Behavioral Ecology Vol. 13 No. 4: 543-550
© 2002 International Society for Behavioral Ecology
Sources of variation in breeding-ground fidelity of mallards (Anas platyrhynchos)
a U.S. Geographic Service, Patuxent Wildlife Research Center, 12100 Beech Forest Road, Laurel, MD 20708, USA b U.S. Fish and Wildlife Service, Patuxent Wildlife Research Center, 12100 Beech Forest Road, Laurel, MD 20708, USA
Address correspondence to P.F. Doherty, Jr., Carlsbad Fish and Wildlife Service Office, 2730 Loker Ave. West, Carlsbad, CA 92008, USA. E-mail: paul_doherty{at}fws.gov .
Received 30 January 2001; revised 30 October 2001; accepted 21 November 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Generalizations used to support hypotheses about the evolution of fidelity to breeding areas in birds include the tendency for fidelity to be greater in adult birds than in yearlings. In ducks, in contrast to most bird species, fidelity is thought to be greater among females than males. Researchers have suggested that fidelity in ducks is positively correlated with pond availability. However, most estimates of fidelity on which these inferences have been based represent functions of survival and recapture—resighting probabilities in addition to fidelity. We applied the modeling approach developed by Burnham to recapture and band recovery data of mallard ducks to test the above hypotheses about fidelity. We found little evidence of sex differences in adult philopatry, with females being slightly more philopatric than males in one study area, but not in a second study area. However, yearling females were more philopatric than yearling males in both study areas. We found that adults were generally more philopatric than yearlings. We could find no relationship between fidelity and pond availability. Our results, while partially supporting current theory concerning sex and age differences in philopatry, suggest that adult male mallards are more philopatric than once thought, and we recommend that other generalizations about philopatry be revisited with proper estimation techniques.
Key words: Anas platyrhynchos, fidelity, mallards, mark—recapture, philopatry.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
More than 20 years ago Greenwood (1980
) published an
influential paper concerning philopatry in birds and mammals. In his
literature review, Greenwood noted a pre-dominance of male-biased dispersal in
mammals and female-biased dispersal in birds. Greenwood hypothesized these
major differences to be a function of differing mating systems. He posited
that in a mate-defense mating system, a system in which resources are not
defendable, females gain from knowledge of a local area and thus are more
likely to be philopatric. In such a system, males compete directly for access
to females and are likely to disperse in search of mates. Polygyny is favored
in this system and is predominant among mammals. In contrast, within
resource-defense systems, in which defendable resources, or territories, are
the prerequisites for obtaining a mate, males use their prior knowledge of a
local area to locate and defend the best territories. In this system females
are free to search for the highest quality mate (and defended resources) and
are more likely to disperse. This situation is common in most birds
(Greenwood and Harvey, 1982).
Although other factors have been suggested to be important in the evolution of
sex differences in dispersal, such as intrasexual competition
(Dobson, 1982; but see
Johnson and Gaines, 1990),
inbreeding avoidance (Wolff,
1992), and parent—offspring competition
(Anderson et al., 1992;
Waser and Jones, 1983), it is
recognized that the mating system perspective not only explains the common
fidelity patterns seen in mammals and birds, but also notable exceptions.
One notable exception in birds is among the waterfowl (family Anatidae). In
contrast to most other bird species, waterfowl show female-biased natal and
breeding philopatry (Clark et al., 1997;
Greenwood and Harvey, 1982).
As reviewed by Rohwer and Anderson
(1988), most waterfowl are
migratory and return to the breeding grounds already paired, with pairing
typically occurring on the wintering grounds. Pairs most often return to the
natal area of the female or where the female previously nested. Females are
generally responsible for incubation and rearing of the young. Knowledge of,
and familiarity with, resources in the breeding area can lead to improved
nesting success and female survival, improved feeding efficiency, and improved
brood-rearing success, especially with older, more experienced birds
(Anderson et al., 1992 and
references within).
Generally, male ducks are not able to defend breeding-ground resources, and
they also abandon their mates during incubation
(Rohwer and Anderson, 1988).
Swans and geese are notable exceptions in which the family unit stays together
at least until fall migration, with males and females having long-term pair
bonds (Rohwer and Anderson,
1988 and references within).
Within waterfowl in general, and ducks in particular, females often have
higher mortality rates, probably due to their greater role in parental care,
and waterfowl generally have a male-biased sex ratio
(Johnson et al., 1992;
Sargeant and Raveling, 1992).
This allows for females to have a choice of mates on the wintering grounds and
to possibly gain benefits from male attendance during the winter as they
accumulate energy reserves (Rohwer and
Anderson, 1988), with males then following females to their
preferred breeding areas.
Thus, following this scenario, female ducks of species such as the mallard
(Anas platyrhynchos) are thought to be more philopatric than males,
and adults are more philopatric than juvenile birds
(Anderson et al., 1992;
Johnson and Grier, 1988;
Rohwer and Anderson, 1988). In
addition, philopatry is thought to covary with wetland conditions on the
breeding grounds. It is clear that wetland conditions influence duck
reproduction (e.g., Miller and Newton,
1999; Pospahala et al.,
1974), and if pond numbers in a region are high, as in a very wet
year, ducks are more likely to be philopatric. If regional conditions are dry,
ducks are more likely to breed elsewhere. Johnson and Grier
(1988) inferred this
relationship from aerial survey data for a number of duck species, including
mallards. Another possible relationship involving wetland conditions is that
mallards may use their experience and wetland conditions from the current year
as an index to future habitat and breeding success
(Danchin and Wagner, 1997,
Danchin et al., 1998).
Despite the widespread interest in philopatry, especially for waterfowl,
robust estimates of fidelity are lacking. Most of the evidence for the
empirical generalizations reviewed by Greenwood
(1980) and Greenwood and
Harvey (1982) is based on ad
hoc estimators such as return rate (typically estimated as the number of
marked animals recaptured or resighted in an area of interest in year
i divided by the number of animals marked in that same area the
previous year, i - 1), and homing rate (sometimes estimated as return
rate divided by average annual survival; e.g., see
Johnson et al., 1992;
Lokemoen et al., 1990). These
ad hoc measures are functions of the probability of survival, the probability
of fidelity given the individual has survived, and the recapture or resighting
probability given the individual has survived and returned to the area. Some
inferences about fidelity are based on distance measures, such as comparisons
of average observed distances moved between breeding seasons by each sex
(Clarke et al., 1997;
Greenwood, 1980;
Robertson and Cooke, 1999;
Waser and Jones, 1983).
However, such measures depend on the detection (recapture, resighting,
recovery) probabilities associated with the different locations in which
animals can be recorded. All of the aforementioned estimators assume that at
least some of the component probabilities are constant for different sexes,
age groups, times, and areas; these assumptions are clearly dubious in many
situations (Anderson et al.,
1992). For example sex-specific behaviors, such as brood
attendance or tendencies for prebreeders to group together, could lead to
detection probabilities not being constant for all birds.
Burnham (1993; also see
Szymczak and Rexstad, 1991)
presented a model to estimate survival, probability of recapture, band
reporting rates, and fidelity by combining both live recapture and dead
recovery data. The basic idea underlying Burnham's
(1993) approach is that
recaptures are restricted to specific sampling locations and that the apparent
survival rates estimated using standard capture—recapture models
(Cormack, 1964;
Jolly, 1965;
Seber, 1965) with such data
thus estimate the product of true survival and fidelity (probability of return
to sampled area given that the bird is alive). Recoveries of dead birds
typically occur over the entire range of the studied population, however, so
that survival rates estimated from these data using band recovery models
(e.g., Brownie et al., 1985)
typically estimate true survival (complement is death). By combining these
types of data, fidelity can be estimated. Despite the potential utility of
this method for estimating fidelity, it has seen little use (Frederiksen and
Bregnballe,
2000a,b;
Szymczak and Rexstad, 1991),
although some researchers have used the intuition underlying this approach in
a less formal manner (Anderson and
Sterling, 1974; Blums et al.,
1996; Hepp et al.,
1987).
We used Burnham's model
(1993) as available in Program
MARK (White and Burnham, 1999)
to estimate the probabilities of recapture, recovery, survival, and fidelity
for two populations of mid-continental mallards (Anas platyrhynchos).
For each of two populations, we tested the following predictions about
philopatry in the mallard: (1) females are more philopatric than males, (2)
after-hatch year birds are more philopatric than hatch year birds, and (3)
philopatry is positively correlated with annual estimates of pond abundance.
The latter prediction is conditional on the existence of positive
autocorrelation in pond numbers for successive years, t and
t + 1, a hypothesis that was also tested.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Data collection
In the 1950s the U.S. Fish and Wildlife Service and the Canadian Wildlife Service organized a cooperative banding program in which ducks are banded on the breeding grounds before the fall migration (Anderson and Henny, 1972
).
This banding is often referred to as "preseason banding" or
"prehunting season banding." Banding crews are sent to reference
areas (Anderson and Henny,
1972) across the breeding ranges. Banding usually takes place
during July-September (most bandings in this study occurred during August),
and both young, fledged birds and adults are banded by capturing birds in bait
traps. Banding takes place in the same general areas every year, but can
change locally based on water conditions and accessibility. Mallard hunting
seasons take place throughout North America generally for some time period
between September and February. Throughout most of the history of the
waterfowl banding program, few live recapture data have been recorded, and the
majority of encounter records in the U.S. Bird Banding Laboratory represent
band recoveries of birds shot and found dead during the hunting season.
Since 1987, banding teams at some localities on the breeding grounds have
made an effort to record recapture data as well as banding data. We were able
to use banding and recapture data from southern Alberta and southern
Saskatchewan. Specifically, we used data from 39 sites within the area bounded
by latitude 49°40'00'' and 53°30'00'' N and
longitude 110°10'00'' and 113°00'00'' W, which
we termed Alberta, and from 89 sites bounded by latitude
49°40'00'' and 53°30'00'' N and longitude
102°00'00'' and 109°40'00'' W, which we denoted
as Saskatchewan. Not all sites were banding locations in all years, due to
habitat conditions, available personnel, and logistics. We pooled sites within
provinces to increase our sample sizes as well as to loosely maintain the
reference area framework established for mallard banding data by Anderson and
Henny (1972).
During this time period, reward band studies with rewards of different
dollar values were conducted on mallards in these areas to estimate the
probability that a banded mallard recovered by a hunter is reported to the
Bird Banding Laboratory of either the U.S. or Canada (Nichols et al.,
1991,
1995). We excluded any birds
with reward bands from our analysis because these reward bands of different
values were reported with different probabilities than standard bands
(Nichols et al., 1991), and we
lacked sufficient data to include this factor in our analyses. We also
excluded anomalous birds, such as those that were injured during banding
operations.
We used banding and live recapture data from ducks banded in Alberta and Saskatchewan, as well as hunting season recovery data from these birds during 1987-1999. There were 109,755 and 187,480 mallards banded in Alberta and Saskatchewan, respectively, during this period, with 2083 and 3160 live recaptures and 14,104 and 29,191 dead recoveries. These data are available from the U.S. Bird Banding Laboratory, Laurel, Maryland.
Analytical approach
We used Burnham's model
(1993), as implemented in
program MARK (White and Burnham,
1999) to estimate parameters for mallards banded in Alberta and
Saskatchewan reference areas by age and sex for the years 1987-1999. We only
considered two age classes; juvenile or hatch-year birds (HY) and adult or
after-hatch-year birds (AHY). The parameters are defined as follows:
- pi = probability that a bird present in the
banding reference area at the time of banding in year i is recaptured
at that time (capture probability);
- ri = probability that a bird that dies during
year i does so during the hunting season and is retrieved and its
band reported to the Bird Banding Laboratory (reporting rate);
- Si = probability that a bird alive at the time
of banding in year i is alive at the time of banding in year
i + 1 (survival rate); and
- Fi = probability that a bird present in the
banding reference area at the time of banding in year i is also
present in the banding reference area at the time of banding in year
i + 1, given that it is alive at i + 1 (fidelity). The
complement of fidelity (i.e., 1 — Fi)
reflects permanent emigration but does not include temporary emigration
(failure to return to a specific breeding area in one year followed by return
in a subsequent year).
Note that the parameterization described above uses the reporting rate
parameter of Seber (1970),
rather than the recovery rate parameter of Brownie et al.
(1985), as Seber's
parameterization has been implemented in MARK
(White and Burnham, 1999).
Note also that the reporting rate parameter, ri,
includes the proportion of annual mortality that occurs during the hunting
season and is not equivalent to the conditional band reporting rate,
i, of Nichols et al.
(1991,
1995). This difference in
parameterization should be kept in mind when comparing to reporting rates from
other studies (see Discussion).
Due to the size of the data set and limitations in computer power, we modeled each reference area (Alberta and Saskatchewan) separately rather than constructing a joint model for the two data sets combined. For each area we first constructed our most general model, a model in which reporting rate, survival, and fidelity were a function of sex (male or female), age (HY or AHY) and time (1987-1999), and the probability of recapture was a function of sex and year (birds banded as HY were AHY by the time of first possible recapture, so there was no need for age-specific capture probability parameters).
In addition, we modeled reporting rate as a function of band type in order
to properly estimate rates of survival and fidelity. During the time period of
data collection for our study, the Bird Banding Lab used two different band
types. Before 1995, the Bird Banding Lab used "address" bands with
the words "AVISE, WRITE WASHINGTON DC" or "WRITE BIRD BAND
LAB, LAUREL MD, USA" stamped on the band. During 1995 a new band type
was instituted with a toll-free telephone number stamped on each band as well.
Since 1995 the address bands have been phased out as stocks of these bands
have been depleted, and essentially the duck-banding program now operates
solely with the toll-free telephone number bands. Because we suspected these
band types affected reporting rates, we included this factor in our modeling
strategy. Our most general model also included all interaction terms in this
model, and we designated it as
S(sex*age*t)
p(sex*t)
r(btype*sex*age*t)
F(sex*age*t), where the
t denotes time and * denotes an interaction between
factors (see Lebreton et al.,
1992)
We performed a bootstrap goodness-of-fit test as available in Program MARK
(White and Burnham, 1999) on
this general model. The bootstrap goodness-of-fit test can detect
overdispersion in the data and, if warranted, a variation inflation factor,
, can be estimated and used to adjust subsequent estimates and
model selection statistics. We took a conservative approach and used
if the bootstrap goodness-of-fit test was significant at
p <.20. The values were calculated by dividing
the deviance of the general model by the average deviance from the results of
the bootstrap simulations.
After the bootstrap procedure, we constructed models in the following way.
For r, we consecutively dropped out interaction terms and, if
possible, main effects using Akaike's Information Criteria (AIC;
Burnham and Anderson, 1998)
for model selection. Specifically, we used the small-sample size and
adjusted AIC (QAICc). We consecutively followed
the same procedure for p, S, and finally for the parameter we were
most interested in, F.
To test the prediction that mallards have higher fidelity rates in years
when wetland conditions are good (Johnson
and Grier, 1988), we additionally modeled fidelity as a function
of pond abundance. Specifically, we used the estimated number of May ponds in
Prairie Canada (U.S. Fish and Wildlife
Service, 2000) as a covariate in modeling fidelity. We assume
there is high spatial covariation among the prairie breeding areas, and, thus,
this estimate would be a good index to wetland conditions across the breeding
range, and we tested for temporal autocorrelation in pond number by using the
Box-Ljung statistic (StatSoft,
1995) with a lag of one year. In these models, other parameters
(rates of survival, recapture, and reporting) were modeled as in the
best-supported model from the previously described efforts. We constructed a
model in which fidelity in year i (Fi)
was a function of pond abundance in year i + 1, as well as a model in
which fidelity was a function of pond abundance in year i. Because
Fi represents the probability that a bird present
in the banding reference area at the time of banding in year i is
also present in the banding reference area at the time of banding in year
i + 1, given that it is alive at i + 1, the first model
focuses on same-year wetland conditions and is based on active habitat
selection using current-year cues (e.g.,
Johnson and Grier, 1988). The
second model is based on past experience of ducks and relies on temporal
covariation such that pond conditions in one year provide information about
future pond conditions (Danchin and
Wagner, 1997, Danchin et al.,
1998).
Because of the large sample sizes of the data sets, we were able to support
models with many parameters. After deciding on the most appropriate models
using AIC model selection criteria (models with 
QAICc > 2
were considered not to fit the data as well;
Burnham and Anderson, 1998),
we examined parameter estimates. When examining year-specific fidelity
estimates in the absence of covariate modeling, it was clear that we could
only marginally estimate fidelity in many years. This problem in estimation
constrained our modeling efforts of fidelity to not include annual
variability.
In addition, we estimated effect sizes for the parameter in which we were
most interested, fidelity, as follows. We estimated effect size for AHY and HY
birds as the ratio of male: female fidelity rates (denoted as 
), with
associated variance:
 |
).
Approximate 95% confidence intervals for these effect size estimates were
constructed as:
 |
If males and females are equally philopatric, then the expected ratio of
male:female fidelity is 1, E() = 1 and the 95% confidence
interval around the estimate should include 1. However, if females are more
philopatric than males, as predicted, then the ratio will be <1, and the
entire 95% confidence interval should be located below 1. All estimates are
presented as estimate (SE) unless otherwise noted.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Model selection
Bootstrap goodness-of-fit test statistics indicated a lack of fit of our most general models, suggesting that our data were overdispersed. Estimated variance inflation factor (
) values from these analyses were
2.56 and 2.39 for Alberta and Saskatchewan, respectively, and we incorporated
these values in our model selection procedures. With the sample sizes
available, we were able to support models with large numbers of parameters.
For both data sets, the best fitting model structured the probability of
recapture, p, as a function of time only, while the reporting rate,
r, was a function of band type, sex, age, and time. However,
interactions between these variables were not supported during the modeling
process, and thus we could focus on the main effects associated with each
variable.
Survival (S) was modeled as a function of sex, age, and time, and
all interactions among these variables. In Alberta the most parsimonious model
coded fidelity (F) as a function of sex and age without an
interaction between these terms. However, a model with an interaction between
sex and age in fidelity ranked similarly according to QAICc. We
chose to use the model without the interaction term. The interaction term
between age and sex was important in modeling fidelity for Saskatchewan. These
models were designated
S(sex*age*t)
p(t)
r(btype+sex+age+t) F(sex+age)
for Alberta and
S(sex*age*t)
p(t)
r(btype+sex+age+t)
F(sex*age) for Saskatchewan. These models were
by far the most supported, with all other models having
QAICc > 20 for both Alberta and Saskatchewan (except for
the aforementioned Alberta model with an age*sex interaction in
fidelity), or having many parameters that could not be estimated. Parameter
estimates based on these best fitting models are discussed below.
Probability of recapture
There was no evidence for sex differences in p. However,
p showed temporal variation, even though the
i were low
overall. The average (obtained from
a reduced model in which temporal variation was not included) for Alberta
(0.0104 [SÊ = 0.0007]) was higher than that for Saskatchewan
(0.0055 [SÊ = 0.0002]).
Reporting probability
Reporting probabilities, ri, varied by age,
sex, time, and band type. We were not able to eliminate any of these
variables. Any model with at least one of these variables being absent had
very high QAICc rankings (
25). Because all interaction
terms were eliminated in the most parsimonious model, we were able to focus
more closely on the main effects. Beta values associated with the main effect
of sex (-1.12 [SÊ = 0.07], -1.36 [SÊ = 0.04])
indicated that females had lower reporting rates in both Alberta and
Saskatchewan, respectively. Similarly, adult birds (-0.64 [SÊ =
0.19], -0.77 [SÊ = 0.07]) showed lower reporting rates than
juvenile birds and address bands (-0.30 [SÊ = 0.09], -0.15
[SÊ = 0.03]) had lower reporting rates than toll-free bands.
Arithmetic mean reporting probabilities for ducks with address bands in
Alberta were greater for HY males (0.24 [SÊ = 0.03]) and AHY
males (0.14 [SÊ = 0.02]) than for HY females (0.09
[SÊ = 0.01]) or AHY females (0.05 [SÊ = 0.02];
Figure 1). These same
qualitative differences were evident for ducks with toll-free bands and for
the Saskatchewan area. The differences among the age—sex classes in this
reporting parameter are a likely result of the timing of mortality during the
year (see Discussion).
|
In 1995, when the toll-free band type was first used, the reporting rates increased for this band type (Figure 2b). In both Alberta and Saskatchewan, over the 5 years the toll-free band type has been used, the reporting rate, r, for these bands has been approximately 1.3 times that of address bands. The advertising of the toll-free band and the existence of an easy way to report bands probably resulted in increases in the reporting rates for all band types. Overall reporting rates from 1995 to 1999 have increased approximately 1.5 times for address bands as compared to the previous 8 years (Figure 2).
|
Probability of survival
In Alberta the probability of survival differed among age and sex classes.
Models lacking at least one of these variables differed from models containing
both variables (QAICc > 30). Estimates from
reduced-parameter models with no time effects indicated that adult males (0.75
[SÊ = 0.01]) survived better than adult females (0.64
[SÊ = 0.02]) and juvenile males (0.84 [SÊ =
0.03]) survived better than juvenile females (0.71 [SÊ = 0.05];
Figure 2). In Saskatchewan,
adult male (0.75 [SÊ = 0.01]), adult female (0.62
[SÊ = 0.01]), juvenile male (0.81 [SÊ = 0.01]),
and juvenile female (0.67 [SÊ = 0.02]) survival rate estimates
were similar to those for ducks nesting in Alberta.
Fidelity
In examining parameter estimates from models incorporating temporal
variation, it was clear that we could only marginally estimate F for
many years, probably because of our low p values, and year-specific
parameters were not needed to model the process that gave rise to the data.
Pond number was postively autocorrelated through time (ri
= 0.53 [0.29], Q = 4.26, p =.03). This evidence of a
positive correlation between pond numbers in successive years i and
i + 1 led us to hypothesize a positive relationship between pond
numbers in year i and the probability of a bird returning the next
year (Fi).
When we modeled fidelity in year i as a function of pond abundance
in year i + 1 or i, these models also performed poorly, with
virtually no support for these models. However, sex and age differences in
fidelity were evident. Models not containing at least one of these parameters
had QAICc values > 20, a strong indication that both age
and sex differences exist in fidelity rates. In Alberta, surviving adult
females returned to breeding sites with an estimated probability of 0.85
(SÊ = 0.05) and adult males returned with an estimated
probability of 0.76 (SÊ = 0.02). Hatch-year females returned at
an estimated rate of 0.40 (SÊ = 0.08) and HY males at an
estimated rate of 0.24 (SÊ = 0.04;
Figure 3). In Saskatchewan, our
estimates of breeding-site fidelity for adult females were 0.85
(SÊ = 0.03) and 0.87 (SÊ = 0.01) for adult
males. Yearling females homed at an estimated rate of 0.99 and yearling males
at an estimated rate of 0.37 (SÊ = 0.04). Because the fidelity
estimate for juvenile females was near the boundary of 1, we could not
estimate the variance for this parameter reliably.
|
We examined the magnitude of the effect size, which is the estimated
proportionate difference between fidelity rates of the sexes
(Figure 4). Philopatry of adult
females was slightly greater than that of adult males in Alberta as indicated
by the ratio of male : female fidelity being < 1 and by the 95% confidence
interval around this ratio not including one. However, this ratio was close to
unity, indicating that adult males also home at a high rate, almost equal to
that of females. No such sex differences could be detected between adults in
Saskatchewan. In juvenile birds, females were much more philopatric than males
as indicated by the ratio being much less than one for both Alberta and
Saskatchewan in Figure 4.
Because we could not estimate the variance reliably for juvenile females in
Saskatchewan, the 95% confidence interval around this ratio should be viewed
as a minimal confidence interval. After viewing these results we constructed a
model in which fidelity was a function of sex in juveniles, but not in adults.
This model was slightly better (more appropriate than the two-sex model) for
the Saskatchewan data (QAICc = 1.42), but not for the
Alberta data (QAICc = 4.15).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This is one of a few studies reporting estimates of fidelity that are not confounded by detection and/or survival probabilities. Although our primary interest was in fidelity, F, our models included three other sets of parameters (p, r, and S), and we begin the discussion with a brief consideration of our estimates of these quantities.
The probabilities of recapture (pi) were extremely low (frequently <.01), and this hindered us in estimating fidelity. When we examined the parameter estimates more closely, we were not able to detect differences between the sexes or areas in pi. Our large sample sizes were fortunate, but an increase in pi would greatly improve our ability to estimate F. We suspect that the low recapture probabilities were largely a consequence of the great mobility of mallards and the large size of the study areas.
The reporting rate parameters (ri) in our models are
not comparable to the band reporting rate parameters
(i) estimated in reward band studies (Nichols et
al., 1991,
1995). The latter parameter
reflects the probability that a hunter will report a bird that has been killed
and retrieved. The reporting rate parameter of Seber
(1970), as used in our
modeling, reflects the probability that a banded bird that dies of any cause
at any time during the year i to i + 1 dies during the
hunting season and is reported to the Bird Banding Laboratory. The larger
reporting rate estimates for males than females likely reflect that the bulk
of male mortality occurred during the hunting season and that a substantial
portion of female mortality occurred during the breeding season (e.g.,
Blohm et al., 1987). Similarly,
the higher reporting rates of young than adult birds within each sex likely
reflect the greater proportion of annual mortality occurring during the
hunting season for young birds (e.g.,
Anderson, 1975).
Band type was an important factor associated with variation in reporting
rates, and we believe that this factor influenced the probability that an
encountered band was reported (e.g., rather than the timing of mortality). The
most recent change to a band with a toll-free number has increased reporting
rates, not only for the toll-free number bands, but also for the address
bands, most probably through the widespread advertisement of the toll-free
number (Figure 2). Bands with
the toll-free phone number showed reporting rate estimates,
i, averaging 1.3 times
those of address bands over the 5-year period of use.
The probability of survival was higher for males than females, with less of
a difference between age classes (Figure
2). This pattern is similar to that found by others (e.g.,
Anderson, 1975;
Chu and Hestbeck, 1989;
Smith and Reynolds, 1991;
Trost, 1987). Our estimates of
mallard survival (e.g., 0.76 for AHY males in Alberta) are generally larger
than previous estimates that have been published for the same area (0.66-0.69;
Chu and Hestbeck, 1989;
Smith and Reynolds, 1991) and
for North America as a whole (Anderson,
1975; Trost,
1987). The estimates from our study cover a more recent time
period, and it may be that mallard survival has increased in recent years.
Our estimates of fidelity for all age—sex classes were high relative
to estimates commonly reported (e.g.,
Lokemoen et al., 1990; see
below). Our analyses supported our prediction that yearling females are more
philopatric than yearling males. The difference between the sexes was not
nearly as large in adults as in HY birds, and there was only marginal support
for the prediction of sex differences in philopatry in AHY birds (Figures
3 and
4). Our prediction that AHY
birds are more philopatric than HY birds was supported.
Although much has been written about philopatry and pair formation in
waterfowl (Rohwer and Anderson,
1988), commonly used estimation methods have not led to strong
inferences. Published ad-hoc breeding-ground fidelity estimates for adult
mallard females range from 0.13 to 0.58
(Anderson et al., 1992;
Coulter and Miller, 1968;
Doty and Lee, 1974;
Johnson et al., 1992;
Lokemoen et al., 1990;
Sowls, 1955), with estimates
for males being lower (e.g., 0.03; Titman,
1983). Rates reported for HY female mallards (0.05-0.56) are
generally lower than those reported for adult females, but higher than
reported for HY males (0.01; Sowls,
1955).
Comparisons of our results with ad hoc estimates from previous studies are
problematic for two reasons. First, there is an issue of scale. We calculated
our estimates for collections of sites within larger reference areas, and our
estimates of fidelity pertain to the probability of returning to one of these
sites within the large reference areas. Defining philopatry in terms of a
larger area will increase fidelity rates. In actuality, 93 of 2083 Alberta and
104 of 3160 Saskatchewan recaptured ducks were in a 10'-degree block
(the finest resolution at which data are recorded in the Bird Banding Lab)
other than the one in which they were originally banded. Although this is not
a large proportion of our recapture data, these recaptures influence our
fidelity rates nonetheless. Other studies have typically involved smaller
areas over which birds were banded and recaptured. Robertson and Cooke
(1999) suggest that
standardized data collection and a study design that would permit comparison
of philopatry rates within a population to study areas of differing scale (1
km2, 10 km2,... 105 km2) would be
valuable. Although our small capture probabilities did not allow us to focus
at a smaller scale, with some adjustments in effort this may be possible.
The second reason for the discrepancies between our fidelity estimates and previous estimates is that most previous work has relied on ad-hoc estimators that represent composite probabilities of survival, recapture, and fidelity. The incorporation of survival and recapture probabilities into our estimation procedure is expected to lead to higher estimates of fidelity than methods that confound fidelity with these quantities. Our estimates for adults (> 0.75) are much higher than previously reported and suggest that mallards exhibit high rates of breeding ground fidelity, at least relative to the scale of our study (Figures 3 and 4).
It is interesting that the few other studies that were not confounded with
survival and recapture probabilities have also reported high estimates of
fidelity. Hepp et al. (1987)
and Blums et al. (1996)
reported fidelity estimates approaching 1 for female wood ducks (Aix
sponsa) in a South Carolina study area and female pochards (Aythya
ferina) and tufted ducks (Aythya fuligula) in Engure Marsh,
Latvia, respectively. Szymczak and Rexstad
(1991) found that models with
F = 1 for gadwalls of both sexes performed well, resulting in the
inference of high fidelity rates for both sexes to a Colorado postbreeding
molting area. Anderson and Stirling (1974) estimated fidelity of male pintails
(Anas acuta) to a molting marsh in Saskatchewan to be about 0.8.
Frederiksen and Bregnballe
(2000a,b)
reported fidelity rates of 0.87 for cormorants (Phalocrocorax carbo)
to a 60-ha island in Denmark, with little evidence for any differences between
age classes. All of the above studies reported high fidelity despite being
conducted at smaller geographic scales than ours.
Our analyses supported predictions from current theory concerning
sex-specific fidelity rates in HY birds, but there was only marginal support
for this prediction in AHY birds. The fidelity rates of adult males were only
slightly less than those of adult females if there was a difference at all
(Figures 3 and
4). We believe that the
inference of sex-biased homing rates in dabbling ducks, with greater fidelity
to breeding grounds of adult females than males, should be revisited in light
of (1) our estimates, (2) the above studies in which fidelity estimates are
not confounded by survival and recapture, and (3) the recent review of
philopatry by Clarke et al.
(1997). Clarke et al.
(1997) found some degree of
male-biased dispersal in 15 species outside the Anatidae, representing many
breeding systems, suggesting that philopatry cannot simply be classified
within a resource-defense versus mate-defense framework and that sex-biased
philopatry is probably not a species constant
(Clarke et al., 1997;
Waser and Jones, 1983).
One possible mechanism underlying our evidence of similar fidelity between
the sexes for adult birds is suggested by observations of Dwyer et al.
(1973) and Blohm and Mackenzie
(1994) that mallard pairs
returned together to a breeding site in multiple years. It may be that
long-term pair bonds in mallards are more common and stronger than previously
thought and that this may be beneficial to the relative fitness of these
individuals. Of course, long-term pair bonds are typical of swans and geese,
with pairs and families commonly wintering together
(Previtt and MacInnes, 1980).
In the absence of long-term pair bonds, a second possibility is that males, as
well as females, may gain from breeding-ground homing tendencies, possibly in
terms of feeding efficiency or in territorial interactions, as suggested by
Titman's (1983) work and his
report of a male that homed with different mates.
A third intriguing possibility is that fidelity rates to wintering areas
are high in both female and male mallards, as suggested by ad hoc estimates
recently reviewed by Robertson and Cooke
(1999), and that these
contribute to high breeding-ground fidelity of both sexes. If females home to
wintering grounds at a high rate, and males do also, males may follow females
to breeding grounds, giving rise to high breeding-season fidelity rates, but
operating through wintering-site fidelity of males and fidelity during both
breeding season and winter by females. Such a mechanism would be especially
likely in situations such as ours, in which fidelity is defined in terms of a
relatively large geographic area (e.g., as opposed to a specific breeding
site). Adequate tests of this third hypothesis will require estimates of
fidelity to wintering grounds that are based on sound statistical methods such
as the approach of Burnham
(1993).
Other hypotheses concerning philopatry can also be addressed with
extensions to the method used here. Fairly good evidence has been provided
that successful birds are more likely to return in following years than
unsuccessful ones (Lokemoen et al.,
1990; Majewski and Beszterda,
1990). A combination of Burnham's
(1993) model and a multistate
modeling approach (e.g., Brownie et al.,
1993; Nichols,
1996) could be used to formally test this idea. Birds encountered
in year i would be classified as successful or unsuccessful in their
breeding attempts, and models could be developed to formally test for
differences in fidelity rates for birds in the two breeding success states.
This approach would avoid the pitfalls associated with confounded
estimates.
Although it is clear that wetland conditions influence duck reproduction
(e.g., Miller and Newton,
1999; Pospahala et al.,
1974), we found no support for our third prediction that fidelity
would covary with pond availability. When there are drought conditions, ducks
are thought to fly farther north in search of appropriate habitat
(Johnson and Grier, 1988). It
may be that our reference areas were sufficiently large that such a response
may have occurred within our reference areas. Another plausible reason for our
result is simply that we had inadequate data (because of our low recapture
probabilities) to permit strong inferences about year-to-year variation in
rates of fidelity. Finally, we note that there are multiple ways to define
fidelity. The complement of the parameter that we have estimated reflects
permanent emigration (commitment to another breeding area resulting in
probability of returning to banding area approaching 0). Responses to changing
wetland conditions such as those discussed by Johnson and Grier
(1988) may reflect temporary
emigration in which birds fly over past breeding areas in years of poor
wetland conditions but return in subsequent years of good conditions. Such
behavior would not be reflected in our fidelity estimates based on the
approach of Burnham (1993).
Temporary emigration can be estimated using capture—recapture methods,
but such estimation requires the robust design of Pollock
(1982; see
Kendall et al., 1997).
In summary, using modern analytical techniques to obtain unbiased estimates of fidelity rates, we supported predictions of age differences in mallard fidelity rates to breeding sites and sex differences in juvenile birds. However, we also found that adult males home at a higher rate than previously thought, almost as high or equal to that of females in some areas. Our estimates of philopatry are higher than those previously published for mallards and agree with more recent studies that have corrected for probabilities of survival and detection. We recommend the further use of analytical techniques such as those we used here to estimate fidelity rates and to further test and revisit hypotheses concerning philopatry.
|
ACKNOWLEDGEMENTS |
|---|
We thank Carl F. Ferguson and Elwood Martin for data collection and especially Sarah Fromme for data entry. The U.S.G.S. Patuxent Wildlife Research Center and the U.S. Fish and Wildlife Service provided funding for this project.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Anderson DR, 1975. Population ecology of the mallard. V. Temporal and geographic estimates of survival, recovery and harvest rates. U.S. Fish and Wildlife Service Resource Publication 125. Washington, DC: U.S. Department of the Interior.
Anderson DR, Henny CJ, 1972. Population ecology of the mallard I. A review of previous studies and the distribution and migration from breeding areas. U.S. Fish and Wildlife Service Resources Publication 105. Washington, DC: U.S. Department of the Interior.
Anderson DR, Sterling RT, 1974. Population dynamics of molting pintails drakes banded in south-central Saskatchewan. J Wildl Manag 38: 266-274.
Anderson MG, Rhymer JM, Rohwer FC, 1992. Philopatry, dispersal and the genetic structure of waterfowl populations. In: Breeding ecology and management of waterfowl (Batt DDJ, Afton AD, Ankney CD, Johnson DH, Kadlek JA, Krapu GL, eds). Minneapolis: University of Minnesota Press; 365-395.
Blohm R, Mackenzie KA, 1994. Additional evidence of migrational homing by a pair of mallards. J Field Ornithol 65: 476-478.
Blohm R, Reynolds RE, Bladen JP, Nichols JD, Hines JE, Pollock KH, Eberhardt RT, 1987. Mallard mortality rates on key breeding and wintering areas. Trans N Am Wildl Nat Resour Conf 52: 246-257.
Blums P, Mednis A, Bauga I, Nichols JD, Hines JE, 1996. Age-specific survival and philopatry in three species of European ducks: a long-term study. Condor 98: 61-74.
Brownie C, Anderson DR, Burnham KP, Robson DS, 1985. Statistical inference from band-recovery data—a handbook, 2nd ed. Resource publication 156, U.S. Fish and Wildlife Service. Washington, DC: U.S. Department of the Interior.
Brownie C, Hines JE, Nichols JD, Pollock KH, Hestbeck JB, 1993. Capture-recapture studies for multiple strata including non-Markovian transition probabilities. Biometrics 49: 1173-1187.
Burnham KP, 1993. A theory for the combined analysis of ringed recovery and recapture data in marked individuals. In: The study of bird populations (Lebreton J-D, North PM, eds). Basel: Birkhauser; 199-213.
Burnham KP, Anderson DR, 1998. Model selection and inference—a practical information-theoretic approach. New York: Springer.
Chu DS, Hestbeck JB, 1989. Temporal and geographic estimates of survival and recovery rates for the mallard, 1950 through 1985. Fish and Wildlife Service Technical Report 20. Washington, DC: U.S. Department of the Interior.
Clarke AL, Saether B-E, Røskaft E, 1997. Sex biases in avian dispersal: a reappraisal. Oikos 79: 429-438.
Cormack RM, 1964. Estimates of survival from sighting
of marked animals. Biometrika 51:
429-438.
Coulter MW, Miller WR, 1968. Nesting biology of black ducks and mallards in northern New England. Vermont. Fish and Game Department Bulletin 68-2. Burlington: Vermont Fish and Game Department.
Danchin E, Boulinier T, Massot M, 1998. Conspecific reproductive success and breeding habitat selection: implications for the study of coloniality. Ecology 79: 2415-2428.[Web of Science]
Danchin E, Wagner RH, 1997. The evolution of coloniality: the emergence of new perspectives. Trends Ecol Evol 12: 342-347.
Dobson SF, 1982. Competition for mates and predominate juvenile male dispersal in mammals. Anim Behav 30: 1183-1192.
Doty HA, Lee FB, 1974. Homing to nest baskets by wild female mallards. J Wildl Manag 38: 714-719.
Dwyer TJ, Derrickson SR, Gilmer DS, 1973. Migrational homing by a pair of mallards. Auk 90: 687.
Frederiksen M, Bregnballe T, 2000a. Diagnosing a decline in return rate of 1-year-old cormorants: mortality, emigration or delayed return? J Anim Ecol 69: 753-761.
Frederiksen M, Bregnballe T, 2000b. Evidence for density-dependent survival in adult cormorants from a combined analysis of recoveries and resightings. J Anim Ecol 69: 737-752.
Greenwood PJ, 1980. Mating systems, philopatry and dispersal in birds and mammals. Anim Behav 28: 1140-1162.
Greenwood PJ, Harvey PH, 1982. The natal and breeding dispersal of birds. Annu Rev Ecol Syst 13: 1-21.[Web of Science]
Hepp GR, Hoppe RT, Kennamer RA, 1987. Population parameters and philopatry of breeding female wood ducks. J Wildl Manag 51: 401-404.
Johnson DH, Grier JW, 1988. Determinants of breeding distributions of ducks. Wildl Monogr 100: 1-37.
Johnson DH, Nichols JD, Schwartz MD, 1992. Population dynamics of breeding waterfowl. In: Breeding ecology and management of waterfowl (Batt DDJ, Afton AD, Ankney CD, Johnson DH, Kadlek JA, Krapu GL, eds). Minneapolis: University of Minnesota Press; 446-485.
Johnson ML, Gaines MS, 1990. Evolution of dispersal: theoretical models and empirical tests using birds and mammals. Annu Rev Ecol Syst 21: 448-480.
Jolly GM, 1965. Explicit estimates from
capture-recapture data with both death and immigration—stochastic model.
Biometrika 52:
225-247.
Kendall WL, Nichols JD, Hines JE, 1997. Estimating temporary emigration using capture-recapture data with Pollock's robust design. Ecology 78: 563-578.
Lebreton J-D, Burnham KP, Clobert J, Anderson DR, 1992. Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecol Monogr 62: 67-118.
Lokemoen JT, Duebbert HF, Sharp DE, 1990. Homing and reproductive habits of mallards, gadwalls, and blue-winged teal. Wildl Monogr 106: 1-28.
Majewski P, Beszterda P, 1990. Influence of nesting success on female homing in the mallard. J Wildl Manag 54: 459-462.
Miller MR, Newton WE, 1999. Population energetics of northern pintails wintering in the Sacramento Valley, California. J Wildl Manag 63: 1222-1238.
Mood AM, Graybill FA, Boes DC, 1974. Introduction to the theory of statistics. New York: McGraw-Hill.
Nichols JD, 1996. Source of variation in migratory movement of animal populations: statistical inference and a selective review of empirical results for birds. In: Population dynamics in ecological space and time (Rhodes OE Jr, Chesser RK, Smith MH, eds). Chicago: University of Chicago Press; 147-197.
Nichols JD, Blohm RJ, Reynolds RE, Trost RE, Hines JE, Bladen JP, 1991. Band reporting rates for mallards with reward bands of different dollar values. J Wildl Manag 55: 119-126.
Nichols JD, Reynolds RE, Blohm RJ, Trost RE, Hines JE, Bladen JP, 1995. Geographic variation in band reporting rates for mallards based on reward banding. J Wildl Manag 59: 697-708.
Pollock KH, 1982. A capture-recapture design robust to unequal probability of capture. J Wildl Manag 46: 757-760.
Pospahala RS, Anderson DR, Henny CJ, 1974. Population ecology of the mallard: II. Breeding habitat conditions, size of the breeding populations, and production indices. U.S. Fish and Wildlife Resource Publication 115. Washington, DC: U.S. Department of the Interior.
Previtt JP, MacInnes CD. 1980. Family and other social groups in snow geese. Wildl Monogr 71: 1-46.
Robertson GJ, Cooke F, 1999. Winter philopatry in migratory waterfowl. Auk 116: 20-34.
Rohwer FC, Anderson MG, 1988. Female-biased philopatry, monogamy, and the timing of pair formation in migratory waterfowl. Curr Ornithol 5: 187-221.
Sargeant AB, Raveling DG, 1992. Mortality during the breeding season. In: Breeding ecology and management of waterfowl (Batt DDJ, Afton AD, Ankney CD, Johnson DH, Kadlek JA, Krapu GL, eds). Minneapolis: University of Minnesota PRess; 396-422.
Seber GAF, 1965. A note on the multiple-recapture
census. Biometrika 52:
249-259.
Seber GAF, 1970. Estimating time-specific survival and
reporting rates for adult birds from band returns. Biometrika
57: 313-318.
Smith GW, Reynolds RE, 1991. Hunting and mallard survival, 1979-88. J Wildl Manag 56: 306-316.
Sowls LK, 1955. Prairie ducks: a study of their behavior, ecology and management. Lincoln: University of Nebraska Press.
StatSoft, 1995. Statistica version 5.1. Tulsa, Oklahoma: StatSoft.
Szymczak MR, Rexstad EA, 1991. Harvest distribution and survival of a gadwall population. J Wildl Manag 55: 592-600.
Titman RD, 1983. Spacing and three-bird flights of mallards breeding in pothole habitat. Can J Zool 61: 839-847.
Trost RE, 1987. Mallard survival and harvest rates: a reexamination of relationships. Trans N Am Wildl Nat Res Conf 52: 264-297.
U.S. Fish and Wildlife Service, 2000. Waterfowl population status, 2000. Washington, DC: U.S. Department of the Interior.
Waser PM, Jones WT, 1983. Natal philopatry among solitary mammals. Q Rev Biol 58: 623-643.
White GC, Burnham KP, 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46: S120-S139.
Wolff JO, 1992. Parents suppress reproduction and stimulate dispersal in opposite-sex juvenile white-footed mice. Nature 359: 409-410.[Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||