Riva T. Madan Wild Turkey Population Change Spring 2015 1 Predictors of Wild Turkey (Meleagris gallopavo) Population Change in California from 1972 to 2013 Riva T. Madan ABSTRACT Wild turkeys never lived in California until humans began introducing them around 1908. The success of this introduced species was variable at first, but in the 1990s there was a noticeable increase in their population and range. The reasons behind what drove these population changes are unknown, especially with only a few studies on wild turkeys in California. With Breeding Bird Survey, National Land Cover Database, PRISM Climate Group, and California Department of Fish and Wildlife data, I used a zero-inflated poisson mixed model to determine the influence of climate, translocations, hunting, and land cover on wild turkey population change. I ran a separate model for three land cover distances. I calculated population change statewide and at individual routes using a generalized linear model. Wild turkey populations increased 10% per year statewide from 1972 to 2013. Hunting was the most influential predictor of population change showing that populations are increasing even as hunting increases. Land cover was a significant predictor, but the effects of different land cover types varied with the scale looked at. Urban land cover was positively related to population change only at 1 and 10 km distances. At the 5 km distance, forest, grassland, and agriculture were inversely related to population change. This study suggests increasing urban land cover, which may increase food supply, can increase wild turkey populations. However, the results also suggest that other factors that are not included in this model have large influences on wild turkey population change. KEYWORDS introduced species, Breeding Bird Survey, zero inflated poisson mixed model, land cover change, ArcGIS
42
Embed
Predictors of Wild Turkey (Meleagris gallopavo) Population ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Riva T. Madan Wild Turkey Population Change Spring 2015
1
Predictors of Wild Turkey (Meleagris gallopavo) Population Change
in California from 1972 to 2013
Riva T. Madan
ABSTRACT
Wild turkeys never lived in California until humans began introducing them around 1908. The
success of this introduced species was variable at first, but in the 1990s there was a noticeable
increase in their population and range. The reasons behind what drove these population changes
are unknown, especially with only a few studies on wild turkeys in California. With Breeding Bird
Survey, National Land Cover Database, PRISM Climate Group, and California Department of Fish
and Wildlife data, I used a zero-inflated poisson mixed model to determine the influence of
climate, translocations, hunting, and land cover on wild turkey population change. I ran a separate
model for three land cover distances. I calculated population change statewide and at individual
routes using a generalized linear model. Wild turkey populations increased 10% per year statewide
from 1972 to 2013. Hunting was the most influential predictor of population change showing that
populations are increasing even as hunting increases. Land cover was a significant predictor, but
the effects of different land cover types varied with the scale looked at. Urban land cover was
positively related to population change only at 1 and 10 km distances. At the 5 km distance, forest,
grassland, and agriculture were inversely related to population change. This study suggests
increasing urban land cover, which may increase food supply, can increase wild turkey
populations. However, the results also suggest that other factors that are not included in this model
have large influences on wild turkey population change.
KEYWORDS
introduced species, Breeding Bird Survey, zero inflated poisson mixed model, land cover
change, ArcGIS
Riva T. Madan Wild Turkey Population Change Spring 2015
2
INTRODUCTION
Humans often accidentally or intentionally transport species to new locations beyond their
natural range. If the species successfully establishes itself, the species becomes an introduced
species. At this point their populations may grow very large and spread over a large area. If the
introduced species has both expanded in range and is causing ecological impacts, the species is
considered invasive (Duncan et al. 2003). Large populations of introduced species can also be
destructive to human activities, making the species become a pest (Baldwin et al. 2012). Wild
turkeys (Meleagris gallopavo) are an example of an introduced species in California that has
expanded in range and increased in population. Although they are potentially a pest, wild turkeys
are not considered invasive due to lack of research on its ecological impacts.
The wild turkey never originally lived in California. Is native range is east of the
Mississippi River, in the southwest United States (Arizona, New Mexico, and Texas), and Mexico
(Gardner et al. 2004). In the late 1800s, wild turkey populations in their native range declined as a
result of deforestation and hunting (Dickson 1992; Barret and Kucera 2005). As a reaction to the
fear of turkeys going extinct, turkeys were introduced into every state except Alaska (Dickson
1992). The first record of wild turkey introduction in California was in 1877, when private ranchers
introduced the wild turkey on Santa Cruz Island. The next recorded introduction was in 1908 when
the California Department of Fish and Wildlife (CDFW) began releasing turkeys for hunting
purposes. Since then the CDFW has released several thousands of turkeys throughout California,
but many of these introduction attempts failed (Gardner et al. 2004). However, the CDFW noticed
an increase in wild turkey population size and range in the 1990s. As a response, they stopped
releasing turkeys in 1999 (Gardner et al. 2004). The reasons behind the recent spread and growth
of wild turkey populations in California are not well understood, but past studies of wild turkeys
in their native range provide knowledge about what effects their population growth.
Nesting failure and food availability are two main factors that determine wild turkey
population growth. These two factors largely depend on habitat type (Spohr et al. 2004; Lehman
et al. 2008; Fuller et al. 2013). The primary reason for nest failure is from predation and the chance
of nest predation depends on the density of vegetation. In areas with somewhat dense vegetation,
turkeys are able to hide their nests from predators well and therefore reduce nest predation. Open
grasslands would have a high amount of predation due to little nest obscurity. However, if the
Riva T. Madan Wild Turkey Population Change Spring 2015
3
vegetation is too dense, such as a dense shrubland, it reduces the ability of a turkey to flee if a
predator arrives, thereby increasing predation (Keegan and Crawford 1997; Lehman et al. 2008;
Fuller et al. 2013). Urban and agricultural areas have been shown to increase food availability.
Wild turkeys are often reliant on agricultural grains and human feeding to maintain large
populations when other food sources are low (Burger 1954; Barrett and Kucera 2005). Although
urban areas may provide additional food, habitat fragmentation from urbanization has resulted in
increased nest predation (Hogrefe et al. 1998; Sphor et al. 2004). Contrastingly, habitat
fragmentation between forests, open areas and agriculture supports large turkey populations
(Glennon and Porter 1999). A majority of these studies on factors affecting population growth took
place in several states in the wild turkey’s native range, but not in California. There have only been
four studies (Burger 1954; Gardner et al. 2004; Barret and Kucera 2005; Wengert et al. 2009) on
wild turkeys in California and three of the four studies are outdated by at least a decade.
Land use changes increasing habitat fragmentation, urbanization and agriculture bring wild
turkeys closer to humans and make them more likely to become pests. Wild turkeys were shown
to damage twenty-three crops, with corn being the most often damaged crop (Tefft et al. 2005).
Additionally, residents of suburban areas have complained about large numbers of turkeys being
a nuisance on their property (Barrett and Kucera 2005). However, no study has analyzed whether
these land use types have encroached on wild turkey habitat or whether wild turkeys have moved
from their original release sites to these land use types. Besides anecdotal and observational
information, there has not been any studies looking at how the population and range of wild turkeys
has changed in California and the possible factors driving the changes.
To better understand wild turkeys in California and as a step towards determining if they
are pests or invasive species, this study maps and analyzes the changes in population and range to
explain the reasons behind the recent spread and growth of wild turkey populations. Climate and
human involvement, such as hunting and translocation, have often influenced population changes
of species, but I hypothesize that land cover, specifically agriculture and urbanization, had the
greatest influence in increasing wild turkey populations in California. To see whether wild turkey
populations are actually increasing, I quantified how wild turkey population changed and mapped
how range changed in California from 1968 to 2012. To determine what factors could have led to
the changes in population, I ran a model to determine the influence of land cover, climate, and
human involvement on wild turkey population changes.
Riva T. Madan Wild Turkey Population Change Spring 2015
4
METHODS
Study system
The sites I looked at were predetermined by the USGS Patuxent Wildlife Research Center’s
North American Breeding Bird Survey that carries out bird surveys to determine presence and
count of many bird species, one of which being the wild turkey. Wild turkeys are generalists; they
can live in many different types of habitats, such as forests, shrublands, grasslands, woodlands and
around agriculture and urban areas (Dickson 1992). Each study site is surrounded by a combination
of these land types. Percentages of each land type: grassland, shrubland, forest, agriculture, and
urban, are given for each site with wild turkeys present in Appendix A.
Data sources
I obtained georeferenced wild turkey population data from the North American Breeding
Bird Survey (BBS) to determine population change. The BBS conducts point count surveys along
40 km long predetermined, non-random roadsides called routes (Figure 1). Citizens skilled in avian
identification conduct surveys between mid-May and early July. Beginning a half hour before
sunrise, the surveys, conducted by one observer, take 4 to 5 hours to complete. Point counts are
taken every 0.8 km along the route, for a total of 50 stops. The point count is conducted by
recording every bird seen or heard within a 400 m radius over a 3 minute timespan (Link and
Sauer 2002; Sauer and Link 2011). The USGS Patuxent Wildlife Research Center has been
conducting the BBS since 1966, but surveys in California began in 1968. I only included data
beginning in 1972 because I originally planned on using LANDSAT satellite imagery that began
in 1972 for land cover type, but the classification was too inaccurate to use. Only one detection of
a wild turkey in 1969 was left out from my analysis. Although the BBS is conducted annually, not
every route is surveyed each year. Turkeys have been observed in 80 routes in California between
1972 and 2013. I only included 76 routes because there was no geospatial information on the path
of these four routes.
Riva T. Madan Wild Turkey Population Change Spring 2015
5
Figure 1: Routes with and without wild turkeys present
I obtained wild turkey translocation and hunting data in California from the California
Department of Fish and Wildlife (CDFW). Translocation data included translocation within
California and between California and another state. Translocation data is from 1959 to 1999 with
the year 1998 missing. For hunting data, CDFW calculated the amount of turkeys hunted by
extrapolating the results from surveys. Mail survey forms were sent out to randomly selected
hunters until approximately 4% of the hunters returned the survey. CDFW has not documented the
specific methodology of the extrapolation. Data spans from 1949 to 2010, with the year 2009
missing. However, there was still hunting in 2009 and after 2010. Instead of leaving zero values
indicating no hunting, which would be incorrect, I used the same data from 2010 for years 2009 to
2013. This was reasonable given that the counties where wild turkeys were hunted had very little
variation since 2000. Both of these datasets are only specific to county, but routes are at a smaller
scale and sometimes span two counties. Therefore it would be incorrect to directly assign each
route a specific value for this data. To account for this inaccuracy, I only included hunting and
translocation as binary data to show whether it occurred in the county the route in. I assigned a
Riva T. Madan Wild Turkey Population Change Spring 2015
6
county to each route based on which county the majority of the route was located in. For a few
routes that were about evenly split between counties, I combined both of the county’s data.
To obtain land cover data, I used USGS National Land Cover Database (NLCD) for 1992,
2001, 2006 and 2011. I only looked at five land cover classes: forest, urban, grassland, agriculture
and shrubland. Because these datasets contain more than just five land cover classes, I combined
similar land cover classes together. For example, I included both evergreen forest, mixed forest,
and deciduous forest under the forest land cover type. I calculated percentage of land cover within
three buffer distances from the route path: 1, 5 and 10 km. Wild turkey flock home range varies
from 1.5 to 10 km2, therefore I will use multiple distances to account for the various home range
sizes as well as to see if habitat has different effects at different scales (Zeiner et al. 1990; Gardner
et al. 2004; Barrett and Kucera 2005). For years that didn’t have a given NLCD, I used the same
land cover values from the next available dataset with the assumption that the land cover didn’t
change very much between these years. Years prior to and including 1992 had land cover values
from the 1992 dataset, years 1993 to 2001 had values from the 2001 dataset, years 2002 to 2006
had values from the 2006 dataset and years 2007 to 2013 had values from the 2011 dataset.
For climate data, I used temperature and precipitation raster images from Oregon State
University PRISM Climate Group. To obtain average temperature and precipitation for each route,
I used ArcGIS v10.2 (ERSI 2014) to average the temperature and precipitation within 10 km from
each route path. I only used one buffer distance for climate variables because it is unlikely that the
climate will vary significantly under 10km from the route. I squared the average temperature and
precipitation because animals tend to have a quadratic relationship to temperature and precipitation
rather than linear.
Data Analysis
To look at range change, I used kriging, an interpolation technique, in ArcGIS v10.2 to
determine wild turkeys range in California for 6 year intervals. To quantify population change of
wild turkeys at each route and California overall from 1972 to 2013, I used a generalized linear
model (GLM) in R (R Core Team 2014) v3.1.2 using the glm package. I did not run a GLM on
routes that only contained one non-zero data point and I excluded the results of routes where the
generalized linear model didn’t converge due to lack of data. To see the effect of different
Riva T. Madan Wild Turkey Population Change Spring 2015
7
predictors on population change, I used a zero-inflated poisson mixed model in R v3.1.2 using the
glmmADMB package on all 77 routes. I included the percentage of land cover type, temperature,
precipitation, the occurrence of wild turkeys translocation and hunting as covariates. Route was
included as a random effect to account for differences between sites that cannot be explained by
the covariates. I standardized all values, excluding binary and percent values, by subtracting the
mean and dividing by the standard deviation to get more comparable covariate estimates. I ran a
separate zero-inflated model for each of the 3 land cover buffer distances.
RESULTS
Population and Range Change
Wild turkeys throughout California increased in population 10.09% (±0.279%) from 1972
to 2013, according to the generalized linear model (p < 0.0001). Different areas in California
experienced various amounts of population change (Figure 2). Majority of routes experienced less
than a 25% increase in population. More specifically, 31 out of the 61 routes included for this
analysis had an increase between 9% and 20% Route 416 near Meadow Valley in Plumas county
experienced the greatest population change of 108% per year, but was statistically insignificant (p
= 0.0644). Route 210 in Mendocino County and route 12 in Sutter County had increases of 88%
and 67%, respectively, but were also not statistically significant. Route 172, located east of
Berkeley in Contra Costa County, had the greatest statistically significant increase of 39% per year
(p < 0.001). Route 422 located south of Yosemite National Park had the next highest significant
increase of 30% per year (p < 0.01). Route 415 near Crescent Mills in Plumas County and route
202 in Sonoma County had a significant increase of about 28% and 27%, respectively (p < 0.5).
Four routes had a decrease in wild turkey population, but only one route had a significant decrease
in population. This route, route 409 located between Burnt Ranch and Hyampom in Trinity
County, decreased in population of 81% per year (p < 0.05). The percent population change per
year, standard errors and p-values for each route is given in Appendix B.
Range expansion can be seen through a significant (p < 0.0001) increase of 0.266%
(±0.035%) per year in the proportion of wild turkey detected on BBS routes from 1969 to 2013
(Figure 3). This is further supported by kriging maps of six year intervals which show wild turkey
Riva T. Madan Wild Turkey Population Change Spring 2015
8
range, indicated in blue, to be increasing (Figure 4). Based on the maps, the largest expansion was
between 1984 to 1989 interval and the 1990 to 1995 interval. After 1990, the population, indicated
by the darkness of blue, increased more than the range changed. From 1972 to 1989, turkeys
increased 24.8% (±4.56%) per year. However, this increase was actually insignificant (p = 0.489).
In the following period, 1990 to 2013, turkey populations increased by 91.3% (±4.56%) per year
(p < 0.0001). Given that there was no significant increase in population before 1990 and that the
kriging maps show very little change in population and range, I decided to only include years 1990
to 2013 when running the zero-inflated poisson model. Additionally, because the land cover data
only begins in 1992 and all years prior to 1992 would have the same land cover value, the result
of the model would be more accurate after removing the years before 1990.
Figure 2: Geographic distribution of population change. Percent population change, calculated by the generalized
linear model, of wild turkeys from 1972 to 2013. Routes with significant population change are indicated by a black
dot.
Riva T. Madan Wild Turkey Population Change Spring 2015
9
Figure 3: Wild turkey presence. Proportion of Breeding Bird Survey routes in California where wild turkeys were
detected from 1972 to 2013.
Figure 4: Wild turkey range expansion. I mapped the change in wild turkey range and population from 1972 to
2013 by averaging BBS counts for six year intervals for each route and then used kriging (in ArcGIS).
0
0.1
0.2
0.3
0.4
0.5
0.6
1971 1977 1983 1989 1995 2001 2007 2013
Perc
enta
ge
Year
Riva T. Madan Wild Turkey Population Change Spring 2015
10
Predictors of population change
According to the zero-inflated poisson mixed model for years 1990 to 2013, the most
significant predictor of wild turkey population change for all buffer distances was hunting (p <
0.01) (Table 1). The estimate for hunting was 0.8 for 1 km buffers, 0.74 for 5 km and 0.73 for 10
km. Year was the second most significant predictor than hunting, but still very influential with
estimates of 0.74 for 1 km, 0.78 for 5 km and 0.75 for 10 km. All other factors were much less
influential predictors with estimates of less than 0.1. Precipitation, temperature, translocation and
grassland were not significant for any of the buffer distances. The 95% confidence intervals
included zero, making the effect these covariates neither positive nor negative.
Different factors were significant and had various influences on population change at each
buffer distances. For the 1 km buffer, the percent of urban land cover was significant (p < 0.01)
with an estimate of 0.096. This indicates that areas with increasing urban land cover will
experience increasing population sizes, given that all other factors are held constant. Similarly, for
the 10 km buffer, urban land cover was also significant (p < 0.05) and positively related to
population change (estimate of 0.047). For the 5 km buffer, forest, shrubland and agricultural land
cover were significant covariates, but urban land cover was not (p = 0.078). Forest, shrubland and
agriculture were all inversely related to population change. Forest, with an estimate of -0.077 was
more of an influential predictor than shrubland and agriculture, both with an estimate of -0.023. A
negative estimate indicates that areas that are decreasing with this land cover type have an
increasing wild turkey population.
Table 1: Parameter values for the 3 buffer distances. Summary of each parameter’s influence on wild turkey
populations according to the zero-inflated mixed model. Significant p-values are indicated with *.