and
brush covers, suggesting the composition and configuration of these covers can conceivably be increased
indefinitely to the benefit of quail. Clearly, though
some intermediate or threshold level exists. Finally, it
should also be noted that landscape structure covariates
depicting bobwhite nesting and brood-rearing covers were
curiously absent from our top models. Factors augmenting
vital rates of quail (DeMaso et al. 2011, Parent et al. 2012)
and their nesting habitat availability (Rader et al. 2011)
can potentially limit bobwhite populations, but this was
not detected by our models. Several explanations exist for
this observation. It is possible habitat features of nesting
and brood-rearing cover types do not translate well to land
cover dataset. Land cover dataset coarsely describe
vegetation and may not effectively model bobwhite cover
that is mediated by fine-scale processes (Gallant 2009).
Thus, our classification scheme, which was developed
according to field studies describing micro-site characteristics of nesting and brood-rearing cover, may not have
adequately represented the broad-scale features of nesting
and brood-rearing habitat. Alternatively, in the presence
of frequent drought, these habitat features may not have
been available uniformly across time and space as indicated
by our land cover dataset.
MANAGEMENT IMPLICATIONS
This study provided support for the concept that
bobwhite distribution in arid landscapes responds to
precipitation and landscape structure. If precipitation is
absent during key periods for bobwhite, then essential
structural cover types such as nesting or brood-rearing
cover may be unavailable or unusable. When configuration and composition at our study area are taken
collectively, and in the presence of drought, relative
counts of bobwhites are expected to be greatest on
landscapes containing large areas of well-interspersed
shrubland and large areas of brush distributed as many
patches. Although shrubland and brush cover types may
buffer bobwhite populations from drought, not all
10
landscapes in the arid portion of the bobwhite’s range
contain these favorable habitats. This suggests management applications that facilitate the conversion of these
woody covers to suitable compositions and configurations
can mitigate future drought-induced ecological bottlenecks that reduce bobwhite populations. These buffer
sites could easily be worked in to existing habitat
management approaches because most strategies involve
the creation of strips or strip–mottes. Performing these
tasks in conjunction with a map of model predictions has
tangible benefits to managers because it can prioritize
where management needs to occur.
ACKNOWLEDGMENTS
B. S. Stevens and A. N. Tri provided a critique of this article.
We thank the Associate Editor and 2 anonymous reviewers.
LAB was supported by the C. C. Winn Endowed Chair,
FCB was supported by the Leroy G. Denman Jr. Endowed
Directorship, and FH was supported by the Alfred C.
Glassell Jr. Endowed Professorship in the R. M. Kleberg, Jr.
Center for Quail Research. This manuscript is Caesar
Kleberg Wildlife Research Institute Publication Number 15118. We thank the Caesar Kleberg Wildlife Research
Institute and the Richard M. Kleberg Jr. Center for Quail
Research for providing financial support. External financial
support also was provided by King Ranch, U.S. Department
of Education’s Title V/PPOHA grant, the South Texas
Quail Coalition, Mr. Rene Barrientos, and the Houston
Safari Club.
LITERATURE CITED
Anderson, J. T., and L. M. Smith. 2000. Invertebrate response to moistsoil management of playa wetlands. Ecological Applications 10:550–558.
Blank, P. J. 2012. Northern bobwhite response to conservation reserve
program habitat and landscape attributes. Journal of Wildlife Management 77:68–74.
Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen,
M. H. H. Stevens, and J.-S. S. White. 2009. Generalized linear mixed
models: a practical guide for ecology and evolution. Trends in Ecology &
Evolution 24:127–135.
Bolker, B. M., B. Gardner, M. Maunder, C. W. Berg, M. Brooks, L.
Comita, E. Crone, S. Cubaynes, T. Davies, P. de Valpine, J. Ford, O.
Gimenez, M. Kery, E. J. Kim, C. Lennert-Cody, A. Magnusson, S.
Martell, J. Nash, A. Nielsen, J. Regetz, H. Skaug, and E. Zipkin.
2013. Strategies for fitting nonlinear ecological models in R, AD
Model Builder, and BUGS. Methods in Ecology and Evolution
4:501–512.
Box, T. W. 1961. Relationships between plants and soils of four range plant
communities in south Texas. Ecology 42:794–810.
Brady, S. J., C. H. Flather, and K. E. Church. 1998. Range-wide declines of
northern bobwhite (Colinus virginianus): land use patterns and population
trends. Gibier Faune Sauvage 15:413–431.
Bridges, A. S., M. J. Peterson, N. J. Silvy, F. E. Smeins, and X. B. Wu. 2001.
Differential influence of weather on regional quail abundance in Texas.
Journal of Wildlife Management 65:10–18.
Campbell-Kissock, L., L. H. Blankenship, and J. W. Stewart. 1985. Plant
and animal foods of bobwhite and scaled quail in southwest Texas.
Southwestern Naturalist 30:543–553.
Comer, P., R. Faber-Langendoen, R. Evans, S. Gawler, C. Josse, G. Kittel,
S. Menard, M. Myne, M. Reid, K. Schulz, K. Snow, and J. Teague. 2003.
Ecological systems of the United States: a working classification of U.S.
terrestrial systems. NatureServe, Arlington, Virginia, USA.
DeMaso, S. J., W. E. Grant, F. Hernandez, L. A. Brennan, N. J. Silvy, X. B.
Wu, and F. C. Bryant. 2011. A population model to simulate northern
The Journal of Wildlife Management
9999
bobwhite population dynamics in southern Texas. Journal of Wildlife
Management 75:319–332.
DeMaso, S. J., F. Hernandez, L. A. Brennan, N. J. Silvy, W. E. Grant, X.
Ben Wu, and F. C. Bryant. 2014. Short- and long-term influence of brush
canopy cover on northern bobwhite demography in southern Texas.
Rangeland Ecology & Management 67:99–106.
Duren, K. R., J. J. Buler, W. Jones, and C. K. Williams. 2011. An improved
multi-scale approach to modeling habitat occupancy of northern
bobwhite. Journal of Wildlife Management 75:1700–1709.
Elliott, L. F., D. German, D. D. Diamon, A. Treuer-Keuhn, K. Ludeke,
and J. Scott. 2009. Texas ecological systems project: phase 1 interpretive
booklet. Texas Parks and Wildlife Department, Austin, USA.
Fies, M. L., K. M. Puckett, and B. Larson-Brogdon. 2002. Breeding
season movements and dispersal of northern bobwhites in fragmented
habitats of Virgina. Proceedings of the National Quail Symposium
5:173–179.
Forman, R. T. T. 1995. Land mosaics: the ecology of landscapes and regions.
Cambridge University Press, Cambridge, United Kingdom.
Fournier, D. A., H. J. Skaug, J. Ancheta, J. Ianelli, A. Magnusson, M. N.
Maunder, A. Nielsen, and J. Sibert. 2012. AD Model Builder: using
automatic differentiation for statistical inference of highly parameterized
complex nonlinear models. Optimization Methods and Software
27:233–249.
Fulbright, T. E., and F. C. Bryant. 2002. The last great habitat. Caesar
Kleberg Wildlife Research Institute, Special Publication No. 1, Kingsville,
Texas, USA.
Gallant, A. L. 2009. What you should know about land-cover data. Journal
of Wildlife Management 73:796–805.
Giuliano, W. M., R. S. Lutz, and R. Patino. 1996. Reproductive responses
of adult female northern bobwhite and scaled quail to nutritional stress.
Journal of Wildlife Management 60:302–309.
Gould, F. W. 1975. Texas plants: a checklist and ecological summary. Texas
Agricultural Experiment Station, College Station, USA.
Guthery, F. S. 1997. A philosophy of habitat management for northern
bobwhites. Journal of Wildlife Management 61:291–301.
Guthery, F. S. 2002. The technology of bobwhite management: the theory
behind the practice. Iowa State University Press, Ames, USA.
Guthery, F. S., J. J. Lusk, D. R. Synatzske, J. Gallagher, S. J. DeMaso, R. R.
George, and M. J. Peterson. 2002. Weather and age ratios of northern
bobwhites in southern Texas. Proceedings of the National Quail
Symposium 5:99–105.
Guthery, F. S., A. R. Rybak, S. D. Fuhlendorf, T. L. Hiller, S. G. Smith,
W. H. Puckett Jr., and R. A. Baker. 2005. Aspects of the thermal ecology
of bobwhites in north Texas. Wildlife Monographs 159:1–36.
Haines, A. M., F. Hernandez, S. E. Henke, and R. L. Bingham. 2004.
Effects of road baiting on home range and survival of northern bobwhites
in southern Texas. Wildlife Society Bulletin 32:401–411.
Hansmire, J. A., D. L. Drawe, D. B. Wester, and M. B. Carlton. 1988.
Effect of winter burns on forbs and grasses of the texas coastal prairie.
Southwestern Naturalist 33:333–338.
Hernandez, F., L. A. Brennan, S. J. DeMaso, J. P. Sands, and D. B. Wester.
2012. On reversing the northern bobwhite population decline: 20 years
later. Wildlife Society Bulletin 37:177–188.
Hernandez, F., and F. S. Guthery. 2012. Beef, brush, and bobwhites: quail
management in cattle country. Texas A&M University Press, College
Station, USA.
Hernandez, F., F. S. Guthery, and W. P. Kuvlesky Jr. 2002. The legacy of
bobwhite research in south Texas. Journal of Wildlife Management
66:1–18.
Hernandez, F., F. Hernandez, J. A. Arredondo, F. C. Bryant, L. A. Brennan,
and R. L. Bingham. 2005. Influence of precipitation on demographics of
northern bobwhites in southern Texas. Wildlife Society Bulletin
33:1071–1079.
Hernandez, F., R. M. Perez, and F. S. Guthery. 2007. Bobwhites on the
south Texas plains. Pages 273–298 in L. A. Brennan, editor. Texas quails:
ecology and management. Texas A&M University Press, College Station,
USA.
Hilbe, J. M. 2011. Negative binomial regression. Second edition. Cambridge
University Press, Cambridge, United Kingdom.
Howell, J. E., C. T. Moore, M. J. Conroy, R. G. Hamrick, R. J.
Cooper, R. E. Thackston, and J. P. Carroll. 2009. Conservation
of northern bobwhite on private lands in Georgia, USA under uncertainty
Parent et al.
Mapping Northern Bobwhite
about landscape-level habitat effects. Landscape Ecology 24:405–
418.
Irwin, B. J., T. Wagner, J. R. Bence, M. V. Kepler, W. Liu, and D. B. Hayes.
2013. Estimating spatial and temporal components of variation for
fisheries count data using negative binomial mixed models. Transactions of
the American Fisheries Society 142:171–183.
Johnson, D. B., and F. S. Guthery. 1988. Loafing coverts used by northern
bobwhites in subtropical environments. Journal of Wildlife Management
52:464–469.
Larson, J. A., T. E. Fulbright, L. A. Brennan, F. Hernandez, and F. C.
Bryant. 2010. Texas bobwhites: a guide to their foods and habitat
management. University of Texas Press, Austin, USA.
Lehmann, V. W. 1946. Bobwhite quail reproduction in southwestern Texas.
Journal of Wildlife Management 10:111–123.
Lehmann, V. W. 1984. The bobwhite in the Rio Grande Plain of Texas.
Texas A&M University Press, College Station, USA.
Li, H., and J. F. Reynolds. 1994. A simulation experiment to quantify spatial
heterogeneity in categorical maps. Ecology 75:2446–2455.
Li, H., and J. Wu. 2004. Use and misuse of landscape indices. Landscape
Ecology 19:389–399.
Lohr, M., B. M. Collins, C. K. Williams, and P. M. Castelli. 2011. Life on
the edge: northern bobwhite ecology at the northern periphery of their
range. Journal of Wildlife Management 75:52–60.
Lusk, J. J., F. S. Guthery, and S. J. DeMaso. 2001. Northern bobwhite
(Colinus virginianus) abundance in relation to yearly weather and longterm climate patterns. Ecological Modeling 146:3–15.
Lusk, J. J., F. S. Guthery, R. R. George, M. J. Peterson, and S. J. DeMaso.
2002. Relative abundance of bobwhites in relation to weather and land use.
Journal of Wildlife Management 66:1040–1051.
Lusk, J. J., F. S. Guthery, M. J. Peterson, and S. J. DeMaso. 2009. Longterm climate trends and northern bobwhite populations in south Texas.
Gamebird 2006: Quail VI and Perdix XII 1:328–337.
Maunder, M. N., and A. E. Punt. 2004. Standardizing catch and effort data:
a review of recent approaches. Fisheries Research 70:141–159.
McGarigal, K. 2014. Fragstats help. http://www.umass.edu/landeco/
research/fragstats/documents/fragstats.help.4.2.pdf. Accessed 15 Mar
2015.
McLendon, T. 1991. Preliminary description of the vegetation of south
Texas exclusive of coastal saline zones. Texas Journal of Science 43:
13–32.
Norwine, J., R. Harriss, J. Yu, C. Tebaldi, and R. Bingham. 2007. The
problematic cliamte of south Texas 1900–2100. Pages 15–55 in J.
Norwine, and J. Kuruvilla, editors. The changing climate of south Texas.
CREST-RESSACA, Texas A&M University—Kingsville, Kingsville,
USA.
Parent, C. J., F. Hernadez, D. B. Wester, and F. C. Bryant. 2012. Temporal
and spatial trends of northern bobwhite survival and nest success.
Proceedings of the National Quail Symposium 7:173–183.
Parmalee, P. W. 1955. Some factors affecting nesting success of the
bob-white quail in east-central Texas. American Midland Naturalist
53:45–55.
Peterson, M. J. 2001. Northern bobwhite and scaled quail abundance and
hunting regulation: a Texas example. Journal of Wildlife Management
65:828–837.
Rader, M. J., L. A. Brennan, K. A. Brazil, F. Hernadez, and N. J. Silvy. 2011.
Simulating northern bobwhite population responses to nest predation,
nesting habitat, and weather in south Texas. Journal of Wildlife
Management 75:61–70.
Rice, S. M., F. S. Guthery, G. S. Spears, S. J. DeMaso, and B. H. Koerth.
1993. A precipitation-habitat model for northern bobwhites on semiarid
rangeland. Journal of Wildlife Management 57:92–102.
Roseberry, J. L., and W. D. Klimstra. 1984. Population ecology of the
bobwhite. Southern Illinois University Press, Carbondale, USA.
Rosene, W. 1969. The bobwhite quail—its life and management. Rutgers
University Press, New Brunswick, New Jersey, USA.
Rusk, J. P., F. Hernandez, J. A. Arredondo, F. Hernandez, F. C.
Bryant, D. G. Hewitt, E. J. Redeker, L. A. Brennan, and R. L. Bingham.
2007. An evaluation of survey methods for estimating northern bobwhite
abundance in southern Texas. Journal of Wildlife Management
71:1336–1343.
Sauer, J. R., J. E. Hines, J. E. Fallon, K. L. Pardieck, D. J. Ziolkowski Jr., and
W. A. Link. 2011. The North American Breeding Bird Survey, results and
11
analysis 1966–2009. Version 3.23.2011. USGS Patuxent Wildlife
Research Center, Laurel, Maryland, USA.
Schairer, G. L., R. H. Wynne, M. L. Fies, and S. D. Klopfer. 1999.
Predicting landscape quality for northern bobwhite from classified Landsat
imagery. Proceedings of the Annual Conference of the Southeast
Association of Fish and Wildlife Agencies 53:243–256.
Schnupp, M. J., F. Hernandez, E. J. Redeker, F. C. Bryant, J. P. Rusk, S. J.
DeMaso, J. P. Sands, T. W. Teinert, L. A. Brennan, D. Rollins, and R. M.
Perez. 2012. An electronic system to collect distance-sampling data during
helicopter surveys of northern bobwhite. Wildlife Society Bulletin
37:236–245.
Stoddard, H. L. 1931. The bobwhite quail: its habits, preservation and
increase. Scribner’s Sons, New York, New York, USA.
Stuen, O. H., and T. K. Spids. 1988. Invertebrate abundance in different
forest habitats as animal food available to capercaillie Tetrao urogallus
chicks. Scandinavian Journal of Forest Research 3:527–532.
Townsend, D. E. II, D. M. Leslie Jr., R. L. Lochmiller, S. J. DeMaso, S. A.
Cox, and A. D. Peoples. 2003. Fitness costs and benefits associated with
dispersal in northern bobwhites (Colinus virginianus). American Midland
Naturalist 150:73–82.
Trewella, I. C. 2014. Timing of northern bobwhite breeding in a semiarid
environment. Thesis, Texas A&M University-Kingsville, Kingsville,
USA.
Tri, A. N., J. P. Sands, M. C. Buelow, D. Williford, E. M. Wehland, J. A.
Larson, K. A. Brazil, J. B. Hardin, F. Hernandez, and L. A. Brennan.
2013. Impacts of weather on northern bobwhite sex ratios, body mass, and
annual production in South Texas. Journal of Wildlife Management
77:579–586.
Veech, J. A. 2006. Increasing and declining populations of northern
bobwhites inhabit different types of landscapes. Journal of Wildlife
Management 70:922–930.
Wallwork, J. A. 1972. Distribution patterns and population dynamics of the
micro-arthropods of a desert soil in southern California. Journal of Animal
Ecology 41:291–310.
White, C. G., S. H. Schweitzer, T. M. Clinton, I. B. P. Iii, and L. A. LewisWeis. 2005. Evaluation of the landscape surrounding northern bobwhite
nest sites: a multiscale analysis. Journal of Wildlife Management
69:1528–1537.
Wiens, J. A. 1977. On competition and variable environments: populations
May experience “ecological crunches” in variable climates, nullifying
the assumptions of competition theory and limiting the usefulness of
short-term studies of population patterns. American Scientist 65:590–597.
Williams, S. E., and J. Middleton. 2008. Climatic seasonality, resource
bottlenecks, and abundance of rainforest birds: implications for global
climate change. Diversity and Distributions 14:69–77.
Wisdom, C. S. 1991. Patterns of heterogeneity in desert herbivorous insect
communities. Pages 151–179 in G. A. Polis, editor. The ecology of desert
communities. University of Arizona Press, Tucson, USA.
Associate Editor: James Sheppard.
12
APPENDIX A. CANDIDATE MODELS
A candidate set of models to test hypotheses depicting how landscape
structure affects northern bobwhite abundance during extreme variability in
precipitation. Each candidate model consists of combinations of
precipitation and pairs of configuration and composition covariates
defining essential bobwhite cover types. The models were built to reflect
hypotheses about how precipitation and land cover influence bobwhite
abundance. We fit models using negative binomial regression with
bobwhite counts representing the response variable. We collected data at
122 landscapes on the King Ranch in southern Texas from 2008 to 2012.
We determined model fit by corrected Akaike’s Information Criterion
(AICc) and ranked models according to DAICc, and weight (wi) with K
representing total number of model parameters. Months indicate a
covariate for precipitation during the month shown, NUC is none-usable
cover, NC is nesting cover, BFC is brood or forage cover, ERTC is escape,
resting, or thermal cover, LH is landscape heterogeneity. We used various
combinations of configuration and composition covariates for specific
vegetation types to define covers.
Candidate models
Oct þ Apr þ Jun
Oct þ Apr þ Jun
Oct þ Apr þ Jun
Oct þ Apr þ Jun
Oct þ Apr þ Jun
Oct þ Apr þ Jun
Oct þ Apr þ Jun
Oct þ Apr þ Jun
All rain
Oct þ Apr þ Jun
NC þ LH
ERTC þ LH
LH
NUC þ LH
BFC þ LH
ERTC
NUC
NC
Oct
Jun
BFC
Apr
þ
þ
þ
þ
þ
þ
þ
þ
ERTC þ LH
NC þ LH
ERTC
LH
NUC þ LH
BFC þ LH
NC
NUC
þ BFC
DAICc
K
wi
0
3
7.1
10.6
11.3
12.4
17.8
21.7
28
29.2
42.7
46.8
52.4
53.3
55.1
56.4
65.4
66
66.7
73.5
77.5
77.8
12
14
10
8
10
10
12
8
6
8
11
9
5
7
7
7
5
9
4
4
5
4
0.8
0.18
0.01
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
SUPPORTING INFORMATION
Additional supporting information may be found in the
online version of this article at the publisher’s website.
The Journal of Wildlife Management
9999