Comparative Analysis of Mourning Dove Population Change in North America

Journal of Wildlife Management 74(5):1059–1069; 2010; DOI: 10.2193/2008-459

Management and Conservation Article

Comparative Analysis of Mourning
Dove Population Change in
North America
JOHN R. SAUER,1 United States Geological Survey, Patuxent Wildlife Research Center, 12100 Beech Forest Road, Laurel, MD 20708, USA
WILLIAM A. LINK, United States Geological Survey, Patuxent Wildlife Research Center, 12100 Beech Forest Road, Laurel, MD 20708, USA
WILLIAM L. KENDALL, United States Geological Survey, Patuxent Wildlife Research Center, 12100 Beech Forest Road, Laurel, MD 20708, USA
DAVID D. DOLTON, United States Fish and Wildlife Service, Division of Migratory Bird Management, P.O. Box 25486 DFC, Denver, CO 80225,
USA

ABSTRACT Mourning doves (Zenaida macroura) are surveyed in North America with a Call-Count Survey (CCS) and the North American
Breeding Bird Survey (BBS). Analyses in recent years have identified inconsistencies in results between surveys, and a need exists to analyze the
surveys using modern methods and examine possible causes of differences in survey results. Call-Count Survey observers collect separate
information on number of doves heard and number of doves seen during counting, whereas BBS observers record one index containing all doves
observed. We used hierarchical log-linear models to estimate trend and annual indices of abundance for 1966–2007 from BBS data, CCS-heard
data, and CCS-seen data. Trend estimates from analyses provided inconsistent results for several states and for eastern and central dovemanagement units. We examined differential effects of change in land use and noise-related disturbance on the CCS indices. Changes in noiserelated disturbance along CCS routes had a larger influence on the heard index than on the seen index, but association analyses among states of
changes in temperature and of amounts of developed land suggest that CCS indices are differentially influenced by changes in these environmental
features. Our hierarchical model should be used to estimate population change from dove surveys, because it provides an efficient framework for
estimating population trends from dove indices while controlling for environmental features that differentially influence the indices.
KEY WORDS Call-Count Survey, hierarchical model, North American Breeding Bird Survey, route regression, trend analysis,
Zenaida macroura.

1

E-mail: jrsauer@usgs.gov

Sauer et al. N Mourning Doves in North America

estimated as part of the survey, the indices vary both due
to real population changes and due to changes in the
environment that are not population-related such as
observer ability, phenology of breeding, counting conditions
such as weather and amount of noise-related disturbance, or
land use around the site (e.g., Bibby et al. 2000). It is likely
that the processes of encountering birds by hearing and
seeing are differentially affected by 1 of these environmental features.
Any analysis of index data must control for factors
influencing detectability, and recent analyses of BBS and
CCS dove data control for baseline observer differences that
are known to influence counts (Link and Sauer 1994).
Different estimates of trends in continental dove populations based on the alternative indices indicate presence of
unmodeled features that are differentially influencing the
indices. Identifying these lurking covariates that are
influencing the indices and controlling for their effects in
the analysis must be a high priority for mourning dove
surveys. Unfortunately, the route-regression analysis presently used in dove analyses is limited in its ability to model
covariates that might influence detectability (Dolton et al.
2007). New statistical methods recently implemented for the
BBS and the Christmas Bird Count provide the flexibility
needed for these covariate analyses (Link and Sauer 2002,
Link et al. 2006).
We implemented a new analysis of CCS data, a Bayesian
analysis of log-linear hierarchical models that permits
accommodation of covariates of population change (Link
and Sauer 2002, Sauer et al. 2008b). We compared results of
these models to results from the presently used routeL

Mourning doves (Zenaida macroura) are widely distributed
and common throughout the continental United States,
with an estimated autumn population of 350 million birds
(Otis et al. 2008). A popular game bird, approximately 20
million doves were harvested in the 2006 and 2007 hunting
seasons (Richkus et al. 2008). Management of the species is
based on roadside, point-count–based surveys, primarily a
Call-Count Survey (CCS), supplemented with data from
the North American Breeding Bird Survey (BBS; United
States Department of the Interior 2005, Dolton et al. 2007,
Sauer et al. 2008a). In the CCS, doves heard cooing are
recorded separately from doves seen; in the BBS, total
numbers of doves perceived by sight or sound are recorded.
Analysis results from both the heard and seen indices from
the CCS and from BBS data have routinely been provided
in annual administrative reports that summarize mourning
dove population status (e.g., Dolton et al. 2007).
Sauer et al. (1994) conducted a comparative analysis of
dove population trends from the CCS-heard index and BBS
dove data. Different patterns of population change as
estimated by the 2 independent surveys were identified in
that analysis in several states, although overall changes were
generally similar. In recent years, differences in trends
estimated from the BBS and the CCS have become
increasingly evident, and trends estimated from CCS-heard
and -seen indices have also begun to show inconsistencies
(Dolton et al. 2007).
These differences between the indices highlight a key
limitation of index data. Because detectability is not

1059

regression analysis approach for CCS data and to BBS
results based on hierarchical models. To identify covariates
that may improve our estimation of dove population change,
we evaluated several hypotheses regarding environmental
features that might be acting to differentially influence the
CCS dove indices. We evaluated whether changes in noise
disturbance along routes differentially affects heard and seen
indices in the CCS; we associated trends in temperature by
state with trends in indices to determine whether phenological changes differentially affected the indices, and we
associated trends in land uses by state with trends in indices
to evaluate whether land-use–associated visibility changes
differentially affected the indices.

METHODS
The CCS is optimized for counting mourning doves. Timed
to coincide with the most stable period of cooing activity,
the CCS collects 2 indices of abundance for the species
(Dolton 1993). The CCS is a roadside survey, initiated
across the continental United States in 1966. Each of the
1,255 survey routes contains 20 listening stations (stops)
located 1.6 km (1 mile) apart, at which all doves heard and
seen are recorded during 3-minute point counts conducted
starting 30 minutes before sunrise on a day between 20 May
and 5 June. At each stop, data on doves heard and seen are
recorded separately; doves seen are also recorded while
traveling between stops. Disturbance, defined as presence of
traffic or other noise likely to interfere with counting, is also
recorded at each stop as a categorical variable with levels of
no, low, moderate, or high disturbance. The example
provided in the survey instructions for low disturbance is
distant tractor noise; for moderate and high disturbance
intermittent and continuous traffic are used as examples (D.
D. Dolton, United States Fish and Wildlife Service,
unpublished report).
Although information collected along a route could be
summarized in several ways, 2 indices of abundance have
historically been used to characterize dove populations: 1)
the CCS-heard index is the sum of the counts of doves
heard over the 20 stops; and 2) the CCS-seen is the sum of
the counts of doves seen at and between stops. See Dolton
(1993) for a detailed summary of survey procedures and
design issues.
The BBS is also a roadside survey, but is generally
conducted in June and has 50 3-minute point-count stops
separated by 0.8 km (0.5 miles). A single index to
abundance, the total number of birds perceived by sight or
sound, is collected for each species at each stop, and species
totals over the 50 stops are used as the route summary.
Counts of numbers of vehicles passing during the surveys
were collected starting in 1998, but no noise disturbance
measure was collected for most of the survey period. The
BBS includes information from .4,000 survey routes in the
continental United States and southern Canada. The survey
was initiated in the eastern United States in 1966 and routes
in the western United States and in Canada were first
surveyed in 1968.
1060

Analysis of CCS and BBS Indices
Between 1995 and 2007, the route-regression analysis
method (Geissler and Sauer 1990, Link and Sauer 1994)
was used for analysis of mourning dove population change for
management reports and other summaries (e.g., Dolton et al.
2007). In the route-regression analysis method, intervalspecific population change (trend) is estimated on each survey
route as the slope of a Poisson regression with log links, in
which observer data are included as covariates to allow each
observer to have a separate intercept (baseline counting
ability). Regional trends are estimated by a weighted average
of the route trends, with weights of mean route abundance, a
factor representing consistency of coverage, and an area
weight. Variances are estimated via bootstrapping. Annual
indices are estimated by calculating the average (on the logscale) difference between yearly counts and predicted counts
based on the regional estimated trends.
Link and Sauer (2002) and Sauer et al. (2008b) described a
log-linear hierarchical model to estimate population change
from count surveys. In hierarchical models, some parameters
(e.g., stratum-specific yr effects) are governed by additional
underlying hyperparameters (e.g., stratum means depend on
a distribution based on a national mean). This structure is
well-suited for national surveys, because the surveys are
naturally hierarchical and we are interested in estimating
attributes at many scales: survey routes that occur within
strata and strata that occur within larger regions. Covariates
can be easily added and evaluated at any scale in the
hierarchical model, an important attribute for use of the
model as a framework for evaluating environmental effects
on counts.
We used the overdispersed Poisson regression model
described in Sauer et al. (2008b). The means li,j,t of counts
Yi,j,t (i indices stratum, j for unique combinations of route
and observer, and t for yr) are a log-linear combination of
several explanatory variables:
log(li,j,t )~Si zbi (t{ t
)zvj zci,t zgI (j,t)zei,j,t : ð1Þ
This model has the standard structure of a linear model, but
model parameters are random rather than fixed effects and
we must specify their distributions. Variables S and b are
stratum-specific intercepts and slopes, v are observer–route
combination effects, c is a year effect for year and stratum, g
is a start-up effect [I(j,t) 5 1 for the first yr of survey for an
observer, 0 otherwise], and e represents overdispersion. We
set the baseline year t* to 1984.
We estimated annual indices of abundance and an
interval-specific estimate of population change (trend) as
functions of components of the log-linear model. Stratumspecific annual indices of abundance (ni,t) index the number
of birds per route in stratum i at year t (Link and Sauer
2002) and we estimated them from year effects, stratum, and
trend effects:


ni,t ~exp Si zbi (t{ t
)zci,t z0:5s2v z0:5s2e :
We defined stratum level population indices Ni,t 5 Aini,t,
where Ai is area of the stratum, and composite indices for
The Journal of Wildlife Management N 74(5)

groups of strata as sums of Ni,t divided by total areas in the
collection. Indices Ni,t are not unbiased estimates of
population totals because ni,t are not area-specific population estimates. We defined trend as an interval-specific
geometric mean rate of change of indices (c.f., Link and
Sauer 2002), presented as a yearly percentage change. From
year ta to year tb, for stratum i, trend is 100(bi 2 1)%, where
1
 t {t
ni,tb b a
bi ~
:
ni,ta

Modeling Disturbance Effects
We incorporated an effect of noise-related disturbance as an
additional component of the model for CCS data. We
summarized route-level disturbance as the sum of stops with
recorded moderate and high disturbance levels, because no
and low disturbance levels reflect either background noise or
noise produced at a distance, whereas the higher levels
reflect traffic and other noises occurring close to the
observers. Disturbance data were only available from 1968
to 2007 and were occasionally missing. To model effects of
disturbance on counts without imposing a predefined shape
of the relationship, we used the flexible 2-parameter family
of models developed for effort adjustments in Christmas
Bird Counts (Link et al. 2006), B (ji,j,tp 2 1) / p, which is
added to the log-linear model. In this model, effort ji,j,t is
scaled to an overall mean effort. The parameter p defines the
shape of the relationship between counts and effort, and B is
the coefficient of the effort adjustment.
To use this model for disturbance, we transformed
disturbance data as d 5 21 2 (sum of moderate and high
disturbance levels), allowing a realistic decline in counts
associated with larger levels of disturbance; d ranges from 1
to 21. We also scaled disturbances to their overall mean and
the model became B (di,j,tp 2 1) / p. If disturbance associated
with count Yi,j,t is equal to overall mean disturbance, then
di,j,t 5 1 and the effort effect is zero.
We conducted an analysis of change in disturbance, using
the log-linear model described above for CCS analysis but
substituting our disturbance measure (sum of moderate and
high levels of disturbance) for the count Yi,j,t. We calculated
annual indices and trend in disturbance as described above.

given our prior knowledge of their distributions (e.g., they
are normally distributed) and the likelihood of data
conditional on the parameters. The posterior distribution
is the basis of all Bayesian inference: its mean (or sometimes
its median or mode) is used as a point estimator; its
percentiles are used to create interval estimates, called
credible intervals. We present 95% credible intervals (CrI)
extending from the 2.5th to the 97.5th percentile.
Historically, direct calculation of posterior distributions
from prior distributions and likelihoods by integration has
been difficult; modeling such as that we described would not
have been feasible using the traditional integration-based
approaches. However, simulation-based numerical procedures such as Markov-chain Monte Carlo (MCMC) allow
us to approximate posterior distributions of the parameters
(Gilks et al. 1996). The MCMC produces samples of
posterior distributions through iterative procedures. One of
the most efficient methods of MCMC is Gibbs sampling, in
which individual parameters are sampled from their full
conditional distributions. The full conditional distribution
of a given parameter is its posterior distribution given fixed
values of the other parameters. Cyclical sampling of full
conditional distributions produces samples from the joint
posterior distributions of the model parameters. We can use
the mean and percentiles of sampled values to compute
point estimates and credible intervals for inference.
Furthermore, we can compute functions of parameters (such
as annual indices) based on the MCMC output, easily
obtaining samples of their posterior distributions. Program
WinBUGS (Lunn et al. 2000) implements Gibbs sampling
and includes diagnostic tools for monitoring the process,
and tools for summarizing the results.
We used WinBUGS to fit the log-linear model for BBS
and CCS data. We assigned Si and bi diffuse normal prior
distributions; other effects were modeled as mean zero normal
random variables, with unknown variances. These variances
were assigned flat inverse gamma prior distributions. The v
were identically distributed, with common variance s2v , e
were identically distributed with variance s2e , and variance of
c varied among strata (s2c,i ). We used standard noninformative priors in the disturbance analysis. For parameter p we
used a uniform prior on the interval [24, 4] and gave B a
diffuse normal distribution (Link et al. 2006).
Breeding Bird Survey and CCS mourning dove data sets
are large, and to implement the model with limited
computer resources we thinned the results, storing each
sixth iteration result for summary. We conducted 60,000
iterations of each chain as a burn-in, and then based the
analysis on an additional 60,000 iterations. We evaluated
convergence by inspection of graphs of iterated results and
autocorrelation functions (e.g., Sauer et al. 2008b).

Fitting Hierarchical Models to CCS and BBS
We used Bayesian methods to fit the hierarchical models to
all data sets. In Bayesian methods, we assume all quantities
to be random variables and the goal of inference is to make
probability statements about these unknown quantities by
estimating their distribution (the posterior distribution),

Geographic Regions Used in Analysis
We conducted analyses of mourning dove population
change for 1966–2007 from CCS and BBS data. The
CCS provided information from the 48 conterminous
United States, and the BBS covered that area and southern
Canada. In this analysis, we analyzed CCS and BBS using

¯ 2 1)%,
We calculated regional trends ¯b analogously as 100(b
using the composite indices Nt 5 Si Ni,t, to calculate
1
 t {t
Ntb b a
b~
:
Nt a
For presentation, we scaled composite indices Nt by total
areas, obtaining a summary on the scale of birds/route, nt 5
Nt / Si Ai.

Sauer et al. N Mourning Doves in North America

1061

Bird Conservation Regions (BCRs) physiographic strata
within states; these state–physiographic regions are used in
BBS analyses as the fundamental strata for analysis (Sauer et
al. 2003). See Sauer et al. (2003) and Sauer et al. (1994) for a
discussion of strata and an analysis of BBS data within
BCR-based strata.
We summarized results by state and for the Eastern
[EMU], Central [CMU], and Western [WMU] Management Units (Dolton et al. 2007; see Table 1 for states
included in the management units) for CCS-seen, CCSheard, and BBS indices. We calculated long-term (1966–
2007) trends and annual indices. Due to small sample sizes
in the CCS, we present composite results for New England
states (denoted as NEE: CT, ME, MA, NH, RI, and VT),
and for Delaware and Maryland (denoted as DM; Dolton et
al. 2007), although we occasionally used trends for
individual states in subsequent analyses.
We compared results by evaluating differences in trends
by states and regions and present 95% CrI to assess
significance of differences. We evaluated whether consistent differences existed in magnitude of trends from
CCS-seen and CCS-heard indices by calculating mean
differences among states and associated 95% confidence
intervals. Dolton (1994) suggested that the CCS-seen index
tended to be less precise than the CCS-heard index. We also
compared relative precision of the indices by comparing
half-width of credible intervals of estimated trends by states,
estimating mean size of credible intervals and associated
95% confidence intervals.
Why Do Trends Differ Among Indices: 3 Hypotheses
Observed differences in population trends estimated from
dove indices suggest that environmental factors differentially
influenced the indices. The indices are collected simultaneously in the CCS; hence, we can interpret changes in
indices as being a consequence of regional changes in factors
that differentially affect the indices rather than actual
population change. If we can document associations
between hypothesized factors and differences in the indices,
we can get insights both into possible causes of differences
between indices and into means of controlling for the factors
in the analysis.
Several possible explanations exist.
1. Differential disturbance effects.—Presumably, increased noise disturbance will influence observers’ ability
to count doves. Effects of disturbance are likely to be
greater on the heard index, because noise competes
with hearing but not with visual effects. We can
directly address this differential disturbance hypothesis
using the hierarchical model and the disturbance index
collected in the CCS. We evaluated the relative importance of disturbance for the indices, estimated the shape
parameter p and the slope parameter B for heard and seen
indices, and compared the magnitude of these parameters
for each index. We also assessed the consequences of
disturbance for the analysis of both indices by estimating
annual indices of abundance and trend while controlling for
disturbance.
1062

2. Differential effects of land-use change along routes.—
Development along roads tends to create perches and increase
observers’ ability to see birds, and the CCS-seen index may
differentially increase after development due to the increased
visibility of doves. We associated state-level trends in dove
populations with 2 sources of information on land-use change:
the National Resource Inventory and change of general landuse types from the National Land Cover Database (NLCD;
Fry et al. 2009). States with larger changes in amount of open
land uses having greater divergence between CCS-heard and
CCS-seen indices would support the land-use hypothesis.
Information on changes in developed versus rural
nonfederal land acreages are summarized by the National
Resources Inventory for 1982–1997 (Nusser and Goebel
1997). We obtained reports of developed and rural
nonfederal land analyses of the National Resources Inventory (NRI) conducted by the Natural Resources Conservation Service (United States Department of Agriculture
2000). The NRI provides estimates at the state level and
estimates were available for 1982 and 1997. We estimated
yearly percentage change between those years and used
simple linear regression to estimate the slope in a linear
regression of population trend estimates on change in
amount of developed land, by state for CCS-seen and CCSheard, estimated for the period 1982–1997.
Land-use change data were also available from the NLCD
1992–2001 Land Cover Change Retrofit Product (Fry et al.
2009). Although based on a short interval, this data set is a
consistent analysis of Landsat data from the starting and
ending time periods using a consistent categorization,
allowing comparisons of broadly defined land-use categories
(open water, urban, barren, forest, grassland–shrub, agriculture, wetlands, ice–snow). We estimated percentage change
in urban land use by state.
We also summarized change in combined agricultural and
grassland–shrub land uses by state, and evaluated change in
these open land uses as a predictor of population trends in
CCS-heard and CCS-seen estimated over the 1992–2001
interval, as well as to differences in the trends from the indices
by state. Although the actual population consequences for
changes in agricultural land uses may vary by crop type
(Martin and Sauer 1993), it is likely that increases in these
land uses will influence visibility of doves and may, therefore,
be associated with differences in trends for the indices.
3. Differential phenology effects.—Spring has been
arriving sooner, and bird reproductive activities are beginning earlier in the year (Dunn and Winkler 1999). Earlier
breeding may cause the cooing rates to be declining during
the dove survey time (see review in Baskett 1993), and
change in the timing of reproduction may have a differential
effect on number of doves heard relative to seen. Although
no phenology data exist that document changes in breeding
times of doves, associations between surface air temperature
increases and earlier spring breeding have been shown in
tree swallows (Tachycineta bicolor) and other taxa (Dunn and
Winkler 1999, Abu-Asab et al. 2001).
We evaluated the hypotheses that differences in the
CCS-seen and CCS-heard indices were associated with
The Journal of Wildlife Management N 74(5)

Table 1. Estimated population trend (%/yr and 95% credible intervals [CrI]) from hierarchical model analysis for 1966–2007, based on Mourning Dove
Call-Count Survey (CCS) heard and seen data and North American Breeding Bird Survey (BBS) data for states (NEE represents combined New England
states; DM is DE and MD) and Eastern (EMU), Central (CMU), and Western (WMU) Management Units.
CCS-heard

CCS-seen

CrI
Region
DM
FL
GA
IL
IN
KY
LA
MI
MS
NC
OH
PA
SC
TN
VA
WI
WV
NEE
NJ
NY
EMU
AR
CO
IA
KS
MN
MO
MT
NE
NM
ND
OK
SD
TX
WY
Central
AZ
CA
ID
NV
OR
UT
WA
Western

BBS

CrI

CrI

%/yr

2.5%

97.5%

%/yr

2.5%

97.5%

%/yr

2.5%

97.5%

20.76
20.06
21.15
20.36
21.43
20.17
2.11
1.25
21.24
0.28
20.28
0.43
20.70
21.96
22.09
0.45
1.38
1.63
22.98
2.44
20.31
20.68
20.13
0.21
20.17
21.19
22.47
20.72
20.88
0.17
0.08
20.66
20.09
20.57
21.60
20.53
20.98
21.90
21.05
22.64
20.89
21.66
20.50
21.47

21.79
20.82
21.9
21.53
22.14
20.94
1.24
0.52
21.9
20.28
20.99
20.63
21.29
22.72
24.54
20.42
0.43
20.29
24.10
1.55
20.61
21.59
21.06
20.56
20.79
22.02
23.22
21.90
21.39
20.67
20.72
21.62
20.79
21.01
22.62
20.74
21.60
22.47
21.93
23.89
21.93
22.68
22.22
21.81

0.34
0.72
20.38
0.68
20.73
0.63
2.98
2.05
20.57
0.85
0.42
1.41
20.08
21.17
21.04
1.31
2.39
2.63
21.80
3.33
20.09
0.29
0.84
0.97
0.46
20.35
21.78
0.45
20.35
1.00
0.88
0.33
0.65
20.14
20.61
20.31
20.37
21.32
20.15
21.33
0.17
20.68
1.23
21.15

0.92
3.03
20.83
0.39
21.08
0.83
3.01
3.25
21.47
0.21
1.28
2.30
0.90
0.06
20.22
3.48
3.57
2.24
20.63
4.66
0.70
20.81
20.13
0.34
0.11
21.63
22.10
20.01
20.37
0.83
0.57
0.31
0.33
0.94
22.83
0.11
23.52
21.89
0.15
22.48
22.21
22.66
1.85
21.95

20.4
1.95
21.71
21.42
22.11
20.28
1.86
2.29
22.35
20.57
0.36
0.8
0.08
20.88
21.33
2.37
1.87
0.05
22.14
3.25
0.39
21.96
21.33
20.55
20.67
22.97
22.91
21.24
21.19
20.26
20.40
20.76
20.58
0.40
24.37
20.17
24.38
22.64
21.15
24.09
23.49
24.20
21.80
22.42

1.63
4.76
0.37
2.79
20.10
2.29
4.58
4.27
0.12
1.19
1.88
3.86
2.02
1.23
1.07
4.23
5.57
3.72
1.02
5.40
1.15
0.58
1.19
1.71
0.87
20.30
21.09
1.18
0.13
2.01
1.43
1.43
1.12
1.60
21.56
0.39
22.31
20.84
2.33
0.50
20.04
21.41
2.03
21.05

0.58
2.82
20.80
1.10
0.04
0.88
2.84
1.72
20.48
0.38
1.55
1.91
0.05
20.18
20.08
1.98
5.19
3.10
0.76
2.29
0.77
0.71
0.23
0.46
0.43
21.17
21.61
20.78
20.17
20.15
0.77
21.32
0.23
20.80
20.75
20.33
21.14
20.52
20.10
20.71
20.46
20.55
0.54
20.70

0.26
2.22
21.25
0.6
20.51
0.36
2.11
1.21
21.27
20.12
1.03
1.44
20.58
20.78
20.54
1.48
4.48
2.62
20.06
1.87
0.63
20.08
20.34
20.12
20.16
21.72
22.15
21.44
20.73
20.94
0.16
21.87
20.37
21.18
21.55
20.51
22.04
21.03
21.17
21.92
21.48
21.54
21.24
21.11

0.91
3.43
20.38
1.59
0.59
1.43
3.55
2.25
0.34
0.87
2.08
2.37
0.67
0.42
0.38
2.47
5.98
3.58
1.51
2.70
0.90
1.5
0.82
1.05
1.03
20.61
21.04
20.08
0.41
0.61
1.39
20.73
0.87
20.43
0.05
20.16
20.27
20.01
0.94
0.52
0.51
0.46
2.28
20.30

changes in temperatures for May, the month of the survey.
Monthly estimated temperature by state was available
from the National Oceanic and Atmospheric Administration (National Climatic Data Center 2009). Most
doves initiate breeding considerably earlier in the year,
and we used the temperature from the survey period as a
measure of the seasonal progression by the time of the
survey. If the heard index decreased as more doves paired,
the trend in birds seen would increase differentially
relative to trends based on the CCS-heard index as
temperature increased. We used simple linear regression
to evaluate this hypothesis. States were the replicates in
the analysis (N 5 47); we did not include Rhode Island in
the analysis due to small sample size of routes in the
state.
Sauer et al. N Mourning Doves in North America

RESULTS
We conducted 5 primary hierarchical model analyses and 2
route-regression analyses. We based the CCS-heard analysis
on 40,013 counts by 8,077 observers, the CCS-seen analysis on
39,894 counts by 8,048 observers, the CCS-heard-disturbance
analysis on 37,065 counts by 7,537 observers, the CCS-seendisturbance analysis on 37,032 counts by 7,519 observers, and
the BBS analysis on 77,116 counts by 13,294 observers. Minor
differences in sample sizes reflect occasional missing data and
other inconsistencies in route data.
Population Trend Estimated From CCS-Heard and
CCS-Seen Indices
Trend estimates based on CCS-heard and CCS-seen indices
differed in some states and regions. In the EMU, the 1966–
1063

2007 mourning dove population trend was 20.31 (CrI 5
20.061, 20.09) based on CCS-heard data and 0.70 (CrI 5
0.39, 1.15) based on CCS-seen data. Trends between CCSheard and CCS-seen indices were different (as defined by
nonoverlapping CrIs) in Florida, Michigan, South Carolina,
Tennessee, and Wisconsin (Table 1). Trends from CCS-seen
indices were more positive in 19 of the 21 EMU states. On
average, CCS-seen trends were larger by 1.5% (CI 5 1.07,
21.95). However, trends based on the CCS-heard index
were more precise, because half-widths of the credible
intervals of the CCS-seen indices were 0.26 (CI 5 0.15,
0.37) larger than those based on CCS-heard indices.
Estimated trends from CCS indices were not consistent
among regions. In the CMU, the 1966–2007 trend was 20.53
(CrI5 20.74, 20.31) based on CCS-heard data and 0.11
(CrI5 20.17, 0.39) based on CCS-seen data. Trends were
different (as defined by nonoverlapping CrIs) only in Texas
(Table 1). Trends based on the CCS-seen index were more
positive in 11 of the 14 CMU states, but on average, CCSseen trends were only slightly larger than the CCS-heard
trends (0.3; CI 5 20.04, 20.64). Trends based on the CCSheard index were more precise, because half-widths of the
credible intervals of the CCS-seen indices were 0.23 (CI 5
0.15, 0.30) larger than those based on CCS-heard indices.
In the WMU, the 1966–2007 trend was 21.47 (CrI 5
21.81, 21.15) based on CCS-heard data and 21.94 (CrI 5
22.42, 21.05) based on CCS-seen data. Trends differed (as
defined by nonoverlapping CrIs) only in Arizona, where the
CCS-seen trend was more negative than the CCS-heard
trend (Table 1). Relative to CCS-heard trends, CCS-seen
trends were more positive in 4 of the 7 WMU states. On
average, CCS-seen trends were similar (20.16 mean difference; CI 5 21.37, 21.05). Trends based on the CCS-heard
index were more precise, because half-widths of the credible
intervals of the CCS-seen indices were 0.58 (CI 5 0.10,
1.06) larger than those based on CCS-heard indices.
Estimating equation (EE) results and hierarchical model
(HM) results were similar for CCS-heard and CCS-seen
indices. Estimates for CCS-heard were similar in magnitude
by region (EMU: 20.31 [HM] vs. 20.37 [EE]; CMU:
20.53 [HM] vs. 20.70 [EE]; WMU: 21.47 [HM] vs.
21.86 [EE]). For heard data, confidence intervals from EE
trend estimates were larger than the CrI from HM estimates
in all regions (0.65 [CI 5 0.43, 0.86] in EMU, 0.73 [CI 5
0.34, 1.11] in CMU, and 0.80 [CrI 5 0.38, 1.21] in WMU).
Estimates for CCS-seen showed similar patterns and were
similar in magnitude by region (EMU: 0.70 [HM] vs. 0.39
[EE]; CMU: 0.11 [HM] vs. 0.06 [EE]; WMU: 21.94 [HM]
vs. 22.09 [EE]). In only 11 of the 42 states were estimating
equation estimates larger than hierarchical model estimates.
For seen data, confidence intervals from EE trend estimates
were larger than the CrI from HM estimates in all regions
(0.71 [CI 5 0.40, 1.01] in EMU, 0.69 [CI 5 0.35, 1.03] in
CMU, and 1.29 [CI 5 0.48, 2.19] in WMU).
Disturbance and CCS Trends
Change in amounts of noise-disturbance varied within and
among regions. In the EMU, Florida, Michigan, Pennsyl1064

vania, South Carolina, and New York had a positive trend in
disturbance; Mississippi and North Carolina had declining
rates of disturbance. Overall in the EMU, disturbance
changed little over 1966–2007 (Fig. 1). In the CMU no
state had increasing disturbance; Nebraska and South
Dakota had negative trends, and overall disturbance
declined at a rate of 0.3%/yr in the CMU. In the WMU,
only Washington had credible intervals indicating an
increase in disturbance over time. Overall in the WMU,
the credible interval of the estimated trend in disturbance of
0.34%/yr overlapped zero (Fig. 1).
Disturbance influenced CCS-heard indices but had less
effect on CCS-seen indices. For CCS-heard, we estimated
the parameter p that determined shape of the relationship
between disturbance and counts as 20.52 (CrI 5 20.82,
20.23) and the coefficient B that described strength of the
relationship as 0.04 (CrI 5 0.03, 0.06). For CCS-seen, the
estimated shape parameter p was 0.73 (CrI 5 21.19, 3.7)
and B was 0.02 (CrI 5 20.01, 0.06). Differences in p had
large effects on the shape of the disturbance effects, as the
significant disturbance effects predicted extensive diminution of the CCS-heard index with higher levels of
disturbance (Fig. 2).
Although effort effects were significant for CCS-heard,
controlling for effort in the analysis of population change
had only minor consequences for regional estimation of
population change. Estimates of long-term trend were
almost identical in the EMU (20.31 without disturbance
covariable, 20.32 with the covariable), similar in the CMU
(20.53 vs. 20.51), and in the WMU the estimate was less
negative when we included disturbance (21.47 without,
21.19 with), although overlapping credible intervals
indicated a lack of significance. Incorporating the disturbance covariate led to trend estimates larger in magnitude in
27 of the 42 states in the analysis.
Incorporating disturbance effects had little influence on
comparison of CCS-heard and CCS-seen indices. In the
EMU, the 1966–2007 trend was 20.32 (CrI 5 20.62,
20.09) based on CCS-heard data and 0.86 (CrI 5 0.56,
1.15) based on CCS-seen data. In the CMU, the 1966–
2007 trend was 20.51 (CrI 5 20.74, 20.29) based on
CCS-heard data and 0.07 (CrI 5 20.25, 0.39) based on
CCS-seen data. In the WMU, the 1966–2007 trend was
21.18 (CrI 5 21.55, 20.83) based on CCS-heard data and
21.55 (CrI 5 22.07, 21.05) based on CCS-seen data.
Temperature and Land-Use Associations with
CCS Trends
Trends based on the CCS-seen index tended to be more
negative in states where there was a positive trend in May
temperatures (Fig. 3; slope of regression 22,335 [CI 5
24,051, 2618]). Trends based on the CCS-heard index
had a weaker association (slope 5 2934 [CI 5 22,300,
432]), and the difference between the 2 indices also was
associated with trends in May temperature (slope 5 21,400
[CI 5 22,369, 2433]).
There was a positive, although weak, relationship between
differences in CCS indices (CCS-seen – CCS-heard) and
The Journal of Wildlife Management N 74(5)

Figure 1. Estimated trend in noise-related disturbance (% change/yr 6 95% credible intervals) for 1968–2007 based on data collected during the Mourning
Dove Call-Count Survey for states across the United States (NEE is New England; DM is DE and MD combined) and Eastern (EMU), Central (CMU),
and Western (WMU) Management Units.

change in the NRI measure of developed land by state (slope
5 0.51 [CI 5 20.15, 1.17]; Fig. 4). This relationship also
appeared when we used the longer term (1966–2007) dove
change estimates in the analysis (slope 5 0.40 [20.03, 0.83]).
For the 1992–2001 land-use change analysis using NLCD
data, associations of CCS trends with change in land use
showed positive associations of nonurban open land-use
changes and trends in both CCS-heard (slope 5 1.26 [CI 5
0.25, 2.27]) and CCS-seen (slope 1.52 [CI 5 0.03, 3.02]);
slopes were similar for the indices. Slopes of the associations
of change in urban land uses were weakly negatively
associated with both CCS-heard (slope 5 21.03 [CI 5
22.78, 0.72]) and CCS-seen (slope 20.98 [CI 5 23.58,
1.62]) indices. There were no apparent differential effects of
changes in either land-use type on the CCS indices.
BBS
The BBS showed a more optimistic view of dove
populations than did the CCS-heard. We note that, due

to variation in the starting year of the BBS by region, the
CMU results represent the interval 1967–2007 and the
WMU results are for 1968–2007. In the EMU, the 1966–
2007 trend was 0.77 (CrI 5 0.63, 0.90) based on BBS,
similar to CCS-seen results (0.70) but larger than the 20.31
based on CCS-heard data. Trends differed between BBS
and CCS-heard in Florida, Indiana, Ohio, Pennsylvania,
Tennessee, Virginia, Wisconsin, and New Jersey, and
magnitudes of trends differed by .1.5% between indices
in several other states (Table 1). Trends based on CCS-seen
were more consistent with BBS results, varying only in New
York. Trends based on the BBS were more precise than
CCS indices; half-widths of the credible intervals of trends
based on the CCS-heard indices were 0.36 (CI 5 0.21,
0.50) larger, and CCS-seen indices were 0.63 (CI 5 0.49,
0.77) larger, than those based on BBS indices.
Patterns of change among CCS and BBS indices were not
consistent among regions. In the CMU, the 1966–2007
BBS trend was 20.33 (CrI 5 20.51, 20.16), more positive

Figure 2. Proportional effect of disturbance (sum of stops with moderate
or high disturbance) on Mourning Dove Call-Count Survey (CCS-heard
[diamonds] and CCS-seen [boxes]) indices based on data from the
contiguous United States, 1968–2007.

Figure 3. Association of yearly percentage change in average May
temperature and Mourning Dove Call-Count Survey (CCS)-heard
(diamonds, solid line) and CCS-Seen (boxes, dashed line) trend estimates
from the contiguous United States over the interval 1966–2007.

Sauer et al. N Mourning Doves in North America

1065

Figure 4. Association of yearly percentage change in developed land and
difference in Mourning Dove Call-Count Survey (CCS)-heard (diamonds,
solid line) and CCS-seen (boxes, dashed line) trend estimates from the
contiguous United States over the interval 1966–2007.

than 20.53 based on CCS-heard data but more negative
than 0.11 based on CCS-seen data. Trends were different
(as defined by nonoverlapping CrIs) only in the CCS-seen
index for Texas (Table 1). Trends based on the BBS were
more precise than CCS indices, because half-widths of the
credible intervals of trends based on the CCS-heard indices
were 0.19 (CI 5 0.11, 0.27) larger, and trends based on
CCS-seen indices were 0.41 (CI 5 0.31, 0.52) larger, than
those based on BBS indices.
In the WMU, the 1968–2007 BBS trend was 20.70 (CrI
5 21.11, 20.30), more positive than 21.47 based on CCSheard data and 21.94 from CCS-seen data. Trends were
different (as defined by nonoverlapping CrIs) between BBS
results and CCS results for both indices for California
(Table 1). Trends based on the BBS were not more precise
than CCS-heard indices in the WMU; half-widths of the
credible intervals of the BBS indices were 0.04 (CI 5 0.05,
0.14) larger. However, half-widths of credible intervals of
trends from CCS-seen indices were 0.54 (CI 5 0.05, 1.03)
larger than those based on BBS indices.
Annual Indices of Abundance
We present annual indices of abundance for EMU, CMU,
and WMU for CCS-heard, CCS-seen, and BBS results
(Fig. 5). The BBS index is larger because it is based on over
twice the number of stops as the CCS. In the EMU,
fluctuations caused by severe winters in 1976–1977 are
evident in the time series, but the CCS-seen and BBS results
plot a generally increasing trajectory, whereas the CCS-heard
results do not show the increase. As mentioned earlier,
trajectories of CCS-seen and CCS-heard indices appear more
divergent in recent years. The CCS-heard index adjusted for
disturbance is slightly larger than the unadjusted heard index
in the early years, but shows similar patterns in later years. In
the CMU, BBS results and CCS-heard data show declines,
whereas CCS-seen shows a stable population. Adjusting for
disturbance does not change the CCS-heard index (Fig. 5).
In the WMU, all annual indices are similar.
Florida and Texas are influential states in which indices
show differing patterns of population change (Fig. 6). We
1066

Figure 5. Annual indices of abundance for mourning doves in the Eastern,
Central, and Western Management Units, based on Call-Count Survey
(CCS)-seen (boxes), CCS-heard (diamonds), indices adjusted for disturbance for the CCS-heard index (x), and North American Breeding Bird
Survey data (triangles) over the interval 1966–2007.

plotted human census data (United States Census Bureau
2009) to illustrate the association of dove counts and human
population. Human population increases with the CCSseen indices in the states, but phenology (as summarized by
mean temp data) has little association with dove indices
(Fig. 6).

DISCUSSION
The log-linear hierarchical model provides a convenient
framework for analysis of mourning dove population change
from CCS and BBS data. For the first time, we were able to
directly include inexperience effects of observers and
disturbance effects in the model and calculate credible
intervals for the annual indices. The hierarchical model is
easily modified to include both additional covariates and
The Journal of Wildlife Management N 74(5)

Figure 6. Annual indices of abundance for mourning doves in Florida and
Texas, based on Call-Count Survey (CCS)-seen (boxes), CCS-heard
(diamonds), North American Breeding Bird Survey data (triangles), and the
CCS-heard-disturbance index (x) for the interval 1966–2007. We also
present time series of mean May temperatures (+) and a linear description of
human population estimates scaled to the dove CCS-seen population index
in 1970. We present the 3 data points for human population with dashes.

spatial effects (e.g., Thogmartin et al. 2007). We directly
documented effects of disturbance, showed that this effect
differed between the CCS-seen and CCS-heard indices,
and estimated population change in doves while controlling
for disturbance in analysis of count data.
Hierarchical model results also documented differences in
trend estimates among the CCS-heard, CCS-seen, and
BBS indices in several states (notably TX, FL, and AZ) and
in Dove Management Units. Estimates of population
change based on the CCS-heard index tended to provide
more negative estimates in areas in which they differed from
the CCS-seen index. Analyses based on BBS results, which
contain information on doves either seen or heard, generally
tended to provide estimates either similar to the CCS-seen
index or intermediate between results of CCS-heard and
CCS-seen analyses.
These inconsistencies complicate our use of dove survey
results for management, because it is unclear whether any of
the indices accurately reflect actual changes in the dove
population. Identification of environmental features that are
differentially influencing the indices and appropriate
incorporation of these features into analyses to control for
these differential effects of counting should be a high
priority for survey analysis. Our hierarchical models provide
an appropriate framework for incorporating covariates that
influence detection of doves.
Sauer et al. N Mourning Doves in North America

Factors That Influence Indices
It is likely that disturbance, phenology, and land-use
changes all influenced the differences we documented in
the CCS indices. All of these influences vary regionally,
complicating analysis and interpretation of results. All
correlations must be viewed with caution, but they do allow
us to evaluate the consistency of hypotheses regarding
changes in the indices.
All roadside bird counters understand the limitations on
counting imposed by ambient noise and traffic, and we
documented that several of the states in which estimated
dove trends differed between CCS indices have also
experienced increases in disturbance over the survey period.
We found that CCS indices were differentially affected by
disturbance. Whereas the CCS-seen index was influenced
little by disturbance the CCS-heard index declined as
amount of disturbance increased. However, disturbance was
not consistently increasing over the survey area, and the
actual consequences of controlling for disturbance were
slight for the CCS indices; controlling for changes in
disturbance did not eliminate differences in estimates.
Most noise-based disturbance is human-mediated. Increases in vehicle traffic, lawn mowing, and other domestic
aspects of increasing urbanization accompany increasing
human population along routes and correlate with development and other land-use changes. Consequently, temporal
trends in disturbance and the covariate effects of disturbance
may reflect the influence of broader land-use change effects.
Although not presently a critical component of the
analysis, there is a clear need to monitor disturbance and
retain it as a possible covariate for future analyses. More
investigation is also needed into refining the disturbance
covariate. As with counts, it is likely that there is much
observer interpretation of disturbance, and although our
analysis of trends in disturbance controlled for these
observer effects, the disturbance covariate did not. There
are also many ways to summarize disturbance, and our
measure tended to emphasize strong disturbance at stops.
Exploration of alternative disturbance metrics may clarify
the incremental effects of disturbance on counts.
Although phenology likely affects both CCS-heard and
CCS-seen data, the negative association of mean May
temperature with indices was stronger for CCS-seen,
contrary to our hypothesis. Changes in May temperatures
were often slight, and May temperatures have exhibited
much year-to-year variation in recent years (e.g., Fig. 6),
and changes in May temperature did not explain disagreements between dove indices in Florida and Texas. However,
phenology is associated with the seen index, particularly in
the central and western United States where more doves are
seen than heard along CCS routes. Patterson (2005) noted
that National Oceanic and Atmospheric Administration
temperature data documented increases in spring temperature in recent decades in the western United States but little
change in the southeastern United States. The influence of
phenology on dove counting clearly merits further study, as
trends in temperature data make it more likely that any
1067

temperature-related effects on phenology will become more
prevalent in the future (Patterson 2005).
Associations we noted between rates of development and
differences in trend estimates suggested that development
may lead to increases in numbers of doves seen. However,
weak associations (even for 1966–2007 trend estimates, the
CrI of the slope estimate still slightly overlapped zero) did not
permit definitive statements about relationships between
indices and land use. Similarities in rate of human population
change and rate of increase in the CCS-seen indices in
Florida and Texas are also consistent with the notion that
human-related development may be differentially influencing
the CCS-seen index. Data were not available for recent years,
even though changes between heard and seen indices may be
most dramatic in recent years. Further exploration of the
association of recent changes in dove indices and land-use
changes may provide better insights into relationships
between doves seen and land-use changes.
Limitations of Association Analyses
Our evaluation of factors influencing seen and heard dove
indices should be viewed as exploratory, because we based it
on observational data and it was limited by availability of
covariates at appropriate spatial and temporal scales. Disturbance could be convincingly accommodated in the analysis
because it was collected as part of the survey; other covariates
were available only at the scale of states and for only a few
times during the 42-year period in which the CCS has been
conducted. States were not a particularly useful replicate for
analyses, because they vary in size and many factors also vary
spatially and are likely confounded in the analysis. Finally,
review of the indices suggests that differences in CCS-seen
versus CCS-heard indices may be more pronounced in recent
years. A more informative analysis of land-use change may be
possible when recent (post-1995) land-use change data
become available for comparison with dove indices.
A critical goal of future analyses must be to more
effectively use covariate analyses to control for factors that
influence detectability of doves. The best way to control for
detectability is to collect information (or obtain better
information as it becomes available) at the scale of survey
routes, allowing for analyses at the scale of the sample units
of the survey. Better data collected at the scale of survey
routes would facilitate direct incorporation of these
covariates into the hierarchical analysis. Texas and Florida
are states that merit more intensive analyses of effects of
land-use change on dove populations, because those states
have large samples of survey routes, notable differences in
index results, and are rapidly urbanizing.

MANAGEMENT IMPLICATIONS
Hierarchical models should be used for analysis of CCS data
in future Mourning Dove Status Reports (Dolton et al.
2007). Hierarchical models provide a coherent framework
for appropriate estimation of estimates and variances,
provide efficient estimates of change derived directly from
the annual indices, and allow for easy accommodation of
covariates. The hierarchical structure is also a natural
1068

framework for 2 additional modeling activities needed for
management: 1) composite analyses of CCS and BBS data,
a topic currently under consideration by managers (National
Mourning Dove Planning Committee 2003; D. R. Otis,
United States Geological Survey, unpublished data), and 2)
development of demographic models that use CCS data and
survival and productivity data to make predictions of future
populations needed for harvest management (Nichols and
Williams 2006). Composite models that integrate surveys
and demographic data would facilitate direct use of CCS
and BBS data in dove population management.
However, differences in results we documented among
analyses based on CCS-seen, CCS-heard, and BBS indices
suggest the need for additional studies before implementation of models that integrate surveys. All indices are not
equal, in terms of both precision and bias in estimation, and
there is a clear need to better understand the limitations of
each index. Further evaluations of effects of development
and changes in phenology on the indices are needed and will
be critical in increasing our understanding of environmental
influences on detectability of mourning doves. Opportunities exist to conduct studies at the scale of individual stops to
associate changes in CCS-heard and CCS-seen indices with
temperature and land-use information, and these sitespecific analyses should provide more powerful tests of
hypotheses about factors influencing detectability. Hierarchical models provide a coherent analysis framework that
will be essential for these multiple-scale analyses of
environmental influences on dove indices.
Finally, our results clearly indicate that even wellimplemented index surveys may not always be interpretable
due to unmodeled changes in detection of doves; reducing
uncertainty in survey results must be a priority for managers.
The most direct way of reducing this uncertainty is to
implement sampling procedures that permit estimation of
detection rates during surveys, and we encourage continued
investigation of ways to efficiently estimate detectability
from both the CCS and BBS.

ACKNOWLEDGMENTS
We thank the BBS and CCS observers and coordinators
who collected and managed the data. J. E. Fallon, K. Parker,
K. Pardieck, and R. Rao provided data and assistance with
data management. Editorial comments by M. D. Koneff, J.
T. Kelley, K. Parker, C. S. Robbins, T. R. Sanders, G. W.
Smith, and anonymous reviewers greatly improved the
quality of the manuscript.

LITERATURE CITED
Abu-Asab, M. S., P. M. Peterson, S. G. Shetler, and S. S. Orli. 2001.
Earlier plant flowering in spring as a response to global warming in the
Washington, DC, area. Biodiversity and Conservation 10:597–612.
Baskett, T. S. 1993. Biological evaluation of the call-count survey. Pages
253–268 in T. S. Baskett, M. W. Sayre, R. E. Tomlinson, and R. E.
Mirarchi, editors. Ecology and management of the mourning dove.
Stackpole, Harrisburg, Pennsylvania, USA.
Bibby, C. J., N. D. Burgess, D. A. Hill, and S. Mustoe. 2000. Bird census
techniques. Second edition. Academic Press, New York, New York,
USA.
The Journal of Wildlife Management N 74(5)

Dolton, D. D. 1993. The call-count survey: historic development and
current procedures. Pages 233–252 in T. S. Baskett, M. W. Sayre, R. E.
Tomlinson, and R. E. Mirarchi, editors. Ecology and management of the
mourning dove. Stackpole, Harrisburg, Pennsylvania, USA.
Dolton, D. D., R. D. Rau, and K. Parker. 2007. Mourning dove population
status, 2007. U.S. Fish and Wildlife Service, Laurel, Maryland, USA.
Dunn, P. O., and D. W. Winkler. 1999. Climate change has affected the
breeding date of tree swallows throughout North America. Proceedings
of the Royal Society of London, Series B 266:2487–2490.
Fry, J. A., M. J. Coan, C. G. Homer, D. K. Meyer, and J. D. Wickham.
2009. Completion of the National Land Cover Database (NLCD) 1992–
2001 Land Cover Change Retrofit product. U.S. Geological Survey
Open-File Report 2008-1379, Reston, Virginia, USA.
Geissler, P. H., and J. R. Sauer. 1990. Topics in route regression analysis.
Pages 54–57 in J. R. Sauer and S. Droege, editors. Survey designs and
statistical methods for the estimation of avian population trends. U.S.
Fish and Wildlife Service, Biological Report 90(1), Washington, D.C.,
USA.
Gilks, W. R., S. Richardson, and D. J. Spiegelhalter, editors. 1996. Markov
chain Monte Carlo in practice. Chapman and Hall, London, United
Kingdom.
Link, W. A., and J. R. Sauer. 1994. Estimating equations estimates of
trends. Bird Populations 2:23–32.
Link, W. A., and J. R. Sauer. 2002. A hierarchical model of population
change with application to Cerulean Warblers. Ecology 83:2832–2840.
Link, W. A., J. R. Sauer, and D. K. Niven. 2006. A hierarchical model for
regional analysis of population change using Christmas Bird Count data,
with application to the American Black Duck. Condor 108:13–24.
Lunn, D. J., A. Thomas, N. Best, and D. Spiegelhalter. 2000. WinBUGS—a Bayesian modelling framework: concepts, structure, and
extensibility. Statistics and Computing 10:325–337.
Martin, F. W., and J. R. Sauer. 1993. The Eastern Management Unit.
Pages 281–304 in T. S. Baskett, M. W. Sayre, R. E. Tomlinson, and R.
E. Mirarchi, editors. The mourning dove: ecology and management.
Stackpole, Harrisburg, Pennsylvania, USA.
National Climatic Data Center. 2009. Climate services and monitoring
division website. ,http://www.ncdc.noaa.gov/oa/climate/research/
monitoring.html.. Accessed 20 Jan 2010.
National Mourning Dove Planning Committee. 2003. Mourning dove
national strategic harvest management plan. U.S. Fish and Wildlife
Service, Washington, D.C., USA.
Nichols, J. D., and B. K. Williams. 2006. Monitoring for conservation.
Trends in Ecology and Evolution 21:668–673.

Sauer et al. N Mourning Doves in North America

Nusser, S. M., and J. J. Goebel. 1997. The national resources inventory: a
long-term multi-resource monitoring programme. Environmental and
Ecological Statistics 4:181–204.
Otis, D. L., J. H. Schulz, D. Miller, R. E. Mirarchi, and T. S. Baskett.
2008. Mourning dove (Zenaida macroura). Account 117 in A. Poole,
editor. The birds of North America online. Cornell Lab of Ornithology,
Ithaca, New York, USA. ,http://bna.birds.cornell.edu/bna/species/
117.. Accessed 20 Jan 2010.
Patterson, J. 2005. An analysis of spring bird migration phenology in
Kansas. Thesis, Kansas State University, Manhattan, USA.
Richkus, K. D., K. A. Wilkins, R. V. Raftovich, S. S. Williams, and H. L.
Spriggs. 2008. Migratory bird hunting activity and harvest during the
2006 and 2007 hunting seasons: preliminary estimates. U.S. Fish and
Wildlife Service, Laurel, Maryland, USA.
Sauer, J. R., D. D. Dolton, and S. Droege. 1994. Mourning dove
population trend estimates from call-count and North American breeding
bird surveys. Journal of Wildlife Management 58:506–515.
Sauer, J. R., J. E. Fallon, and R. Johnson. 2003. Use of North American
Breeding Bird Survey data to estimate population change for bird
conservation regions. Journal of Wildlife Management 67:372–389.
Sauer, J. R., J. E. Hines, and J. Fallon. 2008a. The North American
breeding bird survey, results and analysis 1966–2007. Version 5.15.2008.
U.S. Geological Survey Patuxent Wildlife Research Center, Laurel,
Maryland, USA. ,http://www.mbr-pwrc.usgs.gov/bbs/bbs.html.. Accessed 20 Jan 2010.
Sauer, J. R., W. A. Link, W. L. Kendall, J. R. Kelley, and D. K. Niven.
2008b. A hierarchical model for estimating change in American
woodcock populations. Journal of Wildlife Management 72:204–214.
Thogmartin, W., J. R. Sauer, and M. G. Knutson. 2007. Modeling and
mapping abundance of American woodcock across the midwestern and
northeastern United States. Journal of Wildlife Management 71:376–
382.
United States Census Bureau. 2009. Population estimates website. ,http://
www.census.gov/popest/states/states.html.. Accessed 20 Jan 2010.
United States Department of Agriculture. 2000. Summary report 1997
national resources inventory (revised December 2000). Natural Resources
Conservation Service, Iowa State University Statistical Laboratory, Ames,
USA.
United States Department of the Interior. 2005. Mourning dove national
strategic harvest management plan. U.S. Department of the Interior, U.S.
Fish and Wildlife Service, Washington, D.C., USA.

Associate Editor: Rosenstock.

1069