WILDLIFE SCIENCE: LINKING ECOLOGICAL THEORY AND MANAGEMENT APPLICATIONS - CHAPTER 12 pot

Teer CONTENTS Testing South Texas Deer Counts for Density Dependence.. They presented data from three mule deer and two white-tailed deer populations in Montana, United States, and concl

12 Density Dependence in

Deer Populations:

Relevance for

Management in Variable

Environments

Charles A DeYoung, D Lynn Drawe,

Timothy Edward Fulbright, David Glenn Hewitt,

Stuart W Stedman, David R Synatzske, and James G Teer

CONTENTS

Testing South Texas Deer Counts for Density Dependence 206

Foundation Theory and Management Relevance 215

Acknowledgments 220

References 220

New-world deer of the genus Odocoileus are commonly assumed to respond to food shortage

due to intraspecific competition by reduced recruitment, body mass, and other manifestations McCullough’s (1979) book on the George Reserve deer herd is commonly cited as the

definit-ive work on density dependence in white-tailed deer (O virginianus), and by extension, mule deer (O hemionus) The George Reserve is a 464-ha, high-fenced property in Michigan, United States.

Two males and four females were introduced in 1928, and by 1933, the population was estimated

to be 160 The population fluctuated until 1952, when a series of experiments with deer density began The population was reduced during this time in a series of steps and data collected to form a population model (McCullough 1979) Strong density dependence was evident in this model Reductions in the George Reserve population continued into 1975, when an estimated 10 deer remained (McCullough 1982) No deer were subsequently harvested for 5 years to allow the popu-lation to increase McCullough (1982, 1983) concluded that the popupopu-lation response after 1975 was similar to the original increase from 1928 to 1933

Downing and Guynn (1985) presented a generalized sustained yield table using McCullough (1979) as a starting point Downing and Guynn (1985) used their experience and literature values to

present a table scaled to percent K carrying capacity McCullough (1979, 1982, 1983) and Downing

and Guynn (1985) showed density-dependent responses across the population growth range from

low density to K We define K as the maximum sustainable population level where the deer are in

approximate equilibrium with their food supply (Macnab 1985)

Downing and Guynn (1985) recognized that their generalized model might not apply to all deer populations They wondered if their model would be applicable to low-density populations and areas

203

with poor habitat, which precluded high rates of recruitment They suggested ways their generalized model could be modified for populations that did not fit the mold of those where recruitment was consistently high and relatively stable across time

McCullough (1984) also recognized that if environmental variation is great, density-dependent effects, while present in the mix of factors impinging on a population, may be masked He suggested that these situations were rare and occurred at extreme fringes of whitetail range

Mackie et al (1990) questioned whether density-dependent models have utility for management They presented data from three mule deer and two white-tailed deer populations in Montana, United States, and concluded that there was evidence of density-dependent behavior in one mule deer and one white-tailed deer population They stated that western North America has a high degree

of environmental variation resulting in fluctuating carrying capacity They suggested that some Montana populations had declined, because expected density-dependent responses to harvest did not happen Finally, they suggested that in variable environments, managers should employ techniques providing regular tracking of population size and performance and not depend on predictions from density-dependent models

McCullough (1990) issued a strong caution to the conclusions of Mackie et al (1990) He stressed that density-dependent behavior may be missed because it is obscured by environmental factors and sampling error Also, experimental and statistical design frequently places the burden of proof on density dependence, that is, the null model is a lack of density dependence Finally, time lags, study area scale, environmental homogeneity, life history, behavior, and predation may make density dependence difficult to detect when in fact it is present Importantly, McCullough (1990) hypothesized that several of these factors, singularly or in combination, can result in a population of

deer expressing no density-dependent response until very near K.

Fryxell et al (1991) reported on a white-tailed deer population in southeastern Ontario, Canada, that fluctuated widely over 34 years They concluded that variation in hunting effort strongly affected the fluctuation and that the population showed time-lagged density dependence

McCullough (1992) again emphasized that environmental variation can obscure density-dependent responses in deer populations He stated that this may require study of a population over a large range of densities to detect a density-dependent response McCullough (1999) also reviewed and extended his concept of some populations of ungulates having a “plateau” of constant

growth and then a “ramp” of declining growth in the graph of r on N (Figure 12.1B, b, C, and c)

No density-dependent response would be observed in the plateau phase He hypothesized that this

model could fit more K-selected species with low reproductive rates However, he speculated that a plateau and ramp model may fit Odocoileus deer in desert environments.

Bartmann et al (1992) could not detect differences in fawn survival in response to experimental removal of 22 and 16% in consecutive years in a migratory Colorado, United States, mule deer population They subsequently simulated density-dependent fawn mortality in enclosures with a wide range of density Natural mortality of adult does was low in the free-ranging population but

fawn mortality was relatively high and varied with winter severity Relation of this population to K was unknown, but the authors assumed it to be near or at K.

Keyser et al (2005) studied long-term data sets for nine white-tailed deer populations in the southeastern United States They concluded that eight out of nine populations showed density-dependent responses, but that these responses frequently lagged 1 or 2 years They stated that the population that did not show density-dependent responses occurred on exceptionally poor habitat Shea et al (1992) collected data from a white-tailed deer population in Florida, United States, that declined 75% in density during a 10-year period They found little difference in deer physiological indices during this period and concluded that the habitat, which was characterized by low-fertility soils, produced low amounts of high-quality forage and an abundance of poor-quality forage Lack

of nutritious forage, coupled with abundant poor-quality forage precluded a density-dependent response, because there was little opportunity for intraspecific competition, even when densities were high Shea and Osborne (1995) discussed poor-quality habitat across North America They surveyed

N

A

K

Time

Time

cc

N

cc

N

cc

N

Time

r

N

B

K

r

N

C

(a)

(b)

(c)

K

Density dependence expressed

Density dependence expressed

Density dependence expressed

FIGURE 12.1 Plateau and ramp graphs (A, B, C) showing a range of deer population density-dependent

responses and corresponding graphs (a, b, c) of carrying capacity variation in comparison to population level

variation (Adapted from D R McCullough J Mammal 80:1132 and 1133:1999 With permission.)

state game departments and produced a map within whitetail range where density-dependent response would be lacking or masked

Dumont et al (2000) worked on white-tailed deer on the northern limit of their range in south-eastern Quebec, Canada They stated that severe winters were among the major factors limiting deer populations, but found density-dependent responses during mild winters

Gilbert and Raedeke (2004) worked on Columbian black-tailed deer (Odocoileus hemionus columbianus) and found that minimum temperatures in May and the amount of precipitation in June

affected fawn recruitment However, they also reported that plant production was correlated with deer density in the same year Also, their best models of fawn production included time-lagged dens-ity or forage terms They concluded that the population was expressing densdens-ity-dependent behavior during the study period

McCullough (1999) cited the intrinsic rate of increase of the population, scale of area occupied

by the population, heterogeneity of environment, and general quality of the habitat as factors that might explain why ungulates respond differently to a range of densities Scale of the area occupied

by the population refers to confined areas such as enclosures or islands where limitations on dispersal may change population growth rate A heterogeneous environment allows ungulates more types of high-quality food, leading to competition among different classes of individuals as density increases

Finally, high-quality habitats, including moderate temperatures and precipitation, lead to high plant production On the contrary, habitats with strong limitations on plant growth may result in most forage being of low quality, except perhaps in occasional good years

White-tailed deer have the potential to have a high rate of increase in rich habitats with stable environments, where female fawns commonly breed Such populations may have a

density-dependent function (r on N) that is all ramp, with no plateau (Figure 12.1A) Such populations express

density dependence virtually all the time (Figure 12.1a), even though N may be well below K.

McCullough (1999) listed mule deer among the species that may exhibit a plateau-ramp

density-dependent function (Figure 12.1B and b) He stated that these populations may not reach K very often,

because they tend to live in variable environments, and have lower rates of increase as compared

to white-tailed deer Populations that fit the plateau-ramp hypothesis (Figure 12.1B and b) have a plateau of the density-dependent function where no density-dependent responses would occur Only

when these populations approach K are density-dependent effects observed When these populations are significantly below K, density-dependent effects are not observed (Figure 12.1B and b) McCullough (1999) felt that desert mule deer (O H crooki) may exhibit the density-dependent

function shown in Figure 12.1C, based on the work of Short (1979) This hypothesis fits populations with low intrinsic rates of increase, homogenous habitats with mostly low-quality forage, and variable environments Figure 12.1c shows that these populations only exhibit density-dependent responses during occasional favorable periods

This review of literature shows that Odocoileus population dynamics are complex and frequently

site specific The complex nature of population dynamics in this genus makes formulating a general population model challenging Almost without exception, researchers cite McCullough’s (1979) George Reserve work as the conventional model for density dependence However, as this review

has shown, there are many situations where density-dependent behavior in Odocoileus populations

cannot be detected How widespread are habitats where the assumption of density-dependent beha-vior is not useful to management? Our objectives in this chapter are to (1) analyze three long-term sets of white-tailed deer counts in South Texas for density-dependent behavior, (2) suggest some

uni-fying concepts for considering density-dependent and density-independent behavior of Odocoileus

populations, and (3) suggest regions of deer range where population behavior cannot regularly be predicted with density-dependent models

TESTING SOUTH TEXAS DEER COUNTS FOR

DENSITY DEPENDENCE

Early European explorers in South Texas, United States, found a landscape that was mostly grassland, commonly interspersed with shrub communities (Inglis 1964; Fulbright 2001) White-tailed deer were present in wooded stream bottoms, shrub communities on upland areas, and on the open prairie Little is known about deer populations from this period, except they were commonly mentioned in traveler’s journals (Doughty 1983, 29; Fulbright 2001)

Cattle and horses were at least locally numerous by the mid-1700s (Lehmann 1969) Shrubs probably began to increase at this time, and this trend continued into the twentieth century (Jones 1975) Although famous for cattle ranching, the region harbored millions of domestic sheep in the latter part of the nineteenth century (Lehman 1969) Climatic change may have been a background condition influencing changes in plant ecology in the region, with grazing by domestic livestock being the driving force (Van Auken 2000) A cool, wet period lasted from about 1350 to 1850 (Foster

1998, 9) After 1850, the climate became warmer and dryer, which may also have influenced the increase in shrub density and distribution Removal of fuel by livestock grazing and suppression

by humans reduced or eliminated natural fires that inhibited the increase in woody plants during pre-Columbian times (Van Auken 2000) In the twentieth and early twenty-first centuries, the region has been covered by a canopy of shrubs, frequently in complex taxonomic mixes (Inglis 1964;

Jones 1975) Exclusive of the coastal sand plain and coastal prairie, over 90% of the region has been subjected to≥1 attempts to reduce shrub density to increase cattle-carrying capacity (Davis

and Spicer 1965)

Increased shrub density during at least the past two centuries may have facilitated increased deer populations Deer did not become locally extinct in South Texas after European settlement as they did in much of North America This was the result of low human density in the region, and large land ownerships Roads and highways were scarce until the 1920s when oil exploration began in earnest South Texas, particularly the King Ranch and the Aransas National Wildlife Refuge, provided deer for reestablishing populations elsewhere in the state

Historically, medium- and large-sized predators of deer have included jaguar (Panthera onca), mountain lion (Puma concolor), bobcat (Lynx rufus), black bear (Ursus americanus), gray wolf (Canis lupus), red wolf (Canis niger), and coyote (Canis latrans) Of these, wolves and jaguar are

extirpated Black bears are limited to occasional dispersers from northern Mexico Mountain lions are present in generally low, but apparently increasing density with high densities in localized areas (Harveson 1997) Coyotes and bobcats are present throughout the region, often at high population densities

Mountain lions prey on deer in South Texas, but do not appear to exert a region-wide influence

on populations Bobcats kill deer but do not appear to be an important factor to deer populations (Blankenship 2000; Ballard et al 2001) Studies in the 1960s and 1970s in eastern South Texas showed significant coyote predation on deer fawns (Cook et al 1971; Beasom 1974; Carroll and Brown 1977; Kie and White 1985) Meyer et al (1984) suggested that in addition to coyote predation, poor summer nutrition may be a strong factor in low South Texas fawn survival

Even before there was any formal management of deer populations, South Texas was well known for producing large-antlered bucks (Brothers and Ray 1975; Helmer 2002) Large antlers are

con-sistent with populations well below K carrying capacity (McCullough 1979) Examples of irruptive

behavior in deer populations in the region are lacking, although an irruption was experimentally induced by Kie and White (1985) Fawn survival from birth to fall is erratic (Ginnett and Young 2000) and low compared to white-tailed deer populations in general (Downing and Guynn 1985) Unhunted and otherwise unmanaged deer populations persist in a generally healthy state on some large, remote ranches

The region is virtually all private land, much of which is leased for hunting Intense interest

in deer management has developed among landowners and hunters during the past three decades Wildlife biologists in the region commonly prescribe management practices for deer populations based on the assumption of density-dependent population behavior (Brothers and Ray 1975: 62) This is particularly true of prescriptions to harvest does, as there is a common belief that without significant doe harvest, deer populations will increase to undesirable levels

In variable environments, K carrying capacity (Macnab 1985) varies from year to year The same number of animals may be above K in dry years and below K in wet years (McCullough 1979:

156) The negative feedback of animals on food plants is less important as a population influence

compared to the annual swings in K For South Texas, the CV for annual rainfall varies from 29 to

41% (Norwine and Bingham 1985) Rainfall occurs throughout the year with statistical peaks in May and September The average growing season is about 300 days; however, plant growth can occur any month when moisture and temperature permit (Box 1960; Ansotegui and Lesperance 1973)

We analyzed for density dependence in long-term time-series of deer counts on three study areas The Faith Ranch (28◦15N, 100◦00W) was 16,115 ha in the western portion of South Texas, the

Chaparral Wildlife Management Area (28◦20N, 99◦25W) consisted of 5,930 ha approximately in

the center, and the Rob and Bessie Welder Wildlife Foundation Refuge (28◦6 N, 97◦75W) was

3,158 ha in the eastern portion of the region (Figure 12.2)

For the Faith Ranch, a single helicopter survey of deer was conducted annually during 1975–1977 and 1981–1997 This consisted of flying adjacent belt transects about 200 m wide, at a height of about

20 m and speed of about 55 km/h (DeYoung 1985; Beasom et al 1986) Surveys encompassed various

*

FIGURE 12.2 Location of Faith Ranch, Chaparral Wildlife Management Area, and Rob and Bessie Welder

Wildlife Refuge, South Texas, USA

portions of the Faith Ranch over time (Table 12.1) The time series consisted of raw, unadjusted numbers of deer counted each year and classified as does, bucks, and fawns Because of the variation

in area flown and counted over the time series, deer numbers were transformed into deer density (deer/405 ha)

On the Chaparral Area, population estimates from 1969 to 1975 were made by spotlight counts following methods described by Fafarman and DeYoung (1986) A density estimate was calculated for the spotlight route on the Chaparral Area The density estimate was subsequently projected to a population estimate for the entire area each year During 1975–1997, a single, complete-coverage helicopter survey (DeYoung 1985; Beasom et al 1986) was conducted on the entire area The time series consisted of estimates of deer population size derived from spotlight surveys, or the raw, unadjusted number of deer counted by helicopter each year (Table 12.2) The change in census methods undoubtedly introduced additional variability into the time series Fafarman and DeYoung (1986), working on the Welder Wildlife Refuge, reported that population estimates from spotlight counts were about 10% higher than raw winter helicopter surveys Deer were classified as does, bucks, and fawns during counts by spotlight and helicopter A few unidentified deer were also tallied, but excluded from analysis Because the same area was counted throughout the time series, number of deer estimated or counted was used to form variables for analysis

Census data were available on the Welder Wildlife Refuge from 1963 to 1997, except for 1964 and 1969 During 1963–1976, population estimates were made by spotlight counts (Fafarman and

TABLE 12.1 March–May Rainfall and Number of Deer by Class Coun-ted during Fall Surveys by Helicopter on the Faith Ranch, South Texas, USA, 1974–97

a Mean March–May rain = 13.08; CV = 53.7.

DeYoung 1986) During 1977–1998, estimates were made by a single helicopter survey (DeYoung 1985; Beasom et al 1986) conducted in January each year The change in census method undoubtedly introduced additional variability into the time series, as noted for the Chaparral Area Helicopter surveys were made using procedures similar to those described for the Faith Ranch and Chaparral Area, except belt transects were spaced to result in about 50% coverage of the Refuge The unadjusted number of deer counted was used for all years (Table 12.3) Breakdowns by class of deer were not available for all years However, because an estimate of recruitment was needed for some of the time series analysis, mean number of embryos per mature doe collected each year for scientific purposes were substituted for fawns counted during census (Table 12.3)

The first method used to test for density dependence was that suggested by White and Bartmann (1997, 128) This involved regressing the variable tested for density dependence(V) against the estimate of number of deer (N t ) and N t2without an intercept as follows:

V t = B1N t + B2N t2 and then testing the null hypothesis of B2 = 0 If the test rejects the null hypotheses and B2is less

than 0, then V has been shown to be density dependent.

TABLE 12.2 March–May Rainfall and Number of Deer by Class, Counted during Fall Spotlight Counts (1969–75) or Survey by Helicopter (1976–97)

on the Chaparral Wildlife Management Area, South Texas, USA

March–May

a Mean March–May rain = 17.01; CV = 48.4.

For the Faith Ranch time series, all years (21) available were used and V t = fawns/405 ha,

whereas N t = does/405 ha The Chaparral Area data consisted of 29 years of time series with V t=

number of fawns and N t = number of does For the Welder Refuge, we used 33 years of data (no

counts were available for 1964 and 1969) and V t = mean embryos/adult doe, whereas N t= number

of deer

The second method used to test for density dependence was described by Dennis and Otten (2000) Because Ginnett and Young (2000) showed rainfall influencing fawn: doe ratios in South Texas, a rainfall term was included in the model The model used was written as

N t = N t−1exp(a + bN t−1+cW t−1+σZ t ),

TABLE 12.3 March–May Rainfall, Deer Density Determined by Spotlight Counts (1963–76) or January Survey by Helicopter (1977–98), and Mean Number of Embryos

in Adult Does Collected for Scientific Purposes on the Welder Wildlife Refuge, South Texas, USA

a Mean March–May rain = 21.63; CV = 49.7.

where N t is deer abundance (density in the case of the Faith Ranch) at time t (year: t= 0, 1, 2, …,

number of years in time series), W t is spring (March, April, and May) rainfall total (cm) for time t, and Z t is standard noise (with Z1, Z2, uncorrelated) Unknown parameters to be estimated from the data were a, b, c, and σ The random variables Z trepresent unpredictable fluctuations in growth rate

(logarithmic) over and above fluctuations accounted for by density dependence and precipitation.

Under this model, the population abundances N t (t = 1, 2, …) are random variables correlated

through time, and N ois fixed SeeDennis and Otten(2000) for details on methodology

Four cases of the model were fitted to the data for each study area as separate statistical hypotheses:

H0: b = 0 and c = 0 (no density dependence, no rainfall effect); H1: b = 0 and c = 0 (density

dependence, no rainfall effect); H2: b = 0 and c = 0 (no density dependence, rainfall effect); and

H3: b = 0 and c = 0 (density dependence, rainfall effect).

We calculated maximum-likelihood estimates of unknown parameters in the model for all four hypotheses using time series data from each study area in conjunction with rainfall data (Dennis and Otten 2000) Because of missing years in the time series, for the Faith Ranch we used data from

1982 to 1997 (16 years) and for the Welder Refuge data from 1970 to 1997 (28 years) were used

We tested for density dependence and rainfall effects on density (deer/405 ha) of total deer, adult deer, does, bucks, and fawns using the four hypotheses on the Faith Ranch We tested for density dependence and rainfall effects on total number of deer, adult deer, does, bucks, and fawns for the Chaparral Wildlife Management Area For the Welder Refuge, we tested for density dependence and rainfall effects on total number of deer and mean number of embryos per adult doe (although fall rather than spring rainfall may have more influence on embryos/doe)

Statistical hypotheses were tested using parametric bootstrapping (Dennis and Taper 1994) Four statistical hypothesis tests were conducted for the density dependence–rainfall model, as follows:

H0 versus H1, H0versus H2, H1versus H3, and H2versus H3 For these tests, the null model is contained within the alternative model as a special case, and is obtained by setting one parameter equal to 0 Details of this approach are in Dennis and Taper (1994) and Dennis and Otten (2000) Analysis of the time series by the method suggested by White and Bartmann (1997) showed no density dependence on the Faith Ranch, but did indicate that density dependence was operating at Chaparral and Welder (Table 12.4) The model for Chaparral was heavily influenced by very high census counts in 1979 and 1982 (Table 12.2) If these data points are omitted, density dependence is

not indicated (P= 247)

Covariance analysis by the method of Dennis and Otten (2000) showed a similar trend to the White and Bartmann (1997) model (Tables 12.5–12.8) For the Faith Ranch, hypothesis tests provided

no support for density dependence either with a rainfall covariate (H2versus H3, P≥ 26) or without

the rainfall covariate (H0versus H1, P ≥ 15) for any response variable For the Chaparral Area,

TABLE 12.4

Tests for Density Dependence in Time Series of Deer Abundance and

Reproduction Using the Method Suggested by White and Bartmann

(1987) for the Faith Ranch, Chaparral Wildlife Management Area, and

Welder Wildlife Refuge, South Texas, USA

Study area

Number of years

Nt= doe density t= 0.76

P > 0.05

Nt= does t= −1.99

P= 0.028

Nt= deer t= −9.85

P < 0.0001