 

 

am A— .
_:.

19:;

— :-

.m.-u
.

'V “I V-
- -. g, - ‘
0'“ J I
““4;

'33

Juan ‘

v-
“u...“

m...“ -.
lap-«.4.
q,

L
N
3415!.

a
o:

':1~_ r
”3‘

.4

\. 7354'
"‘ .5116

.4‘:

.d

‘g'

\Ifi.
~1».4.

2.4

n
.if‘

.
I
l‘ﬁ. '
‘ A: .t?’
r mi: N :

i
3 y
if

F 3‘
:H-n' ‘3
I 115'!
t ‘J;'
”.3 a

.g
1‘ ‘ . V
tug. J . {hr ‘ ‘

LR...

' Z3

'3‘;
n’m film
a k\:}.: - _,
‘r

A

4.
RH
, 3‘ :2" '1‘“?! P1 W AT .
g2: ;§L§i§=xrfgigau4..zgu .
- 4%th r ‘Bﬁgx
. a}... “1‘11! It ‘
t diffﬂ i
23:4 53%;;
2‘)
I!

i?“ I, Fiﬁ-I
“if? %?

‘ 55-992. .
.. «'4».

“m

....~-..
2 w:
as;

.05

1‘7 1
We. .
II .
I

F”: .

“-

’Avv;
“AI-"M
,.."~,_
I‘d

'ia‘ii; i
, ~I

i-at’ﬁ‘j?

v-.u'.'."“

.m

w»-

u
.1»
”‘1‘ p.

In
’33:.
a

~‘
.u

3.? “153$
,

‘9

v‘

“I:

.V'
‘A‘

v “r: .«‘
: 3L3}?

and:

:3 " ’ t} 1’1 ‘1‘
H 5"“ W" m '
i '1. ' I ‘
Ta "1 - Lg} 1 (gig-‘9”? z: n
. “a" 11 ‘3' '£

1"? ‘
a.)
4

7 291 "
«74.--...
:13:

 

'w- ‘
s»- 1-.
" ,_._

.1”.

d

 

:1

W (\J'

b

“'3 LIBRARY
Michigan State
7 :2 J g ,6 University

:4)“

This is to certify that the
thesis entitled

ESTIMATION OF BLACK BEAR (URSUS
AMERICANUS) ABUNDANCE IN THE NORTHERN
LOWER PENINSULA OF MICHIGAN USING
MICROSATELLITE DNA MARKERS.

presented by

Brian P. Dreher

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in Fisheries and Wildlife

 

 

Eeﬂl/Mm

Major Professor’ 5 Signature

11—15- 07

 

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

11:0— 9211559005

w

 

102802

 

JUN :019 231.2

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 W Jada-p.15

 

ESTIMATION OF BLACK BEAR (URSUSAMERICANUS) ABUNDANCE IN THE
NORTHERN LOWER PENINSULA OF MICHIGAN USING MICROSATELLITE
DNA MARKERS.

By

Brian P. Dreher

A THESIS
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Fisheries and Wildlife

2004

ABSTRACT
ESTIMATION OF BLACK BEAR (URSUS AMERICAN US) ABUNDANCE IN THE
NORTHERN LOWER PENINSULA OF MICHIGAN USING MICROSATELLITE
DNA MARKERS.
By
Brian P. Dreher

The estimation of black bear population abundance has long been a difﬁcult task,
however valuable when justifying harvest quotas and managing populations. Recent
advancements in molecular genetics have provided a means to identify bears using DNA
contained in tissue and hair samples. We collected hair samples non-invasively using
barbed wire hair snares and tissue samples from harvested bears to genetically identify
individuals and estimate bear abundance in the Northern Lower Peninsula (N LP) of
Michigan using a capture-recapture methodology. In both 2002 and 2003 we derived hair
snare locations using a stratiﬁed random design and collected hair samples over a number
of sampling intervals. As an additional sampling occasion we collected tissue samples
from hunter harvested bears. We genetically analyzed hair and tissue samples with 5
microsatellite loci and quantiﬁed genotyping errors using samples from harvested bears.
We estimated the population of yearling and adult black bears in the NLP with models
that account for genotyping error to be 1,882 bears (95% CI 1,3 89-2,551 bears). We
created a simulation model to quantify the effects of genotyping error and sub-sampling
hair samples and found that genotyping error dramatically biases population estimates
and the selection of 3 hair samples, when multiple hair samples were available, reduced

the variance of population estimates.

ACKNOWLEDGEMENTS

I would like to start by thanking my committee, Scott Winterstein, Kim Scribner,
Guilherme Rosa and Dwayne Etter. Each member contributed in their own way to this
project and I am very thankful to all of them. Thanks especially to Scott, my major
advisor, who gave me guidance throughout the project. Thanks to my wife, Laura, for
moving to Michigan and always supporting me in my endeavors. Thanks to Veronica
Lopez for all of her work from the beginning. Whether it be collecting hair in the bug-
infested swamps, or working in the lab, Lopez was always helpful.

A big thanks goes out to Michigan State University, Michigan Department of
Natural Resources, Safari Club lntemational, and Michigan Bear Hunters Association for
funding this research.

There are many people that assisted in the ﬁeld portion of this project and made
everything possible. First, I would like to thank Eric Dunton, Mike Nattrass, Ryan
Carrasco, and Ryan Fry. These guys did a greatjob working for me in the ﬁeld. Second,
thanks to all the MDNR employees at the Atlanta, Mio, Gaylord, Bay City, Gladwin,
Roscommon, Cadillac, and Traverse City ﬁeld ofﬁces. I would especially like to
acknowledge Dan Moran for all of his efforts in assisting us in both ﬁeld seasons. Dan
made my life much easier getting all the supplies that we needed in the ﬁeld. I would
also like to speciﬁcally thank Larry Copley from whom I learned much and am thankful
for all of his assistance in running 18 hair snares for both ﬁeld seasons. Larry Visser

played a key role in this project both from the beginning and during the project with all

iii

his assistance and advice. I am thankful for all of his help. I would like to acknowledge
the hard work of the United States Forest Service, Grand Traverse Band of Ottawa and
Chippewa Indians, Little Traverse Bay Bands of Odawa Indians and the Little River
Band of Ottawa Indians.

I would like to acknowledge Kristi Filcek, and Scot Libants for all of their work
in the lab. Without their expertise, none of this would have been possible. Mike Wilberg
did a fantastic job creating GENECAP which made my job of data analysis and matching
genotypes more efﬁcient. All the folks in the Winterstein and Campa labs played an
essential role in helping me get through this project and they include Michelle Prehoda,
Dan Walsh, and Brandi Hughey. A special thanks to Ali Felix who helped in a number
of ways from reading my chapters to putting together presentations. Thanks to Dr. Dan
Hayes for his comments and advice about the project.

Thanks to Paul Lukacs from Colorado State University for all his assistance with
modeling, data analysis and advice. Thanks also to Dr. Gary White from Colorado State

University for assisting on the data analysis portion of the project.

iv

TABLE OF CONTENTS

LIST OF TABLES __________________________________________________________________________________________________ viii
LIST OF FIGURES ___________________________________________________________________________________________________ x
LIST OF APPENDICES _________________________________________________________________________________________ XII
CHAPTER I — GENERAL INTRODUCTION AND LITERATURE REVIEW____I
INTRODUCTION ......................................................................................... 1
Management of Black Bear in Michiganwmu, _______________________________ I
Overview of Population Estimation Techniques _______________________________ 4
Field Data Collection ________________________________________________________________________ 7
Lab Analysis of DNA ........................................................................ 8
Population Estimate and Analysis ................................................... 10
OBJECTIVES ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 12
THESIS ORGANIZATION ________________________________________________________________________ l3
LITERATURE CITED ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 15
CHAPTER 2 — FIELD SAMPLING METHODOLOGIES .................................... l9
STUDY AREA ____________________________________________________________________________________________ l9
METHODS ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 20
Hair Snare Placement ...................................................................... 20
Veriﬁcation of Sampling Design ..................................................... 24
Field Location of Hair Snare Sites .................................................. 25
Hair Snare Conﬁguration _________________________________________________________ . ______ 26
Checking Methods ............................................................................ 30
Hunter Harvested Bear Hair and Tissue Collection ________________________ 31
RESULTS ____________________________________________________________________________________________________ 32
Hair Snare Placement ______________________________________________________________________ 32
Veriﬁcation of Sampling Design _____________________________________________________ 33
Hair Snaring ____________________________________________________________________________________ 42
Hunter Harvest Tissue and Hair Collection .................................... 49
DISCUSSION _____________________________________________________________________________________________ 50
Hair Snare Placement ______________________________________________________________________ 50
Hair Snaring ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 53
Hunter Harvested Bear Tissue Collection ....................................... 57
LITERATURE CITED 58

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

CHAPTER 3 — ESTIMATION OF BLACK BEAR ABUNDANCE IN THE
NORTHERN LOWER PENINSULA OF MICHIGAN USING

MICROSATELLITE DNA MARKERS ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 60
INTRODUCTION ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 60
STUDY AREA 62

METHODS 62

....................................................................................................

Sampling Location Identiﬁcation ...................................................... 62

Hair Snaring ...................................................................................... 65
Hunter Harvested Bear Tissue Collection ......................................... 65
Genetic Analysis _______________________________________________________________________________ 66
Sample Selection ___________________________________________________________________ 66

Microsatellite Genetic Analysis ............................................. 67

Quality Control ..................................................................... 68

Genotyping Error Estimation _________________________________________________ 70

Statistical Genetic Analysis ___________________________________________________ 71

Population Estimation ....................................................................... 73
RESULTS ...................................................................................................... 75
Hair and Tissue Sample Collection ................................................... 75
Genetic Analysis _______________________________________________________________________________ 76
Sample Selection and Quality Control ________________________________ 76

Genotyping Error Estimation ................................................ 78

Statistical Genetic Analysis ___________________________________________________ 80

Population Estimate ___________________________________________________________________________ 81
DISCUSSION _______________________________________________________________________________________________ 84
Study Design _____________________________________________________________________________________ 84

DNA Sample Collection .................................................................... 84
Statistical Genetic Analysis _______________________________________________________________ 86
Quality Control and Genotyping Errors ____________________________________________ 88
Population Estimate ___________________________________________________________________________ 89
MANAGEMENT IMPLICATIONS ____________________________________________________________ 9]
LITERATURE CITED 92

CHAPTER 4 — PRECISION AND BIAS: CHALLENGES POSED BY NON—
INVASIVE DNA SAMPLING TO ESTIMATE BLACK BEAR ( URS US
AMERICAN US) ABUNDANCE IN THE NORTHERN LOWER PENINSULA

OF MICHIGAN ________________________________________________________________________________________________________ 96
INTRODUCTION ......................................................................................... 96
Genotyping Error ______________________________________________________________________________ 97
Sub-sampling _____________________________________________________________________________________ 98
Non-invasive Research ...................................................................... 98
METHODS .................................................................................................... 99
Genotyping Error Simulation ____________________________________________________________ 99
Genotyping Error Simulation Model __________________________________ 100

Sub-sampling Simulation ................................................................ 103
Sub-sampling Hair Samples Simulation Model __________________ 103

RESULTS ____________________________________________________________________________________________________ 105
Genotyping Error Simulations ........................................................ 105
Sub-sampling simulations ............................................................... 107
DISCUSSION _____________________________________________________________________________________________ 108
Genotyping Error ............................................................................ 108
Sub-sampling ___________________________________________________________________________________ l 10

vi

LITERATURE CITED _______________________________________________________________________________
CHAPTER 5 — RECOMMENDATIONS TO THE MDNR ___________________________________
Hair Snare Placement __________________________________________________________________________________
Hair Snaring ________________________________________________________________________________________________
Hunter Harvested Hair and Tissue Collection _____________________________________________
Sample Selection _________________________________________________________________________________________
Genetic Analysis _________________________________________________________________________________________
Population Estimate _____________________________________________________________________________________
Personnel and Cost Summary ______________________________________________________________________
Personnel S ummaty ________________________________________________________________________

Cost Summary _________________________________________________________________________________
Comparison to Current Methods and Further Work ___________________________________

LITERATURE CITED _______________________________________________________________________________

APPENDICES

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

vii

113

116
116
117
120
120
121
124
125
125
125
127
129

130

LIST OF TABLES

Table 2.1. Categories for the number of bears harvested per township for 1997-
2001 and 1997-2002 and the resulting number of hair snare locations chosen. .......... 23

Table 2.2. Number of harvests per township between 1997-2002 and the
frequency of townships in each category. ____________________________________________________________________ 33

Table 2.3. Land use /land cover categories, number of telemetry locations and
percent of all telemetry locations collected from 1991-2001. ...................................... 40

Table 2.4. Land use /1and cover categories, number of telemetry locations and
percent of the subset of telemetry locations collected from 1991-2001. ______________________ 41

Table 2.5. Number of hair snares set, number of checking occasions, total number of
checks made, number of snares that were visited by bears, total number of hair samples
collected, and the percent of snares visited for the summers of 2002 and 2003. ......... 42

Table 2.6. Total number of hair samples collected, mean, minimum and

maximum number of hair samples collected given a visitation occurred for 2002

and 2003. ______________________________________________________________________________________________________________________ 48
Table 2.7. Number of hair samples collected in 2003 that contained 1-4, 5-10, 11-

15, and >15 hair follicles. ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 49

Table 2.8. Summary table of other research studies, including this one, to

estimate bear abundance including the author and study area, species of study,

study area size, number of snares, number of snares visited, and the percent of

snares visited. _______________________________________________________________________________________________________________ 56
Table 3.1. Summary of the number of samples genotyped, number of samples

culled with S 3 loci score; number of samples that were contaminated; number of
samples culled with 5 4 loci scored; number culled after mismatch examination;

the total number culled and the ﬁnal number used for estimation analysis for each
selection category. ________________________________________________________________________________________________________ 77

Table 3.2. Summary of the genotyping error rates both before and after quality
control protocols of mismatch identiﬁcation. Error rates are reported per locus
and per allele and reported by locus. ____________________________________________________________________________ 79

Table 3.3. Summary of the number of alleles per locus and the PIsibs single locus
calculations for each locus used to differentiate individuals. 80

viii

Table 3.4. Closed capture estimation models for adult black bears in the NLP of

Michigan including model name, parameters, AICC, Ai, wi, N , and the 95%
conﬁdence interval. Estimates derived with closed capture estimation models in

program MARK (White and Bumham 1999). ............................................................. 82

Table 3.5. Closed capture estimation models for adult black bears in the NLP of

Michigan including model name, parameters, AICC, Ag, wi, N , and the 95%
conﬁdence interval. Estimates derived with models proposed by Lukacs and

Bumham (in press). ______________________________________________________________________________________________________ 83

Table 4.1. Allele and allele frequencies for 5 loci including GlOX, GIOL, GlOD,
UarMU59, and UarMUSO calculated from 544 bears in the northern lower

peninsula of Michigan. ________________________________________________________________________________________________ 101

Table 4.2. Probability (%) of the number of unique animals determined for the
number of hair samples observed for the given number of sites and the summary
of the number of hair samples collected from 239 hair snaring locations across 5
sampling occasions (1,195 possible sampling occasions). Probabilities derived
from genotyping 210 hair samples from 49 snare checking occasions. ,,,,,,,,,,,,,,,,,,,,,

Table 4.3. Mean population Size estimate, standard error of the estimates, and
percent relative bias for each genotyping error rate across 1,000 simulations. __________

Table 4.4. Number of hair samples genotyped per site, the mean of the 1,000
population estimate replicates and the resulting standard error of the estimates. ______

Table 5.1. Number of new bears and number of previously identiﬁed bears for
each sampling occasion conducted in the summer of 2003. ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,

Table 5.2. Summary of the number of possible alleles, number of observed
alleles, and the percentage of these observed alleles, for each locus used in the
2002 and 2003 hair and tissue genetic analysis. _________________________________________________________

Table 5.3. Summary of the personnel hours to perform the necessary tasks as
reﬂected in the portion conducted in 2003/2004 of this project. _______________________________

Table 5.4. Approximate cost for each expense for the 2003/2004 portion of the
project. ________________________________________________________________________________________________________________________

104

106

108

119

122

126

127

LIST OF FIGURES

Figure 1.1. Study area (36,848 kmz) comprised of the Baldwin, Gladwin and Red
Oak bear management units. __________________________________________________________________________________________ 6

Figure 2.1. Single wired hair snare for black bear hair collection comprised of a
single strand of barbed wire (13 cm barb spacing) encircling 2 3 trees at 50 cm

from the ground. Spacing between trees is approximately 2-4 meters apart. Two
warning Signs and ﬂorescent ﬂagging tied around the wire warn humans of the

snare. Bait is suspended over the snare by strings attached to over-hanging limbs.m28

Figure 2.2. Double wired hair snare used for black bear hair collection comprised

of 2 strands of barbed wire (13 cm barb spacing) encircling z 3 trees at 20 cm and

50 cm from the ground. Spacing between trees is approximately 4 meters apart.
Two warning signs and florescent ﬂagging tied around the wire warn humans of

the snare. Bait is suspended over the snare by strings attached to over-hanging

limbs. 29

ooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

Figure 2.3. Public land ownership (11,878 kmz) within the study area. _____________________ 34

Figure 2.4. Potential bear habitat within the study area comprised of emergent
wetland vegetation (emergent wetland/wet meadow, lowland broad-leaved

deciduous shrub, lowland broad-leaved evergreen shrub, lowland broad-leaved
evergreen shrub, other lowland shrub), mixed lowland (mixed lowland
conifer/hardwood, mixed lowland hardwood), and aspen/birch vegetation

categories. ____________________________________________________________________________________________________________________ 35

Figure 2.5. The 500 meter buffer on both sides of county roads, streets and
highways within the study area. __________________________________________________________________________________ 36

Figure 2.6. Final stratiﬁed coverage representing public land, potential bear
habitat, area > 500 meters from streets, county roads, and highways. This
coverage was used for the hair snare selection. ___________________________________________________________ 37

Figure 2.7. Number of bears harvested in each township within the study area
from 1997-2001. 38

..........................................................................................................

Figure 2.8. Number of bears harvested in each township within the study area
from 1997-2002. 39

Figure 2.9. Buffer based on average size of adult female annual home range (131
km2) around the 239 hair snares set in 2003. .............................................................. 43

Figure 2.10. Summer of 2002 hair snare locations. Black dots refer to hair snares

that were visited by bears and unﬁlled dots refer to snares that were not visited by
bears. Documented locations of 5 snares, symbolized by a triangle and referred to

as mislocated, were incorrect and the locations of these snares was approximated.m44

Figure 2.11. Summer of 2003 hair snare locations. Black dots represent the snares
that were visited by bears and the unﬁlled black dots represent snares that were
not visited by bears. ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 45

Figure 2.12. Number of hair snares visited by bears for each week of sampling in
2002. The week of 7/7 — 7/13 represents the l-week interval between sampling
periods. _________________________________________________________________________________________________________________________ 46

Figure 2.13. Number of hair snares visited by bears for each week of sampling in

Figure 2.14. The number of hair samples collected for each one-week of
sampling conducted in the summer of 2003. The week of 7/7 — 7/13 represents
the l-week interval between sampling periods. ___________________________________________________________ 47

Figure 2.15. The number of hair samples collected for each one-week of
sampling conducted in the summer of 2003. ________________________________________________________________ 47

Figure 4.1. Mean estimated population size and associated 95% conﬁdence
intervals across 1,000 simulations for genotyping error rates, 0.0/allele,
0.008/allele, and 0.029/allele. _____________________________________________________________________________________ 106

Figure 4.2. Standard error of the estimate resulting from the number of
individual hair samples chosen to be genotyped across 1000 simulated data sets.____107

xi

LIST OF APPENDICES

Appendix 2.1. 1993 GAP Land Stewardship land cover/ land use categories and
deﬁnitions. _____________________________________________________________________________________________________________________ 13 1
Appendix 2.2. 2000 GAP Land Stewardship land cover/ Land use categories and
deﬁnitions. _____________________________________________________________________________________________________________________ 136
Appendix 2.3. List of hair snare name, year, UTM location, number of visitations

by bears, number of hair samples collected and the county that the hair snare was
located for each management organization operating the snares. _ _ 139

 

Appendix 2.4. List of hair snare name, year, UTM location, number of visitations

by bears, number of hair samples collected and the county that the hair snare was
located for each management organization operating the snares. _________________________________ 154
Appendix 5.]. Hair snare set-up and checking procedures. __________________________________________ 170
Appendix 5.2. List of supplies needed for each task to be performed in the

collection of hair samples and hunter harvested bear hair and muscle tissue

samples. _________________________________________________________________________________________________________________________ 175
Appendix 5.3. Formed that was ﬁlled out upon the set-up of a hair snare. _________________ 176

Appendix 5.4. Copy of the stamp that was placed on all the hair collection
envelopes. ______________________________________________________________________________________________________________________ l 77

Appendix 5.5. Data collection form used when checking hair snares for the
presence or absence of hair samples. ............................................................................ 178

Appendix 5.6. Protocols for the collection of hair, muscle tissue, and tooth
samples. _________________________________________________________________________________________________________________________ 179

xii

CHAPTER 1 — GENERAL INTRODUCTION AND LITERATURE REVIEW

INTRODUCTION

Black bears (Ursus americanus) are a publicly owned resource managed in the
state of Michigan in the public trust by the Michigan Department of Natural Resources
(MDNR). In 1996 the public decided how black bears should be managed in Michigan
using the public referendum process. Two proposals (D and G) were placed on the
ballot. Proposal D was a statutory initiative lobbied by Citizens United for Bear (CUB)
to prohibit the use of bait piles and dogs for hunting black bears in Michigan (CRC
Memorandum 1996). Proposal G was a legislative referendum on Public Act 377 of
1996 that would give the exclusive authority of all hunting regulations to the Natural
Resources Commission (NRC) using “principles of sound scientiﬁc management” (CRC
Memorandum 1996). In the statewide vote, Proposal D was supported by 38 percent of
voters and proposal G was supported by 72 percent of voters (Peyton 1998). Bear
hunting using bait and dogs is still the primary means to manage the Michigan bear
population and the Natural Resources Commission must utilize “principles of sound

scientiﬁc management” in the decision-making process.

Management of Black Bear in Michigan

Bear hunting in Michigan occurs in fall in both the Upper and northern Lower
Peninsulas. Interest in black bear hunting is increasing as evidenced by the number of
people applying to hunt bear in the northern Lower Peninsula (NLP: 5,365 in 1995 and
14,674 in 2001). The number of bears harvested in the NLP increased from 136 bears in

1995 to 357 bears in 2002 (F rawley 2002). Black bear hunting provides a signiﬁcant

source of revenue, estimated at $5.3 million in the 1995 hunting season, to the state
through small local businesses and license fees (USDI 1998).

It is a stated goal of the MDNR Black Bear Management Program “to maintain a
healthy black bear population that provides Michigan residents with a diversity of bear-
related recreational opportunities. A bear population that provides viewing and hunting
opportunities, yet does not create excessive nuisance bear problems for people, is crucial
for the success of this program” (1998, MDNR bear hunting guide). Additionally,
Proposal G mandates that the Natural Resources Commission and the MDNR utilize
“principles of sound scientiﬁc management” in the decision-making process for black
bear management. A population estimation or survey could be considered one means of
obtaining requisite data to conduct “sound scientiﬁc management” (Bailey 1984). A
precise population estimate of black bears in the NLP would help to simultaneously
ensure the maintenance of recreational hunting opportunities and a viable black bear
population by providing a means to evaluate and justify harvest quotas and satisfy public
desire to know “how many bears there are.”

The bear population in the NLP is ecologically and socially important with a wide
array of human placed values (Peyton et al. 2001). According to the MDNR, the bear
population in the NLP is increasing and expanding geographically and this increases the
likelihood of human-bear conﬂicts (Etter et al. 2002). Peyton et al. (2001) surveyed
public attitudes towards bear in the NLP recognizing that the bear population in the NLP
is increasing and expanding into areas that were not previously occupied by the species in
recent history. Peyton et al. (2001) identiﬁed the areas of the NLP where tolerance for

black bears is expected (e.g. traditional bear range) and found an increase in bear

intolerance as residents were surveyed from north to south in the NLP. A precise
population estimate could be used to gauge public opinion about bears as the population
continues to expand its range in the Lower Peninsula (Peyton et al. 2001). This estimate
could allow the MDNR to relate actual black bear numbers with social carrying capacity,
and possibly proactively manage the population to avoid potential negative human-bear
interactions.

Recent research into NLP bear population demography (1991—2001) included the
use of radio-telemetry techniques, which enhanced the MDNR’S understanding of bear in
this region (Etter et al. 2002). Data collected from 126 radio-collared black bears
provided estimates of age- and sex-speciﬁc survivorship, recruitment, movements and
home range. Causes of non-harvest bear mortality in Michigan included collision with
automobiles, illegal killing, nuisance bear control, and unknown causes (Etter et al.
2002). Non-harvest mortality is low for both subadults (bears S 2 years) and adults
(bears > 2 years) with both sex and age classes exhibiting seasonal and annual mortality
rates < 10%. Legal harvest is the single largest cause of mortality (59% annual deaths)

for bears from the NLP (Etter et al. 2002). Home range estimates were some of the
largest reported for the species including 867 km2 for adult males and 131 km2 for adult

females (Etter et al. 2002).

Currently, the MDNR uses a sex- and age-speciﬁc population model developed
for bears from Minnesota to estimate population size and set hunting tag quotas
(Garshelis and Snow 1988). The ﬁve components of the model include initial population
size, birth rates, death rates, immigration, and emigration. For the NLP, empirical data

collected from 1991-2001 provide direct estimates of births, deaths, and movements

(Etter et al. 2002). However, without an accurate estimate of initial population size, the
MDNR model is not complete. Additionally, the MDNR model has not been validated
because indirect or direct estimates of black bear population abundance are currently not
available. Rather, the model has been veriﬁed using population indices such as a bait
station index.

In 1990 the MDNR created Bear Management Units (BMU’s) and limited the
number of licenses allocated in each unit (Frawley 2001). The NLP is comprised of three
units open to hunting: Red Oak, Baldwin and Gladwin BMU’S (Figure 1.1). Hunters that
are successful in harvesting a bear are required to present the entire bear or head at a
MDNR ofﬁce or designated registration location within 72 hours of the kill. Information
collected upon registration includes: hunter name and address, date of the kill, BMU in
which the bear was taken, location of the kill (township, range and section), county of the
kill, sex of the bear, method of take, and comments about the weight of the bear and
general body condition. Additionally, the MDNR attempts to collect teeth from all
harvested bears and reproductive tracts from female bears are voluntarily collected.

Teeth from individual bear are aged using the cementum method (Hildebrandt 1976).

Overview of Population Estimation Techniques

Currently, the MDNR uses an index of bear sightings, an annual bait survey, nuisance
bear reports, and number of bears harvested to survey populations (Etter et al. 2002).
These methods provide only population trends rather than an estimate of population size.
Population estimation techniques are difﬁcult to use on carnivores because of their

secretive nature, large home-range size and difﬁculty of capture (Woods et al. 1999;

Mills et al. 2000; Taberlet et al. 2001). Capture-recapture is one method that has been
used to survey large carnivores. Individuals are captured and outﬁtted with a
recognizable mark and released into the population. A number of individuals are
subsequently recaptured and examined for the previously placed marks. Information
about the number of individuals initially marked and the number of animals in the
recapture with and without marks can provide an estimate of the population (White et al.
1982). Mace et al. (1994) estimated the population size of grizzly bears (Ursus arctos
horribilis) by capturing and marking bears with colored ear tags and streamers.
Researchers then used remote cameras to recapture individuals to obtain a population
estimate. Although this method is feasible for use on large carnivores, capturing and
handling of individuals can result in small sample sizes and logistical difﬁculties.
Garshelis and Visser (1997) used baits laced with the ingested biomarker tetracycline to
non-invasively mark bears in the Upper Peninsula (UP) of Michigan. Individual bears
were then recaptured in the bear harvest when hunters submit tooth samples upon
registration. By examining the teeth or bone, the “mark” was revealed using ultraviolet
light, which ﬂuoresces the tetracycline biomarker. This method can be an effective
means to mark many individuals non-invasively thereby increasing the sample size, and
has proved to be cost-effective in the UP for estimating bear population size. However,
one disadvantage of the tetracycline marking technique is the limited biological
information that can be collected when marking an individual (e.g. sex). Furthermore,
this method provides no means for positively identifying recaptured individuals, which

limits the models that can be applied to the data.

 

 

 

 

 

 

 

 

Advancements in molecular genetics have made it possible to conduct a mark-
recapture study non-invasively using hair and tissue as a DNA source. Utilization of this
method involves three major parts: 1) ﬁeld collection of a DNA source, 2) lab analysis of
the DNA source to produce genotypes of individuals, and 3) estimation of the population

size using mark-recapture models.

Field Data Collection

Many different sources and methods exist for the non-invasive collection of DNA.
Gerloff et al. (1995) extracted DNA from feces of wild, free ranging bonobos (Pan
paniscus) for individual identiﬁcation. Kohn et al. (1999) used DNA extracted from
feces as a non-invasive means to estimate the size of a coyote (Canis latrans) population
in California. Pearce et al. (1997) used eggshells and feathers collected non-invasively in
nests as a DNA source for a conservation genetics study of common eiders (Somateria
fisheri) in both Russia and Alaska. In a population survey of the endangered Pyrenean
brown bear (Ursus arctos), Taberlet and Bouvet (1992) documented the use of hair as a
non-invasive source of DNA. This technique for brown bears included the collection of
hair samples from barbed wire nailed to trees and was a better alternative to handling
individuals because of the risk associated with capture and anesthesia of an endangered
species.

Woods et al. (1999), Mowat and Strobeck (2000), Poole et al. (2001), and
Boerson et al. (2003) successfully captured both black and brown bear hair using a
snaring device made from barbed wire. The basic conﬁguration of the hair snare used a

single strand of four-pronged barbed wire stretched around 3 or more trees at a uniform

height forming an enclosure around an attractant or lure. As bears pass over or under the

barbed wire, hair is deposited on the wire for subsequent collection.

Lab Analysis of DNA

Recent advancements in molecular genetics have provided a means to identify
individuals using small quantities of DNA collected from hair or feces (Higuchi et al.
1988). Extracted DNA in conjunction with microsatellite markers can be ampliﬁed using
the polymerase chain reaction (PCR). PCR results in millions of copies of a target
section of DNA, and occurs in the repetition of three phases including denaturation,
annealing, and extension (Oste 1988). Using microsatellite DNA markers, a speciﬁc
target section of DNA can be isolated (Parker et al. 1998). Microsatellite markers are
composed of forward and reverse oligonucleotide primers that are coded to attach to a
speciﬁc sequence of DNA. Microsatellite markers can be created for a single locus by
scanning the genome and identifying sections of tandem repeats. Once oligonucleotide
primers are created and used in conjunction with PCR, they can be used to isolate and
amplify known loci. The PCR product can then be run on acrylamide gels with
electrophoresis and allele frequencies determined (Parker et al. 1998). The use of this
technique requires that oligonucleotide primers be developed for the speciﬁc species of
interest (Morin and Woodruff 1996). For the genetic differentiation and conservation
genetic study of bears, at least ten forward and reverse primers have been developed
previously (Paetkau and Strobeck 1994; Paetkau et al. 1995).

A pilot study conducted by Scribner and Libants (2001 unpublished data) on

Michigan bears, found that ﬁve polymorphic loci (GlOX, GIOM, GlOD, G10L, GlOB;

Paetkau et al. 1995) would be sufﬁcient in differentiating between individual black bears
with an overall probability of identity of 9.96E-09. Additionally, they analyzed 26 hair
samples and were successful at obtaining genotypes at ﬁve loci in 22 of the hair samples.

When this method was ﬁrst described, it was proclaimed to be an adequate
alternative DNA source to blood or tissue (Taberlet et al. 2001). Subsequently, numerous
studies have identiﬁed that there are many errors associated with using hair samples as a
DNA source including: 1) contamination or errors made during collection and lab
analysis, 2) allelic dropout, 3) null or false alleles, and 4) shadow bands causing scoring
errors. Errors associated with the ﬁeld collection and lab analysis can be generated by
mislabeling samples, cross-contamination of multiple samples, and loading errors
(Gagneux et al. 1997). Errors can be minimized by deﬁning precise study protocols for
ﬁeld collection and by careful lab techniques to minimize error (Taberlet et al. 1996).
Additional contamination can occur when more than one individual deposits hair at a
single collection site (Gagneux et al. 1997). Contamination can be detected by the
presence of more than two alleles at a Single diallelic locus. Allelic dropout and null
alleles are errors that occur when only one of the two alleles is ampliﬁed and detected for
a single diallelic locus (Gerloff et al. 1995; Taberlet et al. 1996; Taberlet et al. 1997;
Gagneux et al. 1997; Taberlet et al. 2001). Occurrence of allelic dropout increases when
using small quantities of degraded DNA. This is of concern when using hair or feces as a
DNA source, because the quantity of DNA can often be in the picogram range (Taberlet
et al. 1996). Error arises when an individual is scored as a homozygote because only one
allele is ampliﬁed. Taberlet et al. (1996) suggests regenotyping all homozygous

individuals to quantify and obtain the real genotype for individuals. Additionally,

detections of allelic dropout errors can be made by examining data for a deﬁciency of
heterozygotes (Brookﬁeld 1996). Simulation of how allelic dropout and null alleles
affect genotypes can be a useful means of estimating the magnitude of error on the
outcome of population estimates (Taberlet et al. 2001). Lastly, shadow bands can cause
scoring errors when genotyping individuals. This can be minimized by having more than
one person score gels and by re-analysis of any disputed or inconclusive scores (Gagneux
et al. 1997).

Although the analysis of hair samples for DNA ﬁngerprinting contains potential
errors, errors can be accounted for by using good lab and ﬁeld practices, designed
experiments to detect errors, and simulation to determine how errors affect the population
estimate. By recognizing and accounting for errors in the study design phase, accurate
estimates of population size are attainable using non-invasive techniques (Waits and

Leburg 2000).

Population Estimate and Analysis

One assumption of using mark-recapture as a means to estimate population size is
that individuals are identiﬁed correctly (White et al. 1982). Using the information of
allele frequencies from a sample of individuals it is possible to calculate a probability of
identity (PI) for each locus examined in the lab. This is given as the probability of two
individuals sampled in a population having the exact same genotype. When analyzing
several loci for a population the product of the PI for each locus gives an overall PI for
the population (Paetkau et al. 1995). The PI is dependent on the amount of variation in

the population. One important factor affecting PI is effective population size and in

10

populations of small size, allelic diversity and heterozygosity will be low (Frankham
1996). As such, more loci may be required to obtain a PI that is sufﬁciently small
enough to differentiate individuals (Paetkau and Strobeck 1994).

Once genotypes have been determined for individual animals they can be used in
a mark-recapture design to estimate population size. Woods et al. (1999), Mowat and
Strobeck (2000), Poole et al. (2001), and Boerson et al. (2003) utilized genetic mark
recapture techniques to obtain population estimates of brown bears in Canada and black
bears in Louisiana. Palsboll et al. (1997) used molecular markers in a mark-recapture
study of hump-backed whales (Megaptera novaeangliae) in the Atlantic Ocean using
skin biopsies and sloughed skin as DNA sources. Genetic markers free from errors
provide an ideal means to mark individuals because they satisfy the marking assumptions
that animals do not loose their marks during the experiment, and marks are noted
correctly on each sampling occasion (White et al. 1982). Genetic methods have no
impact on survival, and marks can be identiﬁed and reported correctly upon recapture
(Woods et al. 1999; White et al. 1982). Additionally, a non-invasive study design allows
for greater sample size required for mark-recapture models and genotypes are permanent
and cannot be lost over time. Because animals are not captured and handled, study
designs can be established to explicitly address spatial issues necessary to meet model
assumptions including population closure (Mowat and Strobeck 2000). One
disadvantage of the technique is the possible under-representation of individuals that visit
a site, but do not deposit hair. Additionally, bears may have an aversion to barbed wire

based on previous negative experience (Woods et al. 1999). Because of the required lab

11

analysis, computation of a population estimate may take longer than with other
conventional capture-recapture techniques.

In the general use of capture-recapture experiments it is generally recommended
that the method used to mark individuals is different from the method used to recapture
individuals to minimize bias (White and Schenk 2001). Utilizing the ability to capture
tissue samples from harvested bears in Michigan provides a means to recapture
individuals using a different method of capture.

Program MARK (White and Bumham 1999) encompasses all current methods to
analyze mark-recapture data. Many different models are available in program MARK,
each with different assumptions and uses, but closed capture models are most commonly
used with non-invasive methodology (Woods et al. 1999; Mowat and Strobeck 2000;
Boerson et al. 2003). For example, closed capture models available in program MARK
are the same models that were in the older program CAPTURE (White et al 1982).
Program MARK offers models that can incorporate individual heterogeneity, but assume
population closure. Program MARK uses multiple competing models to determine the
best ﬁt of the data using Akaike Information Criterion (AIC). These models are ranked
on their AIC score with the lowest score being the best ﬁtting model (Bumham and
Anderson 1998). These aspects make program MARK the most comprehensive software

to analyze mark-recapture data (Cooch 1999).

OBJECTIVES

The main focus of this research is to develop a methodology that the MDNR can

use to estimate the size of the black bear population in the NLP of Michigan. However,

12

non-invasive sampling techniques to estimate population abundance are relatively new
and the effects of genotyping errors and sub-sampling are not well established in the
literature. Therefore, quantiﬁcation and simulation of errors and sub-sampling will both
improve the abundance estimate of black bears in the NLP of Michigan and also
contribute to the profession’s understanding of this technique. The objectives of this
research include:

1.) Perfect a methodology for collecting black bear hair and tissue samples to be
genetically analyzed and used to obtain a mark-recapture based estimate of the
size of the black bear population in the NLP of Michigan.

2.) Obtain an estimate, with conﬁdence intervals, of the size of the black bear
population in the NLP of Michigan.

3.) Quantify and simulate the effect of genotyping errors and sub-sampling on
population abundance estimates of black bears in the NLP of Michigan

4.) Make recommendations to the MDNR on use of this methodology to estimate

bear abundance in the NLP of Michigan

THESIS ORGANIZATION

The thesis is organized in ﬁve chapters. Chapter 1, which precedes, is a general
introduction of the methodologies and literature for utilizing the non-invasive technique
to estimate abundance. Chapter 2 includes the ﬁeld methodologies, results and
discussion for data collected in both years 2002 and 2003. Chapter 3 includes the
introduction, methods, results, discussion and management implications for the

population estimate. Chapter 3 only includes data collected in 2003 because data

13

collected in 2002 were considered preliminary and were used to perfect both the ﬁeld
sampling and laboratory techniques. Chapter 4 includes the introduction, methods,
results, and discussion for a simulation model that we created to examine the effects of
genotyping errors and sub-sampling hair samples on population estimates. The fourth
chapter primarily focuses on data collected in 2003, however portions of model
parameterizations were derived from data collected in 2002. Chapter 5 includes
recommendations to the MDNR for the successful use of this technique and encompasses

information collected throughout the duration of this research project.

14

LITERATURE CITED

Bailey, J. A. 1984. Principles of wildlife management. John Wiley & Sons, New York,
New York, USA.

Boerson, M. R., J. D. Clark, and T. L. King. 2003. Estimating black bear population
density and genetic diversity at Tensas River, Louisiana using microsatellite DNA
markers. Wildlife Society Bulletin 31(1): 197—207.

Brookﬁeld, J. F.Y. 1996. A simple new method for estimating null frequency from
heterozygote deﬁciency. Molecular Ecology 5: 453—455.

Bumham, K. P., and D. R. Anderson. 1998. Model selection and inference: A practical
information theoretic approach. Springer, New York, USA.

Cooch, E. 1999. Program MARK: Techniques for monitoring wildlife populations: First
steps with program MARK: Linear models. In: Fields R., editor; Integrating
People and Wildlife for a Sustainable Future: Proceedings of the Second
International Wildlife Management Congress; GOdOllo, Hungary.

CRC Memorandum. 1996. Statewide ballot issues. Citizens Research Council of
Michigan 107: IOpp.

Etter, D. R., L. G. Visser, C. M. Schumacher, E. Calrson, T. Reis, and D. Rabe. 2002.
Black bear population management techniques. Final Report. Federal Aid in
Wildlife Restoration Project W-127-R-17, I8, 19. Michigan Department of
Natural Resources, Lansing, Michigan, USA.

Frankham, R. 1996. Relationship of genetic variation to population size in wildlife.
Conservation Biology 10(6): 1500—1508.

F rawley, B. J. 2001. 2000 Michigan black bear hunter survey. Wildlife Division Report
3334. Michigan Department of Natural Resources, Lansing, Michigan, USA.

Frawley, B. J. 2002. 2001 Michigan black bear hunter survey. Wildlife Division Report
3364. Michigan Department of Natural Resources, Lansing, Michigan, USA.

Gagneux, 1., C. Boesch, and D. S. Woodruff. 1997. Microsatellite scoring errors
associated with noninvasive genotyping based on nuclear DNA ampliﬁed from
shed hair. Molecular Ecology 6: 861—868.

Garshelis, D. L., and W. Snow. 1988. Minnesota black bear population model. User
Manuel, Version 1.1. Minnesota Department of Natural Resources, Unpublished
Report. 19 pp.

15

GarShelis, D. L., and L. G. Visser. 1997. Enumerating megapopulations ofwild bears
with an ingested biomarker. Journal of Wildlife Management 61(2): 466-480.

Gerloff, U., C. Schlotterer, K. Rassmann, I. Rambold, G. Hohmann, B. Fruth, and D.
Tautz. 1995. Ampliﬁcation of hypervariable simple sequence repeats
(microsatellites) from excremental DNA Of wild living bonobos (Pan paniscus).
Molecular Ecology 4: 515-518.

Higuchi, R., C. H. von Beroldingen, G. F. Sensabaugh, and H. A. Erlich. 1988. DNA
tying from single hairs. Nature 7: 543—546.

Hildebrandt, T. D. 1976. Age determination in Michigan black bear by means of
cementum annuli. Michigan Department of Natural Resources, Wildlife Division
Report No. 2758. Lansing, Michigan, USA.

Kohn, M. H., E. C. York, D. A. Kamradt, G. Haught, R. M. Sauvajot, and R. K. Wayne.
1999. Estimating population size by genotyping feces. The Royal Society 266:
657-663. '

Mace, R. D., S. C. Minta, T. L. Manley, and K. E. Anne. 1994. Estimating grizzly bear
population size using camera sightings. Wildlife Society Bulletin 22: 74-83.

Mills, L. S., J. J. Citta, K. P. Lair, M. K. Schwartz, and D. A. Tallmon. 2000. Estimating
animal abundance using noninvasive DNA sampling: promise and pitfalls.
Ecological Applications 10(1): 283-294.

Morin, P. A. and D. S. Woodruff. 1996. Noninvasive genotyping for vertebrate
conservation. Pages 299—313 in Smith, T. B. and R. K. Wayne, editors.

Molecular Genetic Approaches in Conservation. Oxford University Press, Inc.,
New York, New York, USA.

Mowat, G. and C. Strobeck. 2000. Estimating population size of grizzly bears using hair
capture, DNA proﬁling, and mark-recapture analysis. Journal of Wildlife
Management 64(1): 183—193.

Oste, C. 1988. Polymerase chain reaction. Bio Techniques 6(2): 162—167.

Paetkau, D., W. Calvert, I. Stirling, and C. Strobeck. 1995. Microsatellite analysis of
population structure in Canadian polar bears. Molecular Ecology 4: 347—354.

Paetkau, D. and C. Strobeck. I994. Microsatellite analysis of genetic variation in black
bear populations. Molecular Ecology 3: 489—495.

Palsboll, P. J ., J. Allen, M. Berube, P. J. Clapham, T. P. Feddersen, P. S. Hammond, R.
R. Hudson, H. Jorgensen, S. Katona, A. H. Larsen, F. Larsen, J. Lien, D. K.

16

Mattila, J. Sigurjonsson, R. Sears, T. Smith, R. Sponer, P. Stevick, and N. Olen.
1997. Genetic tagging of humpback whales. Nature 388: 767-769.

Parker, P. G., A. A. Snow, M. D. Schug, G. C. Booton, and P. A. F uerst. 1998. What
molecules can tell us about populations: choosing and using a molecular marker.
Ecology 79: 361—382.

Pearce, J. M., R. L. Fields, and K. T. Scribner. 1997. Nest materials as a source of
genetic data for avian ecological studies. Journal of Field Ecology 63(3):
471—481.

Peyton, R. B. 1998. Deﬁning management issues: dogs, hunting and society.
Transactions of the 63rd North American Wildlife and Natural Resource

Conference pgs. 544-555.

Peyton, B., P. Bull, T. Reis, and L. Visser. 2001. Report summary: public views on bear
and bear management inn the lower peninsula of Michigan. Michigan
Department of Natural Resources, Lansing, Michigan, USA.

Poole, K. G., G. Mowat, D. A. Fear. 2001. DNA-based population estimate for grizzly
bears (Ursus arctos) in northeastern British Columbia, Canada. Wildlife Biology
7: 105-115.

Taberlet, P. and J. Bouvet. 1992. Bear conservation genetics. Nature 358: 197.

Taberlet, P., J. J. Camarra, S. Grifﬁn, E. Uhres, O. Hanotte, L. P. Waits, C. Dubois-
Paganon, T. Burke, and J. Bouvet. 1997. Noninvasive tracking of the endangered
Pyrenean brown bear population. Molecular Ecology 6: 869—876.

Taberlet, P., S. Grifﬁn, B. Goossens, S. Questiau, V. Manceau, N. Escaravage, L. P.
Waits, and .I. Bouvet. 1996. Reliable genotyping of samples with very low DNA
quantities using PCR. Nucleic Acids Research 24(16): 3189—3194.

Taberlet, P., G. Luikart, and E. Geffen. 2001. New methods for obtaining and analyzing

genetic data from free-ranging carnivores. Pages 313—334 in Gittleman, .I. L., S.
M. Funk, D. W. MacDonald, and R. K. Wayne, editors. University Press,
Cambridge, UK.

US. Departemnt of Interior, Fish and Wildlife Service, and US. Department of
Commerce, Bureau of the Census. 1998. 1996 National Survey of Fishing,
Hunting, and Wildlife-Associated Recreation — Michigan. FHW/96-MI.

Waits, J. L. and P. L. Leberg. 2000. Biases associated with population estimation using
molecular tagging. Animal Conservation 3: 191—199.

17

White, G. C., D. R. Anderson, K. P. Bumham, and D. L. Otis. 1982. Capture-recapture
and removal methods for sampling closed populations. Los Alamos National
Laboratory, LA 8787-NERP, Los Alamos, New Mexico, USA.

White, G. C. and K. P. Bumham. 1999. Program MARK: Survival estimation from
populations of marked animals. Bird Study Supplement 46: 120—138.

White, G. C. and T. C. Schenk. 2001. Pages 329-350 in J. J. Millspaugh and J. M.
Marzluff, editors. Design and Analysis of Wildlife Radiotelemetry Studies.
Academic Press, San Diego, California, USA.

Woods, J. G., D. Paetkau, D. Lewis, B. N. McLellan, M. Proctor, and C. Strobeck. 1999.

Genetic tagging of free-ranging black and brown bears. Wildlife Society Bulletin
27(3): 616—627.

18

CHAPTER 2 — FIELD SAMPLING METHODOLOGIES

The main focus of this research is to develop a methodology that the MDNR can
use to estimate the size of the black bear population in the NLP of Michigan. To
accomplish this goal it is necessary to develop a methodology for collecting black bear

hair and tissue samples in the NLP of Michigan to be genetically analyzed.

STUDY AREA

The study area is composed of 3 bear management units (BMU’S) in the NLP of
Michigan; Baldwin, Gladwin and Red Oak (Figure 1.1). The Baldwin BMU
encompasses all of Benzie, Grand Traverse, Lake, Manistee, Osceola, and Wexford
counties and that area of Newaygo County north of highway M-20. The Gladwin BMU
encompasses all of Arenac, Clare, Gladwin, Iosco, and Ogemaw counties. The Red Oak
BMU encompasses all of Alcona, Alpena, Antrim, Charlevoix, Cheboygan, Crawford,
Emmet, Kalkaska, Missaukee, Montmorency, Oscoda, Otsego, Presque Isle, and

Roscommon counties. This area is characterized by the Northern Lacustrine-Inﬂuenced
ecotone as described by Albert (1995). The study area is approximately 36,848 km2 with
elevations ranging from 177 to 526 meters. Well drained, sandy soils in the area
typically support Northern hardwoods, jack pine (Pinus banksiana) barrens, upland
conifer forest, white pine (Pinus strobes) -red pine (Pinus resinosa) forest, hardwood-

conifer swamp, and conifer swamp (Albert 1995). The landscape is highly fragmented

with densities, calculated from digital coverages of county roads, streets and highways,

being approximately 1 km/km2 (MDNR-Land and Mineral Services Division, Resource

l9

Mapping and Aerial Photography 1992). Approximately 11,878 km2 (32.24%) of the

study area is under public ownership (MDNR-Forest, Mineral and Fire Management

Division, Resource Mapping and Aerial Photography 2001).

METHODS
Hair Snare Placement

There has been no similar research using hair snares in Michigan to estimate bear
abundance, thus, we were limited in our knowledge of ﬁeld logistics to execute this
research. We developed a stratiﬁed random sampling design for each of the 2002 and
2003 ﬁeld seasons to identify sampling locations while capitalizing on the information
that we possessed. Additional information came from approximately 700 bait index
locations established by the MDNR from 1991—2000. These locations were utilized by
the MDNR to index the bear population in the NLP and could be used as point locations
for hair snare sampling locations. To maximize our probability of snaring hair given the
large study area, potential sites for hair snare placement were identiﬁed in a series of 5
procedures including: 1) selection of land cover/land characteristics likely to provide
bear habitat, 2) determining buffers on roads to both increase the probability of bear visits
and also decrease sabotage by humans, 3) selection of primarily public lands so that the
number of sites was not limited by personnel access, 4) utilizing townships where hunters
have harvested bear (to indicate bear presence), and 5) selection of hair snare locations.
To perform these procedures we used ArcView 3.2 (Environmental Systems Research
Institute, Redlands, California) using several existing digital coverages obtained from the

MDNR: 1993 and 2001 Gap Analysis Program (GAP) Northern Lower Peninsula land

20

cover datasets (MDNR-Wildlife Bureau and MDNR-Land and Mineral Services
Division, Resource Mapping and Aerial Photography 2000; MDNR-Wildlife Bureau and
MDNR-Land and Mineral Services Division, Resource Mapping and Aerial Photography
2003), black bear management units (MDNR-Wildlife Bureau 2002a), Michigan (Lower
Peninsula) GAP Land Stewardship Coverage (MDNR-Forest, Mineral and Fir
Management Division, Resource Mapping and Aerial Photography 2001), MIRIS Base
Data (MDNR-Land and Mineral Services Division, Resource Mapping and Aerial
Photography 1992), and 1997—2002 black bear harvest locations (MDNR-Wildlife
Bureau 2002b).

Land use/land cover characteristics are important to the distribution of bears
because they represent the distribution of food resources (Rogers 1987). Therefore, hair
snares should be placed in areas having land use/land cover characteristics that represent
bear utilization. For the 2002 hair snare placement, we used the 1993 GAP land
stewardship land cover digital coverage and for the 2003 hair snare placement we used
the 2001 GAP land stewardship land cover digital coverage. The deﬁnitions of land
use/land cover classiﬁcations differed between coverages (Appendix 2.1 and 2.2). We
selected the following land cover/land use categories in both years: aspen (Populus
tremuloides) / birch (Betula papyrifera), emergent wetland vegetation (emergent
wetland/wet meadow, lowland broad-leaved deciduous shrub, lowland broad-leaved
evergreen shrub, lowland broad-leaved evergreen shrub, other lowland shrub), and mixed
lowland (mixed lowland conifer/hardwood, mixed lowland hardwood) (L. Visser,
MDNR, person. comm. 2002). By selecting these vegetation types, we isolated areas

with vegetation types likely to provide bear habitat.

21

Although, black bears in the NLP of Michigan show little avoidance of roads (L.
Visser, MDNR, person. comm. 2002), we chose to set snares an arbitrary distance away
from roads to minimize the occurrence of human tampering with snares. In 2002 we
placed a buffer of 100 meters around all county roads, and a buffer of 1,000 meters was
placed around highways and streets. In the summer of 2003 we placed a 500 meter
buffer on all streets, county roads, and highways and removed this area from the bear
land use/land cover vegetation map. Because access of ﬁeld personnel was limited to
public land in most cases, we also removed inaccessible private land areas from the bear
land use/ land cover vegetation map.

We obtained the locations of black bear harvested from 1997—2001 and grouped

these harvests by township (93 kmz). We then classiﬁed each township based on the

number of bears harvested from 1997-2001. In 2002, categories were based on natural
breaks in the data and included: 1—2 harvests, 3—5 harvests, 6—13 harvests, 14—25
harvests, and 26—48 harvests. These categories were used in the 2002 selection of hair
snares. Before the 2003 ﬁeld season, harvest locations were updated to reﬂect the
harvests that occurred in 2002. Therefore categories for 2003 reﬂect the number of
harvests per township from 1997—2002 and were also based on natural breaks in the data
and included: 1-2 harvests, 3—10 harvests, 11—20 harvests, 21-35 harvests, and 36—59
harvests.

To create the ﬁnal map that was needed to identify hair snare locations, we
overlaid the townships in which bear harvests had occurred onto the map depicting
accessible, potentially suitable vegetation types that were > 500 meters from county

roads, streets and highways.

22

We then used this ﬁnal map to choose hair snare locations and only considered
townships that contained bear harvests from 1997-2001 for 2002 and 1997-2002 for
2003. Table 2.1, contains the frequency of harvest values per township and the resulting
number of snare locations selected in 2002 and 2003 that were based on the number of
snares that ﬁeld personnel could operate.

Table 2.1. Categories for the number of bears harvested per township for 1997-2001
and 1997-2002 and the resulting number of hair snare locations chosen.

 

Number of harvests per Number of harvests per Number of hair snare
Township 1997-2001 Township 1997-2002 locations chosen

0 0 0
1-2 1-2 0
3-5 3-10 1

6-13 11-20 2
14-25 21-35 3
26-48 36-59 4

 

After all the stratiﬁcations were completed and the number of snares for each
township determined, we were ready to choose locations. We selected locations for snare
placement in a series of two steps. First, we wanted to utilize MDNR bait station
locations where possible, so MDNR bait index sites were then overlaid onto the
townships and sites were selected meeting our criteria for the number of snares per
township based on previous visitation history and proximity to other hair snare locations.
Second, in the townships where MDNR bait index sites were not available, a random
script was used in ArcView 3.2 whereby points meeting our criteria for the number of
snares per township and a minimum spacing distance of 5 km between sites were

randomly selected within our ﬁnal coverage.

23

For the summer of 2003, we chose to keep the same snaring locations that we
used in the summer of 2002. However, we added a number of snares to reﬂect the new

harvest categories updated from harvests that occurred in 2002.

Veriﬁcation of Sampling Design

Etter et al. (2002) conducted a research experiment from 1991—2000 to examine
population dynamics and movements of bear in the NLP. Researchers radio-marked 126
bears (64 males, 62 females) varying from 1 year of age to 19 years of age for females
and 1 year of age to 9 years of age for males. Telemetry locations were taken from a
ﬁxed wing aircraft to the nearest quarter, quarter section using a GPS. Locations were
also taken from the ground by triangulation and point locations derived from a minimum
of 2 radio-bearings and estimated using LOCATE II (Nams 1990). Locations with error
> 16 ha were deleted from analysis. A total of 4,873 locations (2,999 locations for
females and 1,874 locations for males) were taken from 5/15/1991 — 11/19/2000.

We used radio telemetry locations to verify stratiﬁcations and snare placement by
examining the distribution of locations on private and public land, proximity of telemetry
locations to streets, roads and highways, land use/land cover characteristics of the
telemetry locations, and the overlap between the telemetry locations and the 2003 snare
locations. Telemetry locations used in this analysis include all telemetry locations and a
subset of locations collected during the months of June and July (i.e. months that we
snared hair) for all years of the study.

To examine bear usage of both private and public lands, we overlaid all the

telemetry locations and the subset of telemetry locations on the public land coverage and

24

determined the number of telemetry locations that were taken on public land, and
calculated the percentage of telemetry locations on public land. To examine the
proximity of all telemetry locations and the subset of telemetry locations to roads, streets
and highways, we overlaid the map with the SOO-meter road buffer on all county roads,
streets and highways with the telemetry locations. We then determined the percentage of
telemetry locations within 500 meters of county roads, streets and highways.

To associate and summarize the different land use/land cover characteristics with
the telemetry locations we overlaid the telemetry locations with the 1993 land use/land
cover digital coverage map (MDNR-Wildlife Bureau and MDNR-Land and Mineral
Services Division, Resource Mapping and Aerial Photography 2000) and associated the
vegetation type with the telemetry location. We then summarized the percentage of
telemetry locations in each vegetation classiﬁcation.

Lastly, to examine the area sampled compared to the telemetry locations, we ﬁrst

buffered all snare locations in 2003 to include the size of a female home range (131 kmz;

Clark and Smith 1994). We then overlaid the telemetry locations onto the buffered area

and calculated the percentage of telemetry locations within the buffered area.

Field Location of Hair Snare Sites

After snaring locations were selected, each snare was given a unique
identiﬁcation name and UTM location. Snares were setup and operated by ﬁeld
personnel from Michigan State University, numerous Department of Natural Resources
ﬁeld offices, United States Forest Service, Little Traverse Bay Bands of Odawa Indians,

Little River Band of Ottawa Indians and the Grand Traverse Band of Ottawa and

25

Chippewa Indians. We held training sessions before the 2002 and 2003 ﬁeld seasons to
ensure consistency of ﬁeld techniques.

To assist the ﬁeld crew in random site location in the ﬁeld, we used Topo USA
3.0 (DeLorme©, Yarmouth, ME, USA) with Gannin III© GPS (Gannin International,
Olathe, KS, USA) units. Field personnel were instructed to drive as close to the random
location as possible and then set out on foot with gear to ﬁnd a suitable snare location.
Criteria for site selection included, upland/lowland edges and features that channel bear
travel such as streams or rivers, edges of lakes, old logging roads, power line or pipeline
right-of-ways, well used deer runways, and upland ridges through swamps (L. Visser,

MDNR, person. comm. 2002).

Hair Snare Conﬁguration

In the summer of 2002 the hair snare conﬁguration used was consistent with the
design used by Woods et. al. (1999), Mowat and Strobeck (2000), Polle et al. (2001), and
Boerson et al. (2003) where a single strand of barbed wire was used to form an enclosure
around a suspended bait (Figure 2.1). In the summer of 2003 we used a snare
conﬁguration that included the use of two separate strands of barbed wire, one placed at a
uniform height of 50 cm and the second placed at a uniform height of 20 cm (Figure 2.2;
Eason et al. 2001). For both summers we used two-stranded barbed wire with four
pronged barbs spaced every 13 cm.

After a general area was selected, ﬁeld personnel were instructed to locate a
suitable grouping of trees. Groupings were to contain 2 3 trees that were Spaced

approximately 4 m apart. If a site was suitable, but lacked a suitable conﬁguration of

26

trees on which to attach the wire, metal fence posts were provided. Trees were large
enough to attach barbed wire and preferably had smooth bark to assist with identiﬁcation
of bear claw marks. Additionally, over-hanging tree limbs were necessary to suspend
baits and the ground needed to be relatively level to ensure the height of the barbed wire
was consistent.

Starting at l of the selected trees, the barbed wire was fastened using a metal
fencing staple. The wire was then hand stretched and tightened to the next tree at a
height of 50 cm. In the summer of 2003 an additional wire was fastened and hand
stretched around all trees at a height of 20 cm. This procedure was continued until an
enclosure was formed (Figures 2.1 and 2.2). Ground irregularities, such as depressions,
were ﬁlled in with woody debris to ensure that the distance between the wire and the
ground remained consistent.

Baits were suspended from trees on opposite sides of the enclosure using twine
and raised to a height of approximately 2—4 meters above the ground (Figures 2.1 and
2.2). Baits used in the summer of 2002 included: one-pound of bacon, liquid scents
(anise, bacon, shellﬁsh, hickory smoke; Bear Scents LLC, Lake Mills, WI, USA),
chicken legs, and sausage links. In the summer of 2003 all baits were consistent
including one-pound bacon and anise extract. Anise extract was contained in a ﬁlm
canisters with holes cut in the sides and stuffed with cotton balls. To warn humans of the
presence of the hair snare, warning signs were placed on 2 trees on either side of the hair
snare site and ﬂorescent ﬂagging was tied to each section of wire at multiple locations.
We recorded UTM coordinates, names of the setup crew, date, estimated time of setup,

and a rough sketch of the hair snare from an aerial perspective.

27

Fluorescent
Flagging

Warning
Sign

 

 

 

Figure 2.1. Single wired hair snare for black bear hair collection comprised of a
single strand of barbed wire (13 cm barb spacing) encircling 2 3 trees at 50 cm from
the ground. Spacing between trees is approximately 2-4 meters apart. Two
warning signs and ﬂorescent ﬂagging tied around the wire warn humans of the
snare. Bait is suspended over the snare by strings attached to over-hanging limbs.

28

Fluorescent
Flagging

 

. ‘ 20 cm 50 cm - . » '
4 m
Figure 2.2. Double wired hair snare used for black bear hair collection comprised
01'2 strands of barbed wire (13 cm barb spacing) encircling 2 3 trees at 20 cm and
50 cm from the ground. Spacing between trees is approximately 4 meters apart.
Two warning signs and florescent ﬂagging tied around the wire warn humans of the
snare. Bait is suspended over the snare by strings attached to over-hanging limbs.

29

Checking Methods

In the summer of 2002, 2 sampling periods consisting of 4 checks in each period
were conducted from May 15 — August 15, 2002. The duration between checks was 5—8
days. Upon the fourth check of the ﬁrst period (checking occasions 1—4) baits were
removed, but the snares remained. The snare was leﬁ without bait for one week and then
revisited to replace bait and clean the wire of any hair that may have been deposited
during the one week pause. Snares were then checked 4 additional times making up the
second period (checking occasions 5—8).

In the summer of 2003 we conducted sampling in 1 period consisting of 5 checks
from June 22 — July 26, 2003. These dates were chosen by examining the number of bear
visitations per week and number of hair samples collected in the summer of 2002
(Figures 2.12 and 2.14). The duration between checks was 5—8 days.

When checking snares, ﬁeld personnel evaluated whether a visitation by a bear
occurred by noting claw marks on trees, absence of bait, or hair on wire. Starting at the
northernmost tree or post each barb of the wire was checked by passing a white index
card behind a barb and looking for the presence of hair. Once hair was found, the
location was documented on the sketch, a height measurement from the ground was
recorded and the hair was removed using forceps. Hair from each independent barb was
considered a single sample and placed in a pro-numbered paper coin envelope. Hair
samples were stored in a dry location awaiting laboratory analysis. We replenished
missing baits or replaced them every two weeks if still present upon checking occasion.

In the summer of 2003, hair samples were only collected from the top strand of

barbed wire and hair samples present on the lower strand of wire were noted, but not

30

collected. In addition, ﬁeld personnel estimated the number of hairs present in each hair
sample and assigned them into categories: 1—4, 5-10, 11—15 and >15 hairs. We also
numerically documented the position of the hair samples on the wire by assigning a
number to each barb of the snare by counting barbs from the northern most tree and

proceeding clockwise.

Hunter Harvested Bear Hair and Tissue Collection

In the fall of both 2002 and 2003 tissue, hair, and tooth samples were collected
from harvested bear for use as a recapture in the mark-recapture methodology. To the
best of our knowledge, this recapture approach is unique and takes advantage of the
mandatory bear harvest registration policy in place in Michigan. To prepare for sample
collection, a training session was held for all MDNR employees who would potentially
register bears. Each MDNR check station was supplied with written collection protocols
and collection kits. The collection kit consisted of a 1.5 mL vile containing tissue buffer
solution (Tris, ECTA, Urea, Sarcocine, NaCl, water) and two small coin envelopes for
tooth and hair collection. All collection kits, including envelopes and the vile, were pre-
numbered to ensure that samples collected from an individual bear were identiﬁable.
Field personnel were instructed to collect a small muscle tissue sample (approximately
1cm by 1cm) and submerge the muscle tissue into the tissue buffer solution. Teeth were
extracted and placed in a small coin envelope for subsequent aging, and hair samples
were plucked so that >10 hair follicles were evident and also placed in a paper coin
envelope. After collection both hair and tissue samples were frozen pending laboratory

analysis.

31

RESULTS
Hair Snare Placement

Because of the large study area required for this project, we chose to use a
stratiﬁed random design to choose the locations of hair snares. Results from the snare
location process will be presented in terms of land area and how this land area was
reduced for the stratiﬁcation of the study area. For brevity, only the stratiﬁcation results
will be covered from the 2003 ﬁeld season. However the results are very similar to these

derived in 2002, only differing in land area values.

The overall size of the study area is 36,848 km2 (Figure 1.1). The area of public

land within the study area is 11,878 km2 (Figure 2.3). The area of land use/land cover

characteristics that we selected as potentially suitable bear habitat is approximately 5,127

km2 (Figure 2.4). The area included within the 500 m road buffer around county roads,

streets and highways were approximately 24,700 km2 (Figure 2.5). The ﬁnal area that

was public land, potentially suitable bear habitat, area > 500 meters from county roads,

streets and highways was 1,420 km2 (Figure 2.6). This technique allowed us to narrow

down the possible area to place hair snares when working with a very large study area.
Additionally, we only choose to place hair snares in townships where bear
harvests had occurred from 1997—2002. There were 464 possible townships in the study
area from which bears could be harvested. There were 225 different townships where
harvests occurred from 1997—2002 and the frequency of harvests for each township also
varied (Table 2.2 and Figure 2.8). Figure 2.7 depicts the distribution of harvest from

1997-2001 that was used to choose hair snare locations for the 2002 ﬁeld season.

32

In some cases we were unable to meet our criteria for the number of snares per
township because access was limited by private land. In addition, some townships did
not have public land, and in these cases we attempted to access private lands.

Table 2.2. Number of harvests per township between 1997-2002 and the frequency
of townships in each category.

 

 

Number of harvests per Number of
Township 1997—2002 Townships
O 239
1-2 71
3-10 109
11-20 34
21-35 8
36-59 3

 

Veriﬁcation of Sampling Design

The percentage of telemetry locations on public land was 63.82% (3,1 10 out of
4,873 telemetry locations) and 64.52% (913 out of 1,415 telemetry locations) from the
subset of telemetry locations. The percentage of telemetry locations within 500 meter of
county roads, street and highways was approximately 49.85% (2,429 out of 4,873
telemetry locations) and 54.28% (768 out of 1,415 telemetry locations) from the subset of
telemetry locations collected in June and July.

Land use/land cover classiﬁcations of the telemetry locations cover almost all
possible categories for all of the telemetry locations and the subset of telemetry locations
(Tables 2.3 and 2.4). The category with the most telemetry locations included mixed
lowland conifer/hardwood with 22.19% of the telemetry locations and 22.06% of the
telemetry locations for all telemetry locations and the subset, respectively. For the

stratiﬁed random design we selected: aspen/birch, emergent wetland vegetation

33

 

- Public Land
E Study Area '

   
  

 

40 0 40 80 Kilometers

 

 

 

Figure 2.3. Public land ownership (11,878 kmz) within the study area.

34

 

Potential Bear Habitat
Study Area

 
  
 
 
 
 
 
 
 
  

Vegetation Classes:
Lowland Deciduous Forest
Lowlmd Coniferous Forest
Lowlmd Mixed Forest
Floding Aquatic

Lowlatd Shrub

Emergent Wetland

Mixed Non—forest Wetland
AspenIBirch

 

 

 

Figure 2.4. Potential bear habitat within the study area comprised of emergent
wetland vegetation (emergent wetland/wet meadow, lowland broad-leaved
deciduous shrub, lowland broad-leaved evergreen shrub, lowland broad-leaved
evergreen shrub, other lowland shrub), mixed lowland (mixed lowland
conifer/hardwood, mixed lowland hardwood), and aspen/birch vegetation
categories.

35

 

 

E Study Area
- Road Buffer

   
 

 

 

 

Figure 2.5. The 500 meter buffer on both sides of county roads, streets and highways
within the study area.

36

 

 

 

- Public Bear Habitat More Than 500 m From Roads
[:lStudyArea
. Jr‘ I! N
.. _ s i
r .. a {55'-
tt J 5"“ 5* (if r" .
3".- " '4‘... n... .
\ s.‘ . tiling in.“
Hi ‘ "1.147;: - a .
1 ".‘.- a. I
‘ ‘ ‘l-‘_ "n ‘q. .
. '. I". ‘3 ‘ \ '6'; 2' -‘ r.
‘ ‘.' -'-\.:"- 1.,-I “"331; . ' ‘
* .,. .‘ -‘=’ '32-; .« an? .. *shg-‘izv?
' It? " -‘ , as: $6.1.
- . Ia. *uf‘gﬁh," -. ' _.'.._ _"
are; "-521” A in. ‘ Sb-
'0 ‘.. 4. ,7 w. h * .._" ,, S
«was ..: .-. -. . . ' ..
’ '.«.o-:. -. '. xt- * m w
\ ‘qg: ‘ . ' ' a . .
:1 "a“: . -' _‘;7‘; I afzf s " :1":
I. '.'g- '5 ' If" _
. ‘ . ‘ 1‘ v8
2" ‘ new - -
sf:
art"
I. ."f:
If ‘- v-
50 0 50 100 Kilometers

 

 

 

 

 

 

Figure 2.6. Final stratiﬁed coverage representing public land, potential bear habitat,

area > 500 meters from streets, county roads, and highways. This coverage was
used for the hair snare selection.

37

 

[:I Study Area

   
    
 

\-

 

50 o 50 100 Kilometers

4

  
  

 
   

. T‘" I
r+~ —

rm

 

 

 

Figure 2.7. Number of bears harvested in each township within the study area from
1997-2001.

38

 

Study Area
1997-2002 Bear Harvest by Township
1 - 2
3 - 10

  
  
  
   

 

      
  

' A

I IIIL

.. .n
i =II.

50 0 50 LOO Kilometers -.“...
re:-

 

 

 

Figure 2.8. Number of bears harvested in each township within the study area from
1997—2002.

39

Table 2.3. Land use /land cover categories, number of telemetry locations and
percent of all telemetry locations collected from 1991-2001.

 

No. of Percent of

Land use/cover Category Locations Locations

High intensity urban 28 0.57%
Low intensity urban 1 0.02%
Extractive - open pit mining O 0.00%
Agricultural crops 278 5.70%
Orchards / vineyards O 0.00%
Herbaceous open land 281 5.77%
Shrubland 82 1 .68%
Other broad-leaved deciduous forest 12 0.25%
Northern hardwood 846 17.36%
Northern hardwood / conifer 64 1.31%
Aspen / birch“ 996 20.44%
Oak 166 3.41%
Oak / jack pine 6 0.12%
Other conifer forest 20 0.41%
White pine 0 0.00%
Red pine 261 5.36%
Upland jack pine 324 6.65%
Cedar / spruce / ﬁr 2 0.04%
Emergent wetland / wet meadow” 59 1.21%
Other lowland shrub“ 15 0.31%
Lowland broad-leaved deciduous shrub“ 236 4.84%
Lowland broad-leaved evergreen shrub“ 5 0.10%
Other forested wetland 2 0.04%
Mixed lowland hardwood“ 11 0.23%
Lowland jack pine 1 0.02%
Black spruce 23 0.47%
Northern white cedar 0 0.00%
Mixed lowland conifer / hardwood” 1075 22.06%
Barren land 0 0.00%
Water 70 1.44%
Urban grasslands 4 0.08%
Lowland needle-leaved evergreen shrub 5 0.10%
Total 4873 100.00%

 

* - Denotes Vegetation Category selected in stratiﬁed design

40

Table 2.4. Land use lland cover categories, number of telemetry locations and
percent of the subset of telemetry locations (June and July) collected from 1991-
2001.

 

No. 01' Percent of
Locations Locations

Land use/cover category

High intensity urban 8 0.57%
Low intensity urban 1 0.07%
Extractive - open pit mining 0 0.00%
Agricultural crops 80 5.65%
Orchards / vineyards 0 0.00%
Herbaceous open land 87 6.15%
Shrubland 29 2.05%
Other broad-leaved deciduous forest 4 0.28%
Northern hardwood 243 17.17%
Northern hardwood / conifer 22 1.55%
Aspen / birch" 277 19.58%
Oak 50 3.53%
Oak / jack pine 3 0.21%
Other conifer forest 5 0.35%
White pine 0 0.00%
Red pine 83 5.87%
Upland jack pine 96 6.78%
Cedar / spruce / ﬁr 2 0.14%
Emergent wetland / wet meadow* 14 0.99%
Other lowland shrub" 3 0.21%
Lowland broad-leaved deciduous shrub" 56 3.96%
Lowland broad-leaved evergreen shrub" 0 0.00%
Other forested wetland 1 0.07%
Mixed lowland hardwood* 1 0.07%
Lowland jack pine 0 0.00%
Black spruce 6 0.42%
Northern white cedar O 0.00%
Mixed lowland conifer / hardwood“ 314 22.19%
Barren land 0 0.00%
Water 25 1.77%
Urban grasslands 2 0.14%
Lowland needle-leaved evergreen shrub 3 0.21%
Total 1415 100.00%

 

"‘ - Denotes Vegetation category selected in stratiﬁed design

41

(emergent wetland/wet meadow, lowland broad-leaved deciduous shrub, lowland broad-
leaved evergreen shrub, lowland broad-leaved evergreen shrub, other lowland shrub), and
mixed lowland (mixed lowland conifer/hardwood, mixed lowland hardwood). The
percentage of telemetry locations in these categories include 47.00% and 49.19% of the
telemetry locations for all telemetry locations and the subset, respectively.

By buffering all 239 hair snares set in 2003 by the radius of an adult female home

range (6.46 km) we made the equivalent of the area of an adult female home range (131
kmz; Figure 2.9). The total area of this buffer is approximately 18,233 km2 (Figure 2.9).

The percentage of telemetry locations within this buffered area was 87.83% (4,280 out of
4,873 telemetry locations) and 87.49% (1,238 out of 1,415 telemetry locations) from the

subset of telemetry locations.

Hair Snaring

The total number of hair snares set, the number of checking occasions, and
consequently the number of checks of hair snares differed between 2002 and 2003 (Table
2.5).
Table 2.5. Number of hair snares set, number of checking occasions, total number
of checks made, number of snares that were visited by bears, total number of hair

samples collected, and the percent of snares visited for the summers of 2002 and
2003.

 

No. Snares No. Checking Total No. No. Snares Percent

 

Year Set Occasions Checks Visited Visited
2002 202 8 1,616 1 18 58.42%
2003 239 5 1,195 122 51.05%

 

42

 

  

Buffered 2003 Hair Snares
l: Study Area

 
     
       

 

 

  
  

    

50 0 50 100 Kilometers “In-I

1 III-II!
III-Ir
IIIUIV

 

 

 

 

Figure 2.9. Buffer based on average size of adult female annual home range (131
kmz) around the 239 hair snares set in 2003.

43

 

 

. Snares Visitedby Bears .
o Snares Not Visitedby Bears .' . N
A Mislocated Snares 0 .° .
:jStudyArea o'o °
00.0 o. o
O
. O
O
. e . e '0 ‘
. O . .0 I. 0 O o O.
O O .
e O . e.
0"}, o .0 e
O C C . .0 O a
O 0 e
o O O 0 0 . x ..O 0
O . . o
o 8 ° 0 . 0 .ﬁ.
0 . .0. A O.
I . O O
. .0 o .0 . o A o . e
0 O O . 9 . e
0.0 . O A . O O
o . 0
o . 9
oo .‘ o A 0Q) o
o (93 O . 9.. 0
0 <8 0 '0
O
O
00 o
O
O
50 0 50
r

 

 

 

 

 

 

 

Figure 2.10. Summer of 2002 hair snare locations. Black dots refer to hair snares
that were visited by bears and unﬁlled dots refer to snares that were not visited by
bears. Documented locations of 5 snares, symbolized by a triangle and referred to
as mislocated, were incorrect and the locations of these snares was approximated.

44

 

 

 

 

 

 

. Snares VisitedbyBear
o Snares Not VisitedbyBear o 0.
[:IStudyArea M.,... N
o..0 0.0 e
ec‘o 0° 8
0 0° 0 I
O. .0. ..
so . o 00'. O .0. ‘
o 00 (T)
O o o. o.
o 000. 0&0. .
00 o
o e o . :0. ‘
coo ' C5
5 0 .0 00’ x .0 o
o. o 0
O o O 0 O {00"0..
o . o o. ..o e 00
0 ° 0
o O o o O. O o - o o 0
o. .000 O0 . o
o ...o .0 o .
.0
o o 0
o o O .
43 00 e. o o 0 .
e O 00 . 0.. no 0
08 o o .0
o o
o
o
. o
O
o
o o
o
O
50 O 50
l

 

 

 

 

 

 

 

Figure 2 11 Summer of 2003 hair snare locations. Black dots represent the snares
that were visited by bears and the unﬁlled black dots represent snares that were not
visited by bears.

45

60

Number of Snares Visited

 

50“

40«

30‘

20w

 

 

 

6/2- 6/9- 6/16- 6/23- 6/30- 7/7- 7/14- 7/21- 7/28— 8/4- 8/11-
6/8 6/15 6/22 6/29 7/6 7/13 7/20 7/27 8/3 8/10 8/17

One-Week Intervals of Sampling in 2002

Figure 2.12. Number of hair snares visited by bears for each week of sampling in
2002. The week of 7/7 — 7/13 represents the l-week interval between sampling

periods.

Number Snares Visited

 

 

 

 

 

6/22-6/28 6/29-7/5 7/6-7/12 7/13-7/19 7/20-7/26
One-Week Intervals of Sampling in 2003

Figure 2.13. Number of hair snares visited by bears for each week of sampling in

2003.

46

250

 

zooww , A 718577“ ,,,,,,,,,,,,1,93,,,L,,-,,,,,,,.

Number of Hair Samples

    

 

 

6/2- 6/9- 6/16— 6/23- 6/30- 7/7- 7/14- 7/21- 7/28- 8/4- 8/11-
6/8 6/15 6/22 6/29 7/6 7/13 7/20 7/27 8/3 8/10 8/17

One-Week Intervals of Sampling in 2002

Figure 2.14. The number of hair samples collected for each one-week of sampling
conducted in the summer of 2003. The week of 7/7 — 7/13 represents the 1-week
interval between sampling periods.

 

Number of Hair Samples

 

 

 

6/22-6/28 6/29—7/5 7/6-7/12 7/13-7/19 7/20-7/26
One-Week Intervals of Sampling in 2003

Figure 2.15. The number of hair samples collected for each one-week of sampling
conducted in the summer of 2003.

47

A complete list of wildlife agencies and MDNR ofﬁces operating hair snares, hair snare
names, year of operation, snare locations, number of bear visitations, number of hair
samples collected, and county is located in Appendix 2.3 and 2.4. The distribution of
snares covered much of the study area in both 2002 and 2003 (Figures 2.10 and 2.11).
The number of snares visited by bears and the percent of snares visited also differed with
the summer of 2003 having more visited hair snares, but the summer of 2002 having a
higher percentage of snares visited (Table 2.5).

Snares were checked on approximately 7-day intervals (min = 3, max = 13 in
2002; min = 4, max = 10 in 2003). The duration of sampling in 2002 was approximately
70 days and the duration of sampling in 2003 was approximately 28 days. The number of
snares visited by bears varied throughout the summer (Figures 2.12 and 2.13), as did the
number of hair samples collected with a decreasing trend in the number of hair samples
collected in 2002 and a slight decreasing trend in 2003 (Figures 2.14 and 2.15). We
collected more hair in 2002 than we did in 2003, but the average number of hair samples
collected in 2002 and 2003 was similar (Table 2.6).

Table 2.6. Total number of hair samples collected, mean, minimum and maximum
number of hair samples collected given a visitation occurred for 2002 and 2003.

 

Total No. Mean No. Min. No. Max. No.
Hair Samples Hair sample/ Hair sample! Hair sample/

 

Year Collected Occasion Occasion Occasion
2002 1,368 4.01 1 24
2003 l ,144 4. 1 1 20

 

In the summer of 2003 we documented the number of hair follicles in each hair

sample and placed them into categories (1—4, 5—10, 11—15, or >15 hair follicles). The

48

number of hair samples in each category differed with 42.83% of the hair samples having

1—4 hair follicles (Table 2.7).

Table 2.7. Number of hair samples collected in 2003 that contained 1-4, 5-10, 11-15,
and >15 hair follicles.

 

1-4 Hair 5—10 Hair 11-15 Hair >15 Hair
Follicles Follicles Follicles Follicles Total

490 250 130 274 1144

Hunter Harvest Tissue and Hair Collection

The duration between the summer hair snaring and the hunter harvest was 35 days
in 2002 and 55 days in 2003. During the 2002 black bear harvest we collected 347 tissue
samples from 351 harvested bears. Additionally, we collected hair samples from nearly
all bears from which we collected tissue samples. The MDNR also submitted 5 different
tissue samples from bears that were killed as nuisance bears or road kills. During the
2003 black bears harvest we collected a total of 414 tissue samples from 418 harvested
bears in the NLP. We were also successful at collecting hair samples from nearly all
harvested bears in 2003. In 2003 we attempted to use the tooth samples as a source for
DNA and collected tooth samples from all but 2 of the 414 harvested bears ﬁ'om which
samples were taken. In both 2002 and 2003 information on the location of the harvest

was submitted by many of the hunters.

49

DISCUSSION
Hair Snare Placement

Design of mark-recapture experiments oﬂen follows a systematic grid design
where grid size is a function of the average female bear summer home range size and
suggests that 4 trapping locations be deﬁned within each grid cell (Mowat and Strobeck
2000; White et al. 1982). This high trap density increases the probability of an animal
encountering a capture location and being captured which increases the precision of the
abundance estimate (White et al. 1982). Therefore, the information required to apply
this technique is an estimate of an adult female summer home range and the ability to

operate a large number of snares. Etter et al. (2002) estimated the annual home range

size of female black bears in the NLP to be 131 kmz. Using the systematic grid design

with our study area size of 36,848 km2 would require approximately 1,125 snares set

across the study area, which was not feasible with available personnel. Research
experiments conducted in British Columbia and Alberta, Canada (Mowat and Strobeck
2000; Poole et al. 2001), Florida (Eason et al. 2001), and Louisiana (Boerson et al. 2003)
utilized the systematic grid design.

There are two primary differences between our research experiment and other
research projects utilizing the systematic grid design. First, the study area that we have
chosen is larger than other such research projects. Second, in most cases, study areas
were made up of landownership that provided researchers with unlimited access across

the entire study area (Mowat and Strobeck 2000; Poole et al. 2001; Boerson et al. 2003),

50

unlike our study area where only approximately one-third (1 1,878 km2 out of 36,848 km2

public land) of the study area provides unlimited access (i.e. public land).

Based on these differences, we had to adapt our study design and we chose a
stratiﬁed random design to select hair snaring locations. Stratiﬁcation of bear habitat
within our study area is biologically relevant because not all land cover/ land use classes
provide the life requisites of bears. Hirsch (1990) found that bears on Drummond Island,
Michigan utilized aspen-birch, coniferous, upland hardwood, and wetland vegetation
types in the spring, summer and fall seasons. In addition, telemetry data collected from
1991—2000 (Etter et al. 2002), suggested that certain land use/ land cover categories (i.e.
aspen/birch and mixed lowland conifer/ hardwood) were used more than others (Tables
2.3 and 2.4).

In our study design we stratiﬁed the study area based on private land because it
would be logistically difﬁcult to sample both private and public lands on a large scale.
Might we expect the sampling of primarily public lands to bias our population estimates?
Veriﬁcation using radio-marked animals would suggest that bears are using both public
and private lands (approximately 64.52% of telemetry locations were on public land from
the subset of telemetry locations). Therefore, we would not expect sampling from public
land to bias population estimates.

The purpose of the road buffer was to prevent the possibility of a Site being
chosen at random in close proximity to major roadways in Michigan. The road buffer
was mostly justiﬁed because we did not want humans vandalizing snares, and keeping the
snares away from roadways would limit the ﬁnding of the snare locations. We might also

expect that there is some biological relevance to a road buffer because bears might

51

exhibit avoidance of roads. Telemetry data collected from 1991—2000 (Etter et al. 2002)
suggested that bears did not show avoidance of roads, as 54.28% of telemetry locations
collected in the months of June and July were within 500 meters of county roads, streets

and highways. In addition, because of the high road densities (approximately 1 km of

2 . . . . .
road from every 1 km ) bears may have little chorce but to utilize areas In close

proximity to roads. In both years of snaring there was only one documented account of a
snare being tampered by people. Therefore we might say that this road buffer was
effective at keeping humans from ﬁnding and vandalizing snares. However, we did not
set this up as an experiment, so inferring that the buffer was the reason for humans not
tampering with snares is inconclusive.

Bears were not equally distributed across this diverse study area. Therefore we
decided to stratify our study area by the townships in which harvests had occurred and
the frequency of harvest from 1997-2002. This assumed that the presence of harvest in a
township indicated the presence of bears within that township. By setting up our snare
allocation hierarchically (i.e. more harvests in a township equates to choosing more
snares), we assumed that frequencies of harvests were predictive indicators of relative
bear density in that township. One therefore might expect that these areas with higher
bear densities may require more snares to adequately sample the bears in the area. In
addition, the population estimate generated will be used for management purposes and
the primary tool that the MDNR uses to manage bears is harvest. Therefore, the
population of bears from which bears were being harvested constituted the population
about which the MDNR must be most informed when making management decisions (i.e.

setting harvest quotas). Radio telemetry data (Etter et al. 2002) suggested that bears were

52

present in areas outside of the townships in which harvests have occurred since
1997—2002. Therefore, a population estimate derived for just the townships where
harvests occurred from 1997—2002 might be a conservative estimate for the size of the
bear population in the NLP. Conservative estimates can be important when managing a
population with harvest, especially when harvest is the highest cause of mortality of bear
in the NLP (Etter et al. 2002).

Ultimately we wanted to know if we were sampling in the correct locations. Once
again telemetry locations collected from 1991—2000 (Etter et al. 2002) provided insight
into this question. Clark and Smith (1994) and Boerson et al. (2003) determined their
effective study area by placing a buffer of an adult female home range on all sampling
locations. Adult female home range size was used because it is smaller than adult male
home range and provides a conservative estimate of the area sampled. To examine the
overlap of telemetry locations with the buffered area we overlaid the telemetry locations
within the buffered area. We found that 87.49% of the telemetry locations collected in
the months of June and July overlapped with this home range buffer. Therefore, we can
conclude that we are sampling in areas that have been utilized by bears from 1991—2000
and this utilization provides evidence that we are sampling in the biologically relevant

locations.

Hair Snaring
Utilization of non-invasive DNA sampling for abundance estimation requires the
success of a ﬁeld methodology for collecting hair samples. Other research experiments

utilizing this methodology have been successful at both collecting hair samples and

53

obtaining population estimates (Mowat and Strobeck 2000; Poole et al. 2001; Eason et al.
2001; Boerson et al. 2003). The basic means to collect hair samples common to all
research experiments, including this study, is a hair snare composed of l (Mowat and
Strobeck 2000; Poole et al. 2001; Boerson et al. 2003) or 2 (Eason et al. 2001) strands of
barbed wire encircling a baited location. Baits utilized by researchers varied from scents
and lures (Mowat and Strobeck 2000; Poole et al. 2001) to pastries, corn and meat scraps
(Boerson et al. 2003; Eason et al. 2002). Our study utilized both bacon and anise scented
lure. Once a snare is baited, researchers then visit the snare at a given interval to check
the wire for the presence of hair. Mowat and Strobeck (2000) and Poole et al. (2001)
checked snares at 10—14 day intervals and this study and Boerson et al. (2003) checked
snares at 7 day intervals. The duration of sampling also varied among studies with Eason
et al. (2002) sampling for 67 days and our research experiment sampling for 55 days in
2003. The total number of hair samples collected and the number of different snares
visited by bears also varied among experiments, with Mowat and Strobeck (2000)
collecting the largest number of hair samples (4,245 hair samples) and Eason et al. (2002)
having the largest percentage of snares visited (86.4%) in the Ocala study area (Table
2.4).

For our project, the main difference between the summers of 2002 and 2003 was
the number of checks (8 checks in 2002 and 5 checks in 2003), the number of strands of
barbed wire (1 strand in 2002 and 2 strands in 2003), and the total number of snares set
(202 in 2002 and 239 in 2003). In 2002 a higher percentage of snares were visited by
bears (58.42%), however, this may be attributed to making 5 checks in 2003 rather than 8

checks in 2002. It does not appear that 2 strands of barbed wire results in the collection

54

of more hair samples because the mean number of hair samples collected given a bear
visitation occurred for each summer is very similar (approximately 4 hair samples). It is
difﬁcult to determine if more individual hairs were collected per sample when using the 2
strands of wire because we did not quantify the number of hairs per sample in the
summer of 2002. In both summers the number of hair samples collected on each
checking occasion increases each week for the ﬁrst three weeks (Figures 2.14 and 2.15),
which would suggest that a minimum number of sampling occasions is 3 weeks to obtain
the a maximum number of visits. In addition, the number of hair samples collected per
week increases to approximately the mid-point of sampling then the number of hair
samples collected per week declines (Figures 2.14 and 2.15). This decline in the number
of hair samples collected could possibly be attributed to bears evading the wire after
encountering the wire on a previous occasion, or adverse weather conditions (i.e. rain,
wind) removing the hair samples from the wire. Based on these data, it appears that the
optimal time of sampling is between mid June and mid to late July (Figures 2.12 and

2.14).

55

 

semis: .383 anew eﬁ

 

$8. _ m N8 8mm NEV— wvwdm SSSSEV .b
$9.3m 3: SN ":2 336m sausage. S 59:22 .328 Sam a5.
$8.2 E Sm Neb— Sm; SecotcEV .3 «Eco; £53.. em A888 .3 Ho .5QO
$8.8 8N 8v NEO— MG SSSSEV .b «ESE .2380 .388 .3 S :85
$8.3 42 E. ”as. Maw assesses S «2.0.... saoo A888 .3 a 83m
$8.8 E _ 8v «Eo— vcnﬁ SSSSEV 5 «ESE .==wm .388 .3 B cemmm
$829. w: Nmm NE: Sm; Secotsﬁw .b «ESE .3293 Em £88 .3 8 :83
$8. _ m m8 mom «no. 8i~ SSSSEV .b 328E .3.8£om3g< .388 .3 Ho :83
$3.3 Nmm m _ m NEV— nmm.» ELEV .b «BE-:00 .325 A _88 .3 .0 Boom
$8.9 RN :8 NE. wowd 88¢. .3 336200 53.5 .888 3890ch 93 326.2
$8.? mm. Em ”Ev. 88% use... .5 «top? .888 338% 93 3302
$2 32 N8 NEV— omm Scoutssw .b 3:33:54 .388 .3 8 comcoom
683$ 6835 83.5 835 05m «can .355 833m «93 .355 can beam .3333—
.eoocom we 62 he .eZ

 

6833 835 no 2.3.3.— 05 can 6833 835. no con—E:— .mo..a:m no cox—Es: .33 no.8 beam .353 he 8327. £93 .35.».
was 35.5 2: ”£122: eons—:53 .32. 82:33 8 .25 m3. mam—5.2: £3.53 £9.38.— ..oﬁe be 2.3. ran—Eam d." 05.9

56

 

Hunter Harvested Bear Tissue Collection

Based on the high success rate (98.86% in 2002 and 99.04% in 2003) of being
able to obtain muscle tissue samples from harvested bears, the MN DR has proven to be
highly effective at sampling harvested bears. Before this research was conducted, we
were unsure if it was necessary to collect muscle tissue samples when tooth samples
could possibly be used as a DNA source, which the MDNR had traditionally collected.
In 2003, there were only 2 tooth samples that were not submitted, therefore it would
appear that ﬁ'om a ﬁeld sampling standpoint (not say anything about whether DNA was
actually in the tooth samples), it may not be necessary to collect muscle tissue samples
and use the tooth samples as a source of DNA.

To date, no other capture-recapture studies utilizing the non-invasive
methodology have used the harvest as an additional capture occasion. Results from this
experiment support that collection of a DNA source from harvested bears is highly

effective means to sample populations.

57

LITERATURE CITED

Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and
Wisconsin: a working map and classification. United States Forest Service, North
Central Forest Experiment Station General Technical Report NC-l78, St.

Paul, Minnesota, USA.

Boerson, M. R., J. D. Clark, and T. L. King. 2003. Estimating black bear population
density and genetic diversity at Tensas River, Louisiana using microsatellite DNA
markers. Wildlife Society Bulletin 31(1): 197-207.

Clark, J. D., and K.G. Smith. 1994. A demographic comparison of two black bear
populations in the interior highlands of Arkansas. Wildlife Society Bulletin 22:
593—603.

Eason, T. H., S. L. Simek, McCowan, J. W., and D. Ziegler. 2002. Statewide assessment
of road impacts on bears in Florida. Research Work Plan. Florida Fish and
Wildlife Conservation Commission, Tallahassee, Florida, USA.

Eason, T. H., S. L. Simek, and D. Ziegler. 2001. Statewide assessment of road impacts
on bears in Florida. Research Work Plan. Florida Fish and Wildlife Conservation
Commission, Tallahassee, Florida, USA.

Etter, D. R., L. G. Visser, C. M. Schumacher, E. Calrson, T. Reis, and D. Rabe. 2002.
Black bear population management techniques. Final Report. Federal Aid in
Wildlife Restoration Project W-127-R-l7, l8, 19. Michigan Department of
Natural Resources, Lansing, Michigan, USA.

Hirsch, J. G. 1990. Black bear habitat utilization and habitat model validation in

Michigan. Masters Thesis, Michigan State University, East Lansing, Michigan,
USA.

Michigan Department of Natural Resources- Wildlife Bureau, Michigan Department of
Natural Resources-Land and Mineral Services Division, Resource Mapping and
Aerial Photography. 2000. Gap Northern Lower Peninsula Land Cover.
Michigan Department of Natural Resources, Lansing, Michigan, USA.

Michigan Department of Natural Resources- Wildlife Bureau, Michigan Department of
Natural Resources-Land and Mineral Services Division, Resource Mapping and
Aerial Photography. 2003. Gap Northern Lower Peninsula Land Cover.
Michigan Department of Natural Resources, Lansing, Michigan, USA.

Michigan Department of Natural Resources-, Mineral and Fire Management Division,
Michigan Department of Natural Resources-Resource Mapping and Aerial
Photography. 200]. Gap Land Stewardship. Michigan Department of Natural
Resources, Lansing, Michigan, USA.

58

Michigan Department of Natural Resources- Land and Mineral Services Division,
Resource Mapping and Aerial Photography. 1992. MIRIS Base Data. Michigan
Department of Natural Resources, Lansing, Michigan, USA.

Michigan Department of Natural Resources- Wildlife Bureau. 2002a. Northern Lower
Peninsula Bear Management Units. Michigan Department of Natural Resources,
Lansing, Michigan, USA.

Michigan Department of Natural Resources- Wildlife Bureau. 2002b. 1997-2002 Black
Bear Harvest. Michigan Department of Natural Resources, Lansing, Michigan,
USA.

Mowat, G. and C. Strobeck. 2000. Estimating population size of grizzly bears using hair
capture, DNA profiling, and mark-recapture analysis. Journal of Wildlife
Management 64(1): 183—193.

Nams, V. O. 1990. Locate lI user’s guide. Pacer, Truro, Nova Scotia.

Poole, K. G., G. Mowat, D. A. Fear. 2001. DNA-based population estimate for grizzly
bears (Ursus arctos) in northeastern British Columbia, Canada. Wildlife Biology
7: 105—115.

Rogers, L. L. 1987. Effects of food supply and kinship on social behavior, movements,
and population growth of black bear in northeastern Minnesota. Wildlife
Monograph 97.

White, G. C., D. R. Anderson, K. P. Bumham, and D. L. Otis. 1982. Capture-recapture
and removal methods for sampling closed populations. Los Alamos National
Laboratory, LA 8787-NERP, Los Alamos, New Mexico, USA.

Woods, J. G., D. Paetkau, D. Lewis, B. N. McLellan, M. Proctor, and C. Strobeck. 1999.

Genetic tagging of free-ranging black and brown bears. Wildlife Society Bulletin
27(3): 616—627.

59

CHAPTER 3 — ESTIMATION OF BLACK BEAR ABUNDANCE IN THE
NORTHERN LOWER PENINSULA OF MICHIGAN USING
MICROSATELLIT E DNA MARKERS

INTRODUCTION

Until recently, the quantitative population management of black (Ursus
americanus) and brown (Ursus arctos) bears has been a difﬁcult task. State and federal
management agencies oﬁen rely upon indirect measures of population abundance through
indices and sex/age of harvested animals (D. Etter, MDNR, person. comm. 2003).
Recent advancements in molecular genetics have provided a means to non-invasively
estimate bear population size for conservation purposes (Boerson et al. 2003; Mowat and
Strobeck 2000; Poole et al. 2001; Taberlet et al. 1997).

These methods could not be timelier for biologists attempting to manage bear
populations with harvest. For many wildlife populations, management is shifting to the
greater involvement of stakeholders and minimizing negative human-wildlife interactions -
(Riley et al. 2002). This stakeholder involvement and public interest often requires
precise data in the decision making process to justify and allocate the resource. Bear
management in the state of Michigan, in particular the northern Lower Peninsula (NLP),
is no exception. The population of bears in the NLP is believed to be increasing and
expanding into areas not occupied by bears in recent history (L.Visser, MDNR, person.
comm. 2002). These population characteristics and recent changes in population
management strategies have resulted in the public desire for precise scientiﬁc
information.

Although non-invasive techniques have demonstrated great success and promise,

they are not without their criticisms. Other researchers have identiﬁed genotyping errors

60

as a potential drawback to using non-invasive methods (Gagneux et al. 1997; Goossens et
al. 1998). Genotyping errors result from the limited quantity and/or quality of DNA
contained in hair samples that is often subjected to diverse environmental conditions
associated with ﬁeld collection (Taberlet and Luikart 1999). Errors result in the
misidentiﬁcation of individuals, which violates the capture-recapture assumptions that
individuals are identiﬁed correctly and that marks are not lost (Stevick et al. 2001; White
et al. 1982). This violation of assumptions results in biased population estimates that
have been demonstrated to overestimate population size (Creel et al. 2003; Waits and
Leburg 2000). Quantiﬁcation of error is an important aspect of any research that utilizes
non-invasive methods because it provides insight into the expected bias (Taberlet et al.
1999). Because of these potential drawbacks, recent statistical methods have been
derived to account for genotyping error within the estimate (Lukacs and Bumham in
press).

In the summer of 2002 we began a research project to estimate black bear
abundance in the NLP of Michigan utilizing non-invasive DNA collection and capture-
recapture methodologies. Although previous research had success using non-invasive
methods, research using non-invasive population estimation for a bear population
managed at a large spatial scale was limited. Because the bear population in the NLP is
managed through harvest, we recognized a unique means to resample the population with
DNA collected from hunter-harvested bears. Our objective was to estimate bear
abundance in the NLP of Michigan while quantifying and accounting for genotyping

error in the population estimate.

61

STUDY AREA

The study area is composed of 3 management units that the Michigan Department
of Natural Resources (MDNR) uses to distribute hunters (Figure 1.1). This area is
bounded by Lake Michigan to the west and north, Lake Huron to the east and north and
increased development and agricultural lands on the south. The study area is

characterized as the northern lacustrine-inﬁuenced ecotone (Albert 1995). The size is

approximately 36,848 km2 with elevations ranging from 177 to 526 meters. Well

drained, sandy soils in the area typically support northern hardwoods, jack pine (Pinus
banksiana) barrens, upland conifer forest, white pine (Pinus strobes) -red pine (Pinus
resinosa) forest, hardwood-conifer swamp, and conifer swamp (Albert 1995). The

landscape is mostly forested, but highly fragmented with densities of county roads, streets

and highways approximately 1 km/km2 (MDNR-Land and Mineral Services Division,

Resource Mapping and Aerial Photography 1992). Approximately 11,878 km2 (32.24%)

of the study area is under public ownership (MDNR-Forest, Mineral and Fire

Management Division, Resource Mapping and Aerial Photography 2001).

METHODS
Sampling Location Identiﬁcation

We used a stratiﬁed random design (Ratti and Garton 1996) to derive sampling
locations in our study area. Our sampling design is unique and experimental because
others utilizing non-invasive marking methodologies have used a systematic grid design

(Mowat and Strobeck 2000; Poole et al 2001; Boerson et al. 2003). We found the

62

systematic grid design impractical because private land in the study area would inhibit
unlimited access and the large study area would require more sampling locations than
feasibly possible to operate. We therefore used a stratiﬁed random design in a series of
ﬁve procedures to identify sampling locations with ArcView 3.2 (Environmental Systems
research Institute, Redlands, California) and several existing digital coverages from the

MDNR spatial database. We ﬁrst stratiﬁed by the frequency of bear harvest grouped by

township (93 kmz) over a 5 year period (1997—2002). This stratiﬁcation was used to

indicate the areas of bear presence because bears are not homogeneously distributed
throughout the area. We divided the frequency of harvest by township into ﬁve
categories which included: 1—2, 3—10, 11—20, 21—35, and 36—59 harvests, respectively.
We allocated snares to each township based on the total number of snares we could
monitor and the frequency of harvest. In townships with 36~59 harvests we allocated 4
snares, 21—35 harvests 3 snares, 11—20 harvests 2 snares, 3—10 harvests l snare, and
townships with 1—2 harvests were excluded from sampling based on logistical
constraints. This hierarchical approach assumes that the frequency of harvest in each
township reﬂects the overall density of bears within that township.

Second, we stratiﬁed for potential bear habitat within the study area. Hirsch
(1990) found that bears on Drummond Island, Michigan, utilized aspen-birch, coniferous,
upland hardwood, and wetland vegetation types in the spring, summer and fall seasons.
We therefore, selected aspen/birch, emergent wetland, and mixed lowland vegetation
classes as potential bear habitat for sampling locations.

Third, because of the high density of roads and humans in the region, we placed a

road buffer of 500 meters on all county roads, streets and highways within the area.

63

Fourth, we selected all public land within the area so that access by ﬁeld personnel was
not limited. After all stratiﬁcations were made, we overlaid all of the coverages and
developed a new coverage which included publicly owned potential bear habitat greater
than 500 meters from county roads, streets and highways.

Lastly, we selected the sampling locations with the criteria developed based on
the frequency of harvest in each township. Sampling site locations were either made at
random using ArcView 3.2, or utilized existing MDNR bear bait index locations. In
some instances, there was no public land in townships where historical harvest had
occurred. In these townships we attempted to gain access to private lands. We used a
conservative spacing distance of 5 km calculated from White et al. (1982) and based on
female bear home range size for Michigan (Etter et al. 2002).

We veriﬁed our sampling locations with a subset of radio-telemetry data collected
in the months June and July (1,415 locations collected from 1991—2000) for 126 radio-
collared bears (64 males, 62 females; Etter et al. 2002). We found that 64.5% of
locations were located on public land, so we would not expect bias by sampling public
lands. We found that 49.2% of the locations were within the vegetation classiﬁcations
that we included in our stratiﬁcation.

Clark and Smith (1994) determined the effective sampling area for black bears by
placing a buffer of an adult female home range around each sampling location. Using

this methodology in our veriﬁcation, we placed a buffer of an average adult female bear

home range (131 kmz; Etter et al. 2002) around each sampling location and determined

the percentage of telemetry locations that were within the buffered area. We found that

87.5% of the telemetry locations were within the buffered area. Based on the high

64

percentage of telemetry locations collected within the equivalent of one female home

range of our sampling locations, we believe that we sampled in the proper locations.

Hair Snaring

We collected hair samples using two strands of barbed wire encircling a baited
location at uniform heights of 8 cm and 20 cm (Figure 2.2). This design differs from the
l-stranded design pr0posed by Woods et al (1999). We attracted bears with l-pound
bacon and anise extract, which was aerially suspended over the snare. We visited snares
at 5—7 day intervals for a total of 5 checking occasions (June 22 — July 26, 2003). We
collected hair samples from each barb on the top strand (strand at 20 cm) independently
and placed them in pre-numbered coin envelopes. Because bear cubs cannot be legally
harvested and therefore included in our last capture occasion (i.e., registered harvest), we
did not collect hair samples from the bottom strand of wire (strand at 8 cm) and set the
top strand at 20 cm to minimize the occurrence of cubs depositing hair on the wire. Hair

samples were stored in a dry location pending laboratory analysis.

Hunter Harvested Bear Tissue Collection

Current management of bear in the NLP of Michigan includes recreational
harvest, which is allocated by the MDNR through a lottery with a preference point
system. Because of the large number of applicants (14,330 applicants for 1,700 available
licenses in 2002; Frawley 2003), the ability to successfully draw a tag may require 4-5
years. There are two bear hunting seasons offered in the study area. The ﬁrst hunting

season is a 7—day ﬁrearm season (19—25 September 2003) where hunters can use bait and

65

hounds. The second is an archery only season (3—9 October 2003) where hunters can use
exclusively bait. Successful hunters must register harvested bears and hunter compliance
is estimated at 98% (Visser, MDNR, person. comm.; Frawley 2002). We trained MDNR
personnel for collecting muscle tissue and hand-plucked hair samples from registered
bears. To determine age of harvested bears, the MDNR collects tooth samples for annuli
aging. Extracted teeth contain small portions of gum tissue and saliva. We examined

using collected tooth samples as the primary DNA source.

Genetic Analysis
Sample Selection

The hair snare is an effective sampling device, which can collect multiple hair
samples with varying numbers of hair follicles across each checking interval. For
example, Mowat and Strobeck (2000) collected 4,245 hair samples in their study of
grizzly bears from British Columbia, Canada. Although this methodology is effective at
collecting hair samples, it may not be possible or necessary to genetically analyze each
sample. Project budgets are often limiting and the quantity of hair in each sample can
affect the genotyping error rate (i.e. few follicles can increase the genotyping error rate;
Goosens et al. 2000). The number of hairs collected for this project exceeded available
funds for genotyping each individual hair sample. Therefore, we developed a series of
selection criteria to sub-sample hair samples trying to account for the spatial and
temporal components of our sampling design.

To ensure an adequate quantity of DNA, but not limit sample size, we considered

only those hair samples with 5 or more hair follicles. Our ﬁrst selection included a single

66

hair sample from each snare where a hair sample with 5 or more follicles was deposited
by a bear. Because we did not move snare locations between checking occasions, it was
possible for any snare to collect hair samples on a maximum of 5 occasions. To account
for this temporal factor of bears visiting snares through time, our second selection criteria
consisted of a single hair sample from each snare that collected hair on each checking
occasion (excluding hair samples that were already selected in our ﬁrst selection). The
ﬁrst two criteria produced a hair sample from every snare that was visited by a bear on
every occasion that a bear visited, but no information about the incidence of multiple
bears visiting a snare during a single checking occasion. Therefore, we randomly choose
a subset of hair samples collected on the same date and location for our third selection
criteria. We used these data to examine the incidence of multiple bears visiting across a
single checking duration. We genotyped all muscle tissue/tooth samples obtained from
harvested bears.
Microsatellite Genetic Analysis

We extracted DNA using Qiagen DNeasy® Tissue Kit (Qiagen Inc.). All hair
follicles were removed and placed in a 1.5-ml centriﬁJge tube. DNA extractions from
hair follicles occurred within approximately 7 days aﬁer collection to prevent degradation
of DNA and increase ampliﬁcation success (Roon et al. 2003). We used tooth samples
collected from registered bears (muscle tissue samples were used when tooth samples
were not available) as the DNA source from hunter-harvested bears. DNA ﬁom muscle

tissue and tooth samples was quantiﬁed using a spectrophotometer and were diluted to a

working concentration of 20ng/ul. The quantity of DNA obtained from hair samples was

67

not quantiﬁed because of the small expected quantities. Aﬁer extraction, each DNA
sample was placed in —70 degrees C freezer pending genetic analysis.

To differentiate individuals we genotyped extracted DNA with ﬁve polymorphic
microsatellite loci including, GlOX, GlOL, GlOD (Paetkau et al. 1995), UarMU59, and
UarMUSO (Taberlet et al. 1997). Loci were ampliﬁed using polymerase chain reaction

(PCR) in 10 pl reaction volumes including 1 ul 10x PCR Buffer (IOmM Tris-HCl at pH
8.5, 1.5 mM MgC12, 50 mM KCl, 10 ug/mL nuclease-free bovine serum albumen,

0.0025% Tween-20), 200 pmol dNTPs, 10 pmol of forward and reverse ﬂuorescently-

labeled primers, 0.2 ul of commercial Taq polymerase (New England Bio Labs, Beverly,
MA), additional MgClz, and Sul of DNA (unknown concentration for hair and for tissue).

Thennocycler conditions included a 2-minute denaturation at 94 degrees C followed by
35 (UarMU59) or 42 cycles of 30 seconds at 94 degrees C (GlOL at 94 for 1 minute), 1
minute at locus-speciﬁc annealing temperature, 1 minute at 72 degrees C and one cycle

of 5 additional minutes at 72 degrees C. We used denaturing acrylamide gels (6.5%) for

electrophoresis visualized by LI-COR® 1R2 Global Edition DNA Sequencer. We used

molecular weight standards and individuals of known genotypes collected from muscle
tissue to standardize all genotype scores. SagaTM genotyping software (LI-COR, inc.,
Lincoln, NE, USA) was used to determine genotypes. All individual genotypes were
manually checked by two experienced lab personnel.
Quality Control

We used quality control protocols developed by Paetkau (2003) to minimize

genotyping errors and provide reliable genotypes which included, development of a

68

database around the ﬁeld database, conservative genotype scoring, culling of samples
with incomplete or contaminated genotypes, re-analysis of samples with uncertain scores,
and careful scrutiny of samples with similar genotypes for the detection of genotyping
errors.

We used SagaTM genotyping software to develop a database around our ﬁeld
database to limit notation errors made during manual entry. We initially genotyped all
samples for all 5 loci and conservatively scored genotypes by not recording scores for
samples that could not be conﬁdently scored by 2 experienced lab personnel. After the
initial genetic analysis we culled all samples where unambiguous and veriﬁed scores for
3 of the 5 loci were not assigned. Sample culling is important because genotyping errors
are related to sample quality. Thus culling samples that do not produce genotypes from
the majority of loci helps to ensure that the remaining samples are of a higher quality
(Paetkau 2003). After samples were culled, we attempted to replace culled hair samples.
Because we only chose a single hair sample from a visited snare on every date the, snare
was checked, we often had other hair samples that could replace culled samples. When
possible, we choose 2 hair samples to replace a culled sample. We did this in an attempt
to obtain a genotype from sites where samples were culled and minimize the possibility
of a replacement sample also being culled. After all samples were genotyped a single
time for all loci and all necessary culling carried out, the remaining samples with missing
genotypes by locus were then reanalyzed and scored. After this reanalysis, samples that
did not have conﬁdent scores for 4 loci were culled and no replacements identiﬁed and

selected.

69

To identify and correct genotyping errors we used the mismatch methodology
described by Paetkau (2003). This methodology involves the careful scrutiny of all
genotypes that are nearly identical, differing by one or two loci [referred to as 1-
mismatch (l-MM) and 2-mismatch (2-MM) pairs, respectively]. The basis for this data
scrutiny is that the probability of genotyping errors is quite low in most cases. The most
important samples in which to identify errors are those that matched, but a genotyping
error caused them to miss-match, thus creating an additional individual (Paetkau 2003).
We identiﬁed mismatched samples with program GENECAP (Wilberg and Dreher 2004)
and the scores of each sample in the mismatch were scrutinized by the raw genetic data
for possible errors. If the mismatch could not be resolved by examining the raw genetic
data, the samples were reampliﬁed and genetically reanalyzed to determine if a
genotyping error occurred. All incidences of miss-scored genotypes were noted and
corrected within the database.

Genotyping Error Estimation

For purposes of population estimation, the accuracy of genotypes is critical
because genotyping errors can positively bias population estimates (Waits and Leberg
2000; Creel et al. 2003). To minimize errors, Paetkau (2003) suggested careful data
scrutiny, which was used in this study. We recognized an additional means to estimate
genotyping errors using samples obtained from harvested bears. Upon the registration of
a legally harvested bear, ﬁeld personnel were instructed to collect a tissue sample and a
hand-pulled hair sample. We randomly choose 110 harvested bear hair samples for our
error analysis (matching muscle tissue samples were already within the genetic analysis).

We handled error samples consistent with snare collected samples including extraction

70

and quality control protocols. Extracted hair samples were genotyped concurrent with
snare-collected hair samples. To ensure that the hair and tissue samples from the same
bear were independently genotyped, sample identiﬁcation numbers were randomly
assigned and the key was not made available to laboratory personnel.

Paetkau (2003) suggests the careful scrutiny of genotypes with the identiﬁcation
of mismatching samples as one means to detect and correct genotyping errors. To
examine the efﬁcacy of this methodology of error detection we matched genotypes from
hair and tissue samples before the careful scrutiny of mismatches and identiﬁed and
quantiﬁed inconsistencies. We then included both hair and tissue samples within the data
scrutiny and corrected any inconsistencies. Finally, we matched hair and muscle tissue
genotypes and quantiﬁed errors following review of data. We compared results obtained
before and after review of data.

Statistical Genetic Analysis

Individual discrimination using highly polymorphic microsatellite loci
necessitates that a signiﬁcant number of loci be analyzed to minimize the probability of
unique individuals sharing the same genotype (Waits et al. 2001). Some non-zero
probability of unique individuals sharing the same genotype can result in two different
bears being identiﬁed as the same individual. This can result in negative bias in
population estimates because fewer individuals are uniquely identiﬁed than truly exist in
the population (Mills et al. 2000). The probability of identity (PI) is a statistical measure

to determine the probability of matching genotypes (Paetkau and Strobeck 1994). We

calculated PIsibs single locus with program GENECAP (Wilberg and Dreher 2004), which

provides the probability of two full siblings in the population sharing the same genotype

7]

for the loci examined and represents the upper range of possible PI values in the

population (Taberlet and Luikart 1999). The following is the equation for

Phil” single locus 3
P] ‘b . — 0 2 2 2 2 4
S’ 3 Single locus — - 5+ (0.521),. )+[0-5(2pi ) l-(O-ZSZPI- )

where p,- is the frequency of alleles and the overall PIsibs single locus is calculated as the

product of the PIsibs single locus values for each locus.

Identical genotypes are assumed to be from the same individual, but because there

is a probability of two individuals sharing the same genotype, statistical rigor must be
used for the match declaration (Woods et al. 1999). We used Psib calculations from
Woods et al. (1999) in program GENECAP (Wilberg and Dreher in press) to calculate
the probability of a matching genotype with the minimum criteria for a match of P5,}, <

0.05 before a match was declared with the following equations:

(1 + 2p, +p2,.)
Psib(Homozygotes) : 4
P _(l+pi+pj+2pipj)
sib(Heterozygotes) — 4

 

 

where p,- and pj are the frequencies of the ith and jth alleles. Samples that matched, but

did not meet this criteria where excluded from capture histories as matching individuals

(Woods et al. 1999; Mowat and Strobeck 2000; Boerson et al. 2003).

72

Population Estimation

Previous research using non-invasive techniques to estimate bear abundance used
closed capture models in program CAPTURE (White et al. 1982; Boerson et al. 2003;
Mowat and Strobeck 2000; Poole et al. 2001). For our initial population analysis we used
closed capture models in program MARK (White and Bumham 1999). One of the major
advantages of program MARK is the ability to create estimation models that account for
a capture and behavioral response by capture occasion.

We deﬁne parameter notation as presented in Otis et al. (1978) to assist in

understanding our population estimation models as follows:

p,- Probability of an animal’s initial capture in time period 1'
c,- Probability of an animal’s subsequent recapture in time period 1'
N Population size

We created 4 a priori models to estimate population size by incorporating our
knowledge of ﬁeld methods and bear biology. All models, when applicable, were
constructed to account for the harvest of bears being a different capture method than

samples collected with hair snares (denoted as “hunt” in model names). We constructed

our ﬁrst model to account for a probability of capture that varied by occasion (model M,,

parameters p1, p2, p3, p4, p5, p6, N). Our second model accounts for a differential

probability of recapture, also known as a behavioral response, which might be expected

when bait is used as an attractant (model Mb+hunb parameters p1-5, c2_5, p6, N). Our

third model accounts for a behavioral response, a time varying response, and the hunt

being a separate capture method (model M,+b+hu,,,, parameters pl, p2, p3, p4, p5, c2-5, p6,

73

N). Our last model was constructed to account for a behavioral response, the harvest, and

a change in the number of hair samples collected in occasions 1-3 as compared to
sampling occasions 4—5 (model M, 1- 3,4- 5+b+ hum, parameters pl_3, p4-5, 02-5, p6, N).

This last model was formulated after a perceived difference in the number of hair samples
collected in weeks 1—3 and 4—5 by ﬁeld personnel.

Using these a priori models in program MARK we ﬁrst derived population
estimate using all 6 capture occasions. Waits and Leberg (2000) demonstrated through
simulation and Creel et al. (2003) through empirical analysis that genotyping errors tend
to cause a positive bias in population estimates. Upwardly biased population estimates
can be particularly problematic when population estimates are used to set harvest
objectives. Second, we used estimation models proposed by Lukacs and Bumham (in
press) to account for genotyping error in the population estimate. These models are an
extension of the class of closed capture models offered in program MARK (White and
Bumham 1999). The major assumptions in these models are geographic and
demographic closure, incorporation of enough loci to differentiate individuals, and that
genotyping errors create unique genotypes that do not exist in the population (Lukacs and
Bumham in press).

These models were written in SAS (SAS Inc. 2002), but are similar to those
contained in program MARK (White and Bumham 1999) because they allow the user to
account for differential probability of capture values attributed to time, and behavioral
factors. Unlike the closed capture models in program MARK, the Lukacs and Bumham
(in press) models include an additional parameter denoted as o. and deﬁned as the

probability that the animal was genotyped correctly, given its ﬁrst observation. The

74

value of a is derived using the capture histories, or incorporating a pre-measured estimate
of genotyping error (Lukacs and Bumham in press).

Because we estimated our genotyping error rate using hair and tissue samples
from harvested bears, we used ajoint likelihood function to correct population estimates
for our measured error rate. The following is the equation for the binomial portion of the

joint likelihood function:
’7 x n — x
L(a|n,x) =( )a (l —a)
x

where n is the number of samples tested and x is the number of correctly assigned
genotypes. Because we incorporated a pre-measured error rate into the models, the value
of 01 remained consistent for all models used to estimate population size.

For all models and methods of analysis we ranked competing models based on
Akaike Information Criterion (AICC) to determine which models were best supported by

the data.

RESULTS
Hair and Tissue Sample Collection

We set a total of 239 snares between 22 June 2003 and 26 July 2003. Each snare
was checked 5 times for a total of 1,195 checks. We collected 1,144 hair samples from
122 (51%) snares. Laboratory examination of the hair samples determined that 687 of
the 1,144 (60%) hair samples had 5 or more hair follicles. DNA was extracted from

samples with 5 or more hair follicles within approximately 7 days of their collection.

75

We registered 423 harvested bears in the fall of 2003 and obtained a tissue sample
from 421 (99.2%). We were able to obtain a quantiﬁable DNA product from all cases
where a tooth sample was used for the DNA extraction (n = 419). In addition we

obtained hand-plucked hair samples from approximately 75% of the harvested bears.

Genetic Analysis
Sample Selection and Quality Control

We genotyped 1,026 hair and tissue samples including 416 hair snare collected
samples; 110 randomly selected hair samples from harvested bears for our error analysis;
421 harvested bear tissues; and 79 replacement hair samples after our quality control
culling procedure (Table 3.1).

After the initial genetic analysis of all samples at 5 loci, we culled 162 samples
because they lacked scores for 23 loci and 13 samples because they showed evidence of
contamination (multiple genotypes indicating hair was deposited on the same wire barb
by 2 or more individuals; Table 3.1). After we culled samples, we then reanalyzed
inconclusive scores and genotypes that contained samples that failed to amplify. In this
process we reanalyzed 238 locus genotypes from previous PCR product and re-ampliﬁed
and analyzed 38] locus genotypes.

After the reanalysis, we then culled 27 samples that did not have _>_4 loci scored
(Table 3.1). We excluded mismatches between harvested bear samples because these
samples were known to be collected from different individuals. We identiﬁed 97
genotypes that differed by 1 allele and 229 samples that differed by 2 alleles. These

identiﬁed mismatches include samples from harvested bear hair and tissue samples.

76

 

 

3: :3 h 5 m— NE :2: 38,—.
pm me o F o 3 2. 8.95% has 28:58.3:
Ev m o o o m :3 8E8?“ “822 Sam
8 E _ _ _ _ _ o. _ 8353 :2 Sum
8m 02 0 mm .2 R: 0:. $358 vacuum
3:8: Z 3:3 ass—5::— houua 3.8% 32 VW :oBEEaEeU 3.33. to. am 3:52.00 Cowman 258:5
3...: .83. 3:3 .oZ >5 3:3 .oZ con—5.2 \3 10:3 62 Lon—5.2

 

$.5ng 553.3. :23 .5. mam—«:5 zeta—:38 .8.— ..3: hon—5:. :3: 2: can 69:3 33:5:
_53 of Earns—=83 €35.35 coca 3:5 3:55. 6958 .3. v w :33 3:3 8.95% no hon—5:. maﬁaﬁﬁﬁneo 29$ :2:
33.53 he .855:— 838 .8. m w :33 3:3 8383. me 3:55. 699328» 8.95% he 3:5...— 2: he fin—:5 .—.m «Bah

77

All mismatched samples were then examined and scrutinized for possible errors. We re-
ampliﬁed 47 samples in this process to identify possible errors. We identiﬁed 55
samples where genotyping errors occurred and changed scores accordingly.
In addition, we culled 7 samples based on incomplete scores and evidence of
contamination (Table 3.1). We used 816 hair and tissue samples after the quality control
procedures to estimate the rate of genotyping error and to estimate population size (Table
3.1).
Genotyping Error Estimation

We used 95 hair and tissue samples from harvested bears to estimate the rate of
genotyping error (14 hair samples and 1 tissue sample were culled during the quality
control methods). We estimated the overall misidentiﬁcation rate before mismatch
examination to be 20% because 19 out of the 95 pairs differed. The error rate differed
across loci with the mean error rate per locus calculated at 0.044 and at 0.029 per allele
(Table 3.2). The GlOX locus had the highest error rate and UarMU59 locus had the
lowest (Table 3.2).

The error rate decreased after we examined mismatched genotype pairs, because
15 out of 19 mismatching samples were corrected after mismatch examinations.
Therefore, we did not correct 4 out of 95 sample pairs giving us an overall individual
misidentiﬁcation rate of 4.21% after quality control procedures. Of the 15 samples that
were corrected with the mismatching procedures, 9 resulted from scoring errors differing
by 2 base pairs, 1 was the result of allelic dropout, and 5 resulted from loading errors and

were discovered as a result of including error samples.

78

 

 

.83 39:393.. 3...: 9.82. 29:333.. 3...”.— ..83 38:03.. 3...: 0.8.3.. 2.9233. 3...”:

w::.: 3:: — 8.: 3:: .302
: 2:: : _No: 3:23:
2:: 3:: 3:: 3:: 3323:
m::.: 2:: 2:: N8: :20
_ 5.: _No: _ 5: N8: 020
m::.: m::.: _ S: 3:: X30
2:35:35 338:... 533.5538 538:... 5:35:35 538:... 5.33.538 538...... 2.25

 

.25..— .E 53.8%.. 6.... 29:3 .8: E... 2.2.. .3.— ..oteno. 9... 83.. 3...“: 4.3395232 535%...
a: 288...... 3.2.8 3:5... .83 6.3 98.3: :3: 83.. 3...... :39»...qu o... .3 b.2555 .N.m 933,—.

79

The remaining 4 sample mismatches that were not corrected resulted from differences
greater than 2 alleles (2 sample pairs) and pairs that differed by 2 alleles (2 sample pairs),
but not corrected. The mean error rate per locus and per allele decreased as a result of
examining mismatched samples (Table 3.2).

Statistical Genetic Analysis

We estimated PIsibs single locus across all loci and unique genotypes at 0.00865 (n
= 544). The number of alleles and the value of PIsibs single locus differed across loci with
locus G 1 0L being the most informative based on the smallest PIsibs single locus value

(Table 3.3). All matches met the criteria of PS”, < 0.05 except 1 of 5 matching harvested

bear genotypes. We identiﬁed 165 bears from samples collected from hair snares. We
identiﬁed 412 unique genotypes from harvested bears and 33 of these genotypes were
captured by hair snares. One harvested bear was deleted from the analysis because it was

harvested outside of the study area in the Upper Peninsula of Michigan.

Table 3.3. Summary of the number of alleles per locus and the PIsibs single locus
calculations for each locus used to differentiate individuals.

 

 

No. of
Locus Alleles PISibs single locus
max 10 0.4385
(3101) 9 0.3876
GlOL 14 0.3257
UarMUSO 10 0.3635
UarMU59 7 0.4298

 

80

Population Estimate
Results of model selection based on AICC were very similar between the program

MARK and the Lukacs and Bumham (in press) models (Table 3.4 and Table 3.5). The
largest difference between the results was in the derived population estimates. MARK
estimates were consistently larger, with the most supported model being 176 bears higher
than that of the Lukacs and Bumham (in press) models (Table 3.4 and Table 3.5). This
result was suspected because we measured a non-zero error rate within our genetic data
and suspected that the MARK estimates would be inﬂated because of this error rate.
Therefore, the remaining results and the population estimate will focus on the estimates

derived from the Lukacs and Bumham (in press) models.

Bumham and Anderson (1998) suggest that when AAICi is less than 2 there is
reasonable support for a model by the data. Two models, Mb+hum and Mtl-3,4-5+b+hunt
had AAIC; values less than 2 (Table 3.5). Model Mb+hunt appeared to have the best ﬁt

(w,- = 0.63) however, there was also support for model Mtl-3,4-5+b+hunt (w,- = 0.35).
There appears to be a behavioral response as illustrated by the strong support for both the

Mb+hunt and Mtl-3,4-5+b+hunt modelS-
Because there was support for the Mb+hunb M,1-3, 4-5+b+hum, and Mt+b+hunt

models with w,- vaules greater than 0, we chose to use model averaging to derive the

p0pulation estimate (Buckland et al. 1997). Using this methodology we derive a

population estimate of 1,880 bears (95% Cl l,454—2,306).

81

0.0.5.8 530.30:

.0

._ 0. E30 0. 30:08.0: 0.0 00:55:: 000:... ..0£o:0 .05 Eve... 0:0 :0 00:0E>0 :0 BE?» 0:. 8:03.50. £023 . BE?» 002 0

.07.. :_ 80:20:20 0

0.0: So :_ 55.03683
:o 85:30 0: m_ 0.0:. 003000: 07.. .26 007.. :00: 03 :50. 3:00.200 05 3:03:60 :0 .33 8:25 5:06.85 0.00:8? 0

 

 

2 31.8.. a... 08... 20.; 20.2.8. 2 a: é .a .8 .8 .a .2

3.015.. 03.. ~ch 50 8:: :V. 2 a: .20 .8 .a .8 .3 .a Eff...

E200... So... 2...: S... 08.2.... 2 an. .30 .3: .2: s..ff3...-:2

33.32 08.0 .8... 8.; 2.0.42... 2 a: .20 .2: BEL:
_0 $8 0 a. 0.... .6 no... 320580.. 2:02 .282

 

.33. 52.5.... 0.... 83>: 0:22
E050...— :_ 0.0.0:... 5:05:00 0.5.9.0 @0020 5.3 .0033: 00.0530“: .3202: 00:06:50 3m: 0.: 6:0 . 2 0.0. a: .002

<

5.0.0803: .050: 3.5:. ”5:20... 5.3.—0:2 a: .52 0... 5 E00: 0.00:. .160 .8.— m_0:::. 5:08:00 0.5:00 come—U 0.0. 050,—.

82

000508 500.000: 0

._ 00 5:0 00 0002058: 0.0 8250:: 08::- 8.00:0 8>0 _0005 0:0 00 008030 .00 £303 05 8:08.00. 0053 . £0.03 002 0

002 :_ 80:08-05 0
.0000 .00 :_ 88808083

.00 008030 0: 8 0.0.0 080000 02 .0>0 00?. 00m: 03 5.8 0:000:00 85 _0:00_000 :0 .00: 58:8 5005.005 0.000002 0

 

 

000.708 00... 000.0 000.00 08.0.00. 2 .0 .8 .8 .8 .8 .8 .8 .2
20.0.80; 80.. 000.0 000.0 000.800- 2 .0 .8 0-8 .8 .8 .8 .8 .8 3.08:2
000.0403 80.. 000.0 00... 850:.- z .0 .8 0-8 0-8 0-8 .5210: 0.0 -0 .2
800-000.. 000.. 000.0 000.0 00000:.- 2 .0 .8 0-8 0-8 22:02

_0 $3 0 Q 0.0: 0.00 0002 08850.0: 0502 _0002

 

8.8.3 05 500:...5 0:0 800:1— »: 00800.:— m_0005 :33 002.00 8.050% 00.0.0.5 00:00::00 {cma 05 0:0 . 2 0.: .04 .00;-

<

88850.0: .050: .0005 050305 50020:: .00 .52 0:. :_ 0.00: :00... 2:00 .0.— m_0005 :000508 0.5000 0000.0 m.n 030-0

83

DISCUSSION
Study Design

When estimating animal abundance it is critical to consider the spatial scale at
which the species is managed. Therefore, our study design focused on the scale at which
bear management decisions are made in Michigan. Other researchers conducting non-
invasive DNA sampling have utilized the systematic grid design as suggested by White et
al. (1982) to increase snare density and the probability of an animal encountering a
trapping location (Mowat and Strobeck 2000; Poole et al. 2001; Eason et al. 2001;

Boerson et al. 2003). However, all previous studies were conducted at smaller spatial
scales. Mowat and Strobeck (2000) estimated grizzly bear abundance in a 9,866 km2
study area in British Columbia, Canada. Boerson et al. (2003) estimated the black bear
population at the 329 km2 Tensas River Tract in Louisiana. Using the systematic grid
design with 4 or more traps in each female bear’s home range described by White et al.

(1982) with our study area size of 36,848 km2 would require approximately 1,125 snares.

This was not feasible with available funding, personnel and private land access issues.
Therefore, we used a stratiﬁed random sampling design to identify biologically relevant
and accessible sampling locations. We veriﬁed our design with radio telemetry data and

believe that we sampled in the proper locations.

DNA Sample Collection
Non-invasive DNA sampling with the use of hair snares is an effective means to

sample bear populations. One major advantage of such a methodology is the large

84

sample sizes that can be obtained. We identiﬁed 165 different bears in 5 weeks of
sampling. Poole et al. (2001) identiﬁed 98 different grizzly bears (Ursus arctos) in
approximately 9 weeks of sampling. Comparable sample sizes obtained from live
trapping would be hard to achieve. The MDNR live-captured and handled 126 different
bears over a lO-year period from 1991—2001 in the NLP of Michigan (Etter et al. 2002).
Non-invasive methods can be much more time, labor and cost effective at capturing
individuals.

The new challenge with this method is to determine the best way to sub-sample
hair samples if project budgets are limited. For example, we collected 1,144 hair samples
and 687 had 5 or more hair follicles. Because our project budget did not allow for all hair
samples to be genotyped, we developed a hair selection strategy to sub-sample our data.
Boerson et al. (2003) collected 448 hair samples with greater than 10 hair follicles and
choose 116 samples to derive a population estimate. Unlike other capture methods, hair
snares can not, in any relative sense, be saturated. Therefore, it is possible to sample
multiple individuals during a single snaring duration (i.e. that is a 7 day interval in our
study). For instance, on one occasion we genotyped 7 hair samples from a single snare
and detected 4 different bear genotypes. On another occasion we genotyped l4 hair
samples and only detected a single genotype. Poole et al. (2001) detected as many as 7
different bears at an individual snare during a snaring session. Therefore, the researcher
has the potential to impact an individual bear’s probability of capture based on which hair
samples are selected to be genotyped. Further work is needed to identify the optimal

number of hair samples to genotype and how sub-sampling impacts population estimates.

85

We found the use of the harvest as an additional capture occasion to be useful for
several reasons. First and most importantly, we were unable to converge on a population
estimate in program MARK without harvest as a capture occasion. This can be attributed
to the large sample size in which the harvest provided (approximately 22% of the
estimated population). Second, we found that sufﬁcient quantities of DNA could be
extracted from bear incisors that were historically collected by the MDNR for aging
bears. Excellent registration compliance by hunters (approximately 98%; F rawley 2003)
and the fact that we collected DNA samples from 99.2% of the registered bears provided
a large re-capture sample for our population estimate. Third, ampliﬁcation success as
related to the quality of DNA in teeth/tissue samples is greater than that of hair samples.
We found 29 out of a possible 1,110 locus scores (2.61%) for hair samples did not
amplify, while 26 out of a possible 2,080 locus scores (1.65%) for harvested bear
teeth/tissue samples did not amplify. Fourth, the ability to collect both hair and tissue
samples from harvested bears allowed us to quantify genotyping error, which we then
used to correct the bias in our population estimate. Overall, the sampling of harvested

animals, when possible, can increase sample size and data quality.

Statistical Genetic Analysis

The ability to genetically resolve unique individuals is an important aspect of non-
invasive sampling. 1f truly unique individuals share genotypes when examined, it
violates one of the assumptions of the Lukacs and Bumham (in press) models. Mills et
al. (2000) found through simulation that truly unique, but matching genotypes biased

population estimates low because fewer individuals were identiﬁed than existed in the

86

population. The resolution of genetic markers can be examined through the PI

calculations. We calculated PIsibs single locus to be 0.00865 with a likelihood ratio of l

in l 16 of two siblings sharing the same genotype. Boerson et al. (2003) used 12 loci and
estimated PIsibs single locus to be 0.00134, or a 1 in 745 chance of observing identical

sibling genotypes.

We identiﬁed 4 out of 4 1 7 genotypes from harvested bears that matched. This

suggests that we may not have used enough loci, resulting in a PIsibs single locus value

that was not sufﬁciently low for our population. Our ﬁnding of matching genotypes for
the 5 loci that we examined may be related to the high productivity of the NLP
population of black bears. Etter et al. (2002) found that 27 of 28 radio-collared sows bred
by 3 years of age and mean litter size of 76 sows was 2.6 cubs/sow. This compares to
work conducted in the Upper Peninsula of Michigan where DeBruyn (1997) found the
mean age of ﬁrst reproduction was 4.4 years and mean litter size was 2.0—2.7 cubs/sow.
Because of the high productivity of the NLP population related to younger ages at ﬁrst
reproduction, we might expect that the relatedness of the population is greater, resulting
in greater similarity of genotypes. Therefore, future genetic identiﬁcation of individuals
from non-invasive sampling in the NLP of Michigan might require additional loci, or the
analysis of additional loci for matching genotypes to further limit the probability of two
unique bears in the population sharing identical genotypes. Overall, we believe that there
is a low occurrence of matching genotypes from unique individuals. The directional bias
from such matches has been demonstrated to be downward (Mills et al 2000), which

provides a more conservative estimate of population size.

87

Quality Control and Genotyping Errors

Genotyping error has been a large focus of non-invasive methodologies because it
biases population estimates. Waits and Leberg (2000) found population estimates
through simulation to be biased on the magnitude of 2 times larger than true population
size with a genotyping error rate of 0.05/locus. We found that the quality control
protocols proposed by Paetkau (2003) substantially decreased our individual genotype
misidentiﬁcation rate. This can be attributed to culling of samples that did not produce
conﬁdent scores, conservative scoring, and the identiﬁcation of genotypes that are similar
(l-MM and 2-MM sample identiﬁcation). Goosens et al. (1998) documented an error
rate of 4.86% per locus aﬁer extractions of 3 hair follicles and using a multi-tubes
approach. This compares to our mean error rate of 4.4% per locus before the quality
control protocols, but is greater than our ﬁnal genotyping rate per locus of 1.1% after
quality control protocols.

We found the use of hair and tissue samples from harvested bears as very
beneﬁcial for examining genotyping errors. Others have used two sources of DNA from
the same individual to examine genotyping errors (Parsons 2001; Sloane et al. 2000).
Interspersing error testing samples within the data increased our ability to identify
potential loading errors which can result in numerous samples being misidentiﬁed.
However, we still identiﬁed a non-zero probability of genotyping errors. Two of the 4
remaining errors were associated with errors that differed by more than two alleles and
could potentially be corrected if we choose to examine samples that differed by 3 and 4
alleles. The time required to execute the quality control protocols was extensive, but

decreased the number of genotyping errors. Because of our non-zero error rate, we

88

suggest using the quality control protocols in conjunction with a methodology to account

for remaining errors within the estimate.

Population Estimate

Biased population estimates caused by genotyping error have raised suspicion
about the efﬁcacy of non-invasive sampling to estimate wildlife population abundance
(Creel et al. 2003). This bias occurs because the capture-recapture assumptions of correct
identiﬁcation of individuals and tag loss are violated. Is the violation of marking
assumptions with non-invasive methodologies a new phenomena, or have they been
present all along with other marking methods, but too difﬁcult to quantify? Stevick et al.
(2001) examined the occurrence of natural tag misreads in a humpback whale
(Megaptera novaeangliae) capture-recapture experiment and found an error rate of 0.125
with sub-premium photo quality. The ability to quantify and examine tag misreads is one
advantage of non-invasive methodologies because it provides the opportunity to correct
for assumption violation in the estimate. The effectiveness of the non-invasive technique
with respect to large samples sizes and ﬁeld logistics cannot be argued, but genotyping
errors must be accounted for.

Lukacs and Bumham (in press) propose models that account for genotyping
errors in the population estimate. They liken these models to other models that have been
created to deal with assumption violations, such as those created by Otis et al. (1978),
Huggins (1991) and Pledger (2000). One assumption that remains for all closed capture-
recapture models is geographic and demographic population closure. Movement data

collected by the MDNR from 1991—2001 from 126 radio-collared black bears did not

89

document a single bear movement outside of our study area (D. Etter, MDNR, person.
comm. 2004). Therefore, it is valid to assume that Lakes Michigan and Huron to the
north, east and west and areas devoid of bears to the south provide geographic population
closure. In addition, survival rates during the summer months were estimated between
96% and 99% for adult and sub-adult males and females (Etter et al. 2002). Based on
these ﬁndings, we feel that we met the assumption of demographic population closure.
For comparison purposes, we derived population estimates that accounted for
genotyping error with the Lukacs and Bumham (in press) models and estimates that do

not account for genotyping error in program MARK (White and Bumham 1999). We

found that the AICC ranking of the models to be consistent between program MARK and

Lukacs and Bumham (in press) models (Table 3.4 and Table 3.5). We found the model
averaged estimate from program MARK to be 2,054 bears (95% CI 1,453—2,655). This
estimate was expected to be greater than the averaged estimate obtained from the Lukacs
and Bumham (in press) models because we violated the marking assumptions of the
capture-recapture models in program MARK, and this violation has shown to
overestimate population size (Waits and Leberg 2000). Thus, we feel that the estimates
from the Lukacs and Bumham (in press) models are conservative.

Overall, the estimate of 1,880 bears (95% CI 1,454—-2,306) is consistent, but
slightly higher than present MDNR population models (D. Etter, MDNR, person. comm.
2004). Because we used the harvest as a recapture method, this estimate reflects the
number of adult and sub-adult black bears before the 2003 hunting season. This estimate
excludes the cub portion of the population because cubs can not be legally harvested and

were excluded from sampling based on our hair snare design.

90

MANAGEMENT IMPLICATIONS

Historically, the MDNR monitored the bear population in the NLP of Michigan
with a bait index. The bait index provided managers with information about the trend in
the population, but did not provide estimates of population size. In addition to the bait
index, management decisions were based on an empirical population model, which
accounts for age- and sex-speciﬁc survival and reproduction (Garshelis and Snow 1988).
The MDNR conducted a 10-year research study to collect demographic information to
parameterize the population model (Etter et al. 2002). The critical parameter that was
missing in the model was an initial estimate of population size. Our results can be used
in the MDNR population model to aid in making decisions about future bear
management.

We found great success in using harvested bears as an additional capture
occasion. Utilization of this information drastically increased our sample size and
allowed us to estimate genotyping errors. Collection of samples was relatively simple
and proper personnel training ensured success. We found that tooth samples, that were
already being taken systematically to determine age, contained adequate quantities of
DNA for genotyping and can obviate collecting alternative sources of DNA (e.g. muscle
tissue).

Quality control protocols proposed by Paetkau (2003) drastically reduced the
amount of genotyping error in our data. The use of these protocols in conjunction with
the estimation models proposed by Lukacs and Bumham (in press), can provide unbiased

estimates of population abundance when using non-invasive techniques.

91

LITERATURE CITED

Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and
Wisconsin: a working map and classiﬁcation. United States Forest Service, North
Central Forest Experiment Station General Technical Report NC-178, St.

Paul, Minnesota, USA.

Boerson, M. R., J. D. Clark, and T. L. King. 2003. Estimating black bear population
density and genetic diversity at Tensas River, Louisiana using microsatellite DNA
markers. Wildlife Society Bulletin 31(1): 197—207.

Buckland, S. T., K. P. Bumham, and N. H. Augustin. 1997. Model selection: an integral
part of inference. Biometrics 53: 603-618.

Bumham, K. P., and D. R. Anderson. 1998. Model selection and multimodel inference:
a practical information-theoretic approach. Springer-Verlag, New York, NY.

Clark, J. D., and K.G. Smith. 1994. A demographic comparison of two black bear
populations in the interior highlands of Arkansas. Wildlife Society Bulletin 22:
593—603.

Creel, S., G. Spong, J. L. Sands, J. Rotella, J. Zeigle, L. Joe, K. M. Murphy, and D.
Smith. 2003. Population size estimation in Yellowstone wolves with error-prone
noninvasive microsatellite genotypes. Molecular Ecology 12: 2003-2009.

DeBruyn, T. D. 1997. Habitat use, food habits and population characteristics of female
black bears in the central upper peninsula of Michigan: a geographic information
system approach. Ph.D. Thesis, Michigan Technical University, Houghton,
Michigan, USA.

Eason, T. H., S. L. Simek, and D. Ziegler. 2001. Statewide assessment of road impacts
on bears in Florida. Research Work Plan. Florida Fish and Wildlife Conservation
Commission, Tallahassee, Florida, USA.

Etter, D. R., L. G. Visser, C. M. Schumacher, E. Calrson, T. Reis, and D. Rabe. 2002.
Black bear population management techniques. Final Report. Federal Aid in
Wildlife Restoration Project W-127-R-17, 18, 19. Michigan Department of
Natural Resources, Lansing, Michigan, USA.

Frawley, B. J. 2003. 2002 Michigan black bear hunter survey. Wildlife Division Report
3413. Michigan Department of Natural Resources, Lansing, Michigan, USA.

Gagneux, 1., C. Boesch, and D. S. Woodruff. 1997. Microsatellite scoring errors

associated with noninvasive genotyping based on nuclear DNA ampliﬁed from
shed hair. Molecular Ecology 6: 861—868.

92

Garshelis, D. L., and W. Snow. 1988. Minnesota black bear population model. User
Manuel, Version 1.1. Minnesota Department of Natural Resources, Unpublished
Report. 19 pp.

Goosens, B., L. P. Waits, and P. Taberlet. 1998. Plucked hair samples as a source of
DNA: reliability of dinucleotide microsatellite genotyping. Molecular Ecology 7:
1237—1241.

Hirsch, J. G. 1990. Black bear habitat utilization and habitat model validation in

Michigan. Masters Thesis, Michigan State University, East Lansing, Michigan,
USA.

Huggins, R. M. 1991. Some practical aspects of a conditional likelihood approach to
capture experiments. Biometrics 47: 725—732.

Lukacs, P. M. and K. P. Bumham. in press. Estimating population size ﬁom DNA based
closed capture-recapture data incorporating genotyping error. Journal of Wildlife
Management.

Michigan Department of Natural Resources- Mineral and Fire Management Division,
Michigan Department of Natural Resources-Resource Mapping and Aerial
Photography. 2001. Gap Land Stewardship. Michigan Department of Natural
Resources, Lansing, Michigan, USA.

Michigan Department of Natural Resources- Land and Mineral Services Division,
Resource Mapping and Aerial Photography. 1992. MIRIS Base Data. Michigan
Department of Natural Resources, Lansing, Michigan, USA. '

Mills, L. S., J. J. Citta, K. P. Lair, M. K. Schwartz, and D. A. Tallmon. 2000. Estimating
animal abundance using noninvasive DNA sampling: promise and pitfalls.
Ecological Applications 10(1): 283-294.

Mowat, G. and C. Strobeck. 2000. Estimating population size of grizzly bears using hair
capture, DNA proﬁling, and mark-recapture analysis. Journal of Wildlife
Management 64(1): 183—193.

Paetkau, D. 2003. An empirical exploration of data quality in DNA-based population
inventories. Molecular Ecology 12: 1375—1387.

Paetkau, D. and C. Strobeck. 1994. Microsatellite analysis of genetic variation in black
bear populations. Molecular Ecology 3: 489—495.

Paetkau, D., W. Calvert, I. Stirling, and C. Strobeck. 1995. Microsatellite analysis of
population structure in Canadian polar bears. Molecular Ecology 4: 347—354.

93

Parsons, K. M. 2001. Reliable microsatillite genotyping of dolphin DNA from faeces.
Molecular Ecology Notes 1: 341—344.

Pledger, S. 2000. Uniﬁed maximum likelihood estimates for closed capture-recapture
models using mixtures. Biometrics 56: 434-442.

Poole, K. G., G. Mowat, D. A. Fear. 2001. DNA-based population estimate for grizzly
bears (Ursus arctos) in northeastern British Columbia, Canada. Wildlife Biology
7: 105—115.

Otis, D. L., K. P. Bumham, G. C. White, and D. R. Anderson. 1978. Statistical inference
from capture data on closed animal populations. Wildlife Monograph 62: 1-135.

Ratti, J. T. and E. O. Garton. 1996. Research and experimental design. Pages 1-23 in
T.A Bookhout, editor. Research and Management Techniques for Wildlife and
Habitats. Allen Press, Inc., Kansas, USA.

Riley, 8. J ., D. J. Decker, L. H. Carpenter, J. F. Organ, W. F. Siemer, G. F. Mattfeld, and
G. Parsons. 2002. The essence of wildlife management. Wildlife Society
Bulletin 30(2): 585—593.

Roon, D. A., L. P. Waits, and K. C. Kendall. 2003. A quantitative evaluation of two
methods for preserving hair samples. Molecular Ecology Notes 3: 163—166.

SAS Institute. 2002. SAS/STAT user’s guide. Version 8. SAS Institute, Cary, North
Carolina, USA.

Sloane, M. A., P. Sunnucks, D. Alpers, B. Beheregaray, and A. C. Taylor. 2000. Highly
reliable genetic identiﬁcation of individual northern hairy-nosed wombats from
single remotely collected hairs: a feasible censusing method. Molecular Ecology
9: 1233-1240.

Stevick, P. T., P. J. Palsbéll, T. D. Smith, M. V. Bravington, and P. S. Hammond. 2001.
Errors in identiﬁcation using natural markings: rates, sources, and effects on
capture-recapture estimates of abundance. Canadian Journal of Fisheries and
Aquatic Sciences 58: 1861—1870.

Taberlet, P., J. J. Camarra, S. Grifﬁn, E. Uhres, O. Hanotte, L. P. Waits, C. Dubois-
Paganon, T. Burke, and J. Bouvet. 1997. Noninvasive tracking of the endangered
Pyrenean brown bear population. Molecular Ecology 6: 869-876.

Taberlet, P., L. P. Waits, and G. Luikart. 1999. Noninvasive genetic sampling: look
before you leap. Trends in Ecology and Evolution 14(8): 323-327.

94

Taberlet, P. and G. Luikart. 1999. Non-invasive genetic sampling and individual
identiﬁcation. Biological Journal of the Linnean Society 68: 41—55.

Waits, J. L. and P. L. Leberg. 2000. Biases associated with population estimation using
molecular tagging. Animal Conservation 3: 191—199.

Waits, L. P., G. Luikart, and P. Taberlet. 2001. Estimating the probability of identity

among genotypes in natural populations: cautions and guidelines. Molecular
Ecology 10: 249—256.

White, G. C., D. R. Anderson, K. P. Bumham, and D. L. Otis. 1982. Capture-recapture
and removal methods for sampling closed populations. Los Alamos National
Laboratory, LA 8787-NERP, Los Alamos, New Mexico, USA.

White, G. C. and K. P. Bumham. 1999. Program MARK: Survival estimation from
populations of marked animals. Bird Study Supplement 46: 120—138.

Wilberg, M. J. and B.P. Dreher. 2004. GENECAP: a program for analysis of nuclear

data for capture-recapture population estimation. Molecular Ecology Notes 4:
783—785.

Woods, J. G., D. Paetkau, D. Lewis, B. N. McLellan, M. Proctor, and C. Strobeck. 1999.

Genetic tagging of free-ranging black and brown bears. Wildlife Society Bulletin
27(3): 616—627.

95

CHAPTER 4 — PRECISION AND BIAS: CHALLENGES POSED BY NON-
INVASIVE DNA SAMPLING TO ESTIMATE BLACK BEAR (URSUS
AMERICANUS) ABUNDAN CE IN THE NORTHERN LOWER PENINSULA OF
MICHIGAN
INTRODUCTION

As public participation increases in the management of wildlife and the
abundance of declining species becomes of greater concern, accurate abundance
estimates become essential for the conservation and management planning. The use of
non-invasive techniques to estimate population abundance for rare and secretive species
has become a very popular and effective method to meet management needs. The non-
invasive sampling of black bear (Ursus americanus) and brown bear (Ursus arctos) for
purposes of abundance estimation has gained popularity in recent years. One basic
methodology uses a barbed wire hair snare to collect hair samples from bears at baited
locations (Woods et a1. 1999). Non-invasive sampling has shown great promise in
increasing sample sizes because the negative stimulus of handling is eliminated (Woods
et al. 1999). In addition, more stations can be operated and study designs be spatially
based because the animal is not captured in the snare (Mowat and Strobeck 2000).
However, even though the non-invasive methods for sampling bears show great promise,
non-invasive methods also pose new challenges. In particular, estimating the bias caused
by genotyping error and changes in precision caused by sub-sampling samples must be

rigorously tested.

96

Genotyping Error

Since recent advancements in DNA technology have provided a means to
efﬁciently and effectively estimate the population size of large free ranging animal
populations using a capture-recapture methodology, their application has been increasing
for both conservation and management purposes (Boerson et al. 2003; Woods et al. 1999;
Mowat and Strobeck 2000; Palsboll et al. 1997 ; Sloane et a1. 2000). However, with the
rise in popularity of non-invasive techniques utilizing hair and fecal samples as a DNA
source, researchers have identiﬁed potential drawbacks, including genotyping errors,
which bias population estimates (Creel et al. 2003; Waits and Leburg 2000). Research
examining the cause of genotyping errors found that they result from a number of factors
including the quality and quantity of the DNA (Gagneux et al. 1997). Genotyping errors
result in the misidentiﬁcation of individuals within the population and violate two
capture-recapture assumptions: 1) the correct identiﬁcation of individuals and 2)
individuals do not loose their marks (White et al. 1982).

Paetkau (2003) proposed quality control lab protocols to decrease genotyping
error. These protocols include development of a database around the ﬁeld database,
conservative genotype scoring, culling of samples with incomplete or contaminated
genotypes, re-analysis of samples with questionable scores, and careful scrutiny of
samples with similar genotypes for the detection of genotyping errors. These quality
control methods have shown to substantially decrease genotyping error and provide
reliable genotypes. However, no work has been conducted to evaluate population

estimate bias attributed to using, or not using these protocols.

97

Sub-sampling

Non-invasive sampling methods can often result in large numbers of potential
DNA samples. For example, Mowat and Strobeck (2000) collected 4,245 hair samples
from black bear (Ursus americanus) and grizzly bears (Ursus arctos) in British
Columbia, Canada. It is well known that multiple bears can visit a snaring location
during a sampling interval. Poole et al. (2001) detected as many as 7 different bears
during a 12-day snaring session. Conversely, it is possible to detect the same individual
when multiple hair samples are genetically analyzed. We detected the same individual
black bear when we genotyped 14 hair samples from the same site in Michigan.
Precision of capture-recapture population estimates is a function of the number of
individuals sampled within the population, or the probability of capture (Otis et al. 1978).
The ﬁnding of multiple hair samples from the same individual does not increase the
information for the population estimate. Rather it increases the cost of the project.
Conversely, if multiple bears are detected when multiple hair samples are genotyped, the
precision of the population estimate will increase. Overall, the researcher can inﬂuence
an individual’s probability of capture and the precision of an estimate if not all samples
are analyzed (i.e. sub-sampling DNA samples). The affect of sub-sampling hair samples
must be better explored and through this exploration better decisions can be made about

the number of hair samples to sub-sample.

Non-invasive Research

In the summer of 2002 we began a research project in the northern Lower

Peninsula (N LP) of Michigan to estimate black bear abundance. Because this black bear

98

population is managed by the MDNR through regulated harvest, it was important that we
obtain an unbiased and precise population estimate. However, given the concerns about
bias caused by genotyping error and a limited project budget that did not allow us to
genotype all hair samples, we wanted to examine through simulation the affects of quality
control methods proposed by Paetkau (2003) and sub-sampling hair samples on

population estimates.

METHODS
Genotyping Error Simulation

SuccessﬁJl hunters are required to register all harvested bears and compliance is
estimated at 98% (L. Visser, MDNR, person. comm.; Frawley 2002). We collected
muscle tissue and hair samples from harvested bears to examine rates of genotyping
error. We assigned different identiﬁcation numbers to hair and tissue samples from the
same bear and extracted DNA from hair samples with 5 or more hair follicles 7 or more
days after collection. We then used the quality control protocols proposed by Paetkau
(2003) for both error examination and genetic analysis for purposes of population
estimation. First, we developed a database around the ﬁeld database to minimize the
probability of transcriptional errors associated with data entry. Second, we genotyped
110 hair and 110 tissue samples (1 each from 110 different bears) at 5 microsatillite loci
including GlOX, GIOL, G10D(Paetkau et al. 1995), UarMU59, and UarMU50 (Taberlet
et a1. 1997). Two experienced lab personnel conservatively scored genotypes and
unassigned scores were left blank in the database. Third, we conducted an initial cull of

samples that did not have scores for 3 or more loci from our error analysis. Fourth, we

99

attempted to obtain a genotype using a second round of PCR for remaining samples that
did not have scores assigned. Fifth, we culled samples that did not have scores for 4 or
more loci. We then made our ﬁrst evaluation of error by matching the remaining hair and
tissue samples from the same bear. We matched 95 sample pairs (15 samples were
culled) and found 19 out of 95 (20%) pairs differed and we calculated the error rate per
allele for all loci. We found the mean error rate across all loci to be 0.029/allele.

We then went through the mismatching protocols described by Paetkau (2003),
which involves the identiﬁcation of samples that differ by 1 and 2 alleles. We identiﬁed
mismatches and re-examined and/or reanalyzed samples to verify Scores. We again
examined error aﬁer the mismatching procedures by matching hair and tissue samples
from the same bear and identifying inconsistencies. We found that 4 of the 95 (4.21%)
sample pairs differed and thus the error rate per allele for all loci decreased. The mean
error rate across all loci decreased to 0.008/ allele.

Genotyping Error Simulation Model

To examine the affects of different rates of genotyping error, we randomly created

a virtual population of 2,000 genotypes from the allele frequencies at 5 loci (GlOX,

G10L, G10D, UarMU59, and UarMU50) obtained from 544 bears in the NLP of

Michigan (Table 4.1). We calculated the PIsibs single locus (Taberlet and Luikart 1999)
to be 0.00865 (n = 544) across all 5 loci. We found 10 of the randomly created genotypes
matched other genotypes for all 5 loci in the virtual population of 2,000 and included

these matching genotypes in the simulation to examine the affect of matching genotypes

on the population estimates.

100

 

 

 

 

- - - - 333886 3. - - - -
- - - - 5Q. 2.336 NS - - - ..
- - - - 3.2.3306 om: - - - -
- - - - $3333. ”2 - - 353256 wo—
- - nommmﬂo—d ow. .8232 ﬁe of - - ”mg—N— 23 co—
- - $33386 mi mmmmwwmomd v2 _NNoE mmod mm: Gammonmood X:
- - Som— mcomod v2 $03826 N2 233 _ m _ .c on. _mmmioo—d N2
oomooi _d o2 mmoo _ wowed NS 33$~vod of $.88 _ .o 3: 09: $256 of
~38 _ mo _ .o vm. emwnmaoood om. woomovmmod v3 nwvmogood NM: 3 33 _ cod mm—
vchmwmcﬂ .o NQ wmwvvmc _ m6 3..— SSmevod o3 mvvcwmvvod ow. 33¢me _ .o v3
oomnmoomvd o2 wmmwmwm: .6 mg vmmnmwmm _ .o ”2 mammomwood w: mhwomon _ o6 mm—
oommmoocod 2 _ oaomhvc _ Nd 2: 3333A; on. mmov _ mmovd on. 033336 a:
Ebowmhood 3 _ Ramvowm _ .o E _ nonwvmoood N2 393de v: :83 _ md cv—
mocw _ moo _ .o S _ aonm _ o _ cod 3 _ movwmood om. vwmm _ 33.0 N: coo _ 8236 N2
muaosaonm o_o=< ﬁaoacoum 0.0—7.. zuaosvuuh 0.0.7.. 5:259:— o_o=< mono—:3..."— o_o=<
$323: $22.53 430 :30 X30
mace..—

 

.EwwEﬁS he 3.55:2. .833 505.8: 2: 5 Es“:— vem SP...
6823.3 cmD—ZED .55 imp—2.3: 530 {—39 .xc—G mics—2: Bo. m .8.— mﬂouozco...‘ 0.0:.“ can o_o__< .—.v nigh.

101

We wrote the simulation in SAS (PROC IML; SAS Inc. 2002), whereby we
randomly sampled our virtual population with a capture probability of 0.10 on each of 5
sampling occasions with the assignment of a uniform random number to each genotype
on each occasion. Genotypes were captured on each occasion if the random number was
less than or equal to the capture probability (0.10). To introduce genotyping errors into
the captured genotypes, we assigned each allele in the captured genotypes a uniform
random number. If the random number was less than or equal to the pre-speciﬁed error
rate (0029/ allele or 0.008/ allele) for the particular simulation, the allele was randomly
replaced by a different possible allele for the locus using an additional uniform random
number routine. Genotypes with and without errors were then matched across all
occasions and capture histories formed for purposes of population estimation. We
estimated population size using closed capture estimation models similar to models in

program CAPTURE (Rexstad and Bumham 1992), but coded into the SAS environment

(P. Lukacs, person. comm. 2004). We used the M0 model because our probability of

capture did not change through time (Otis et al. 1978).

We simulated each scenario of genotyping error rate (0.0/ allele, 0.029/ allele, and
0.008/ allele) 1,000 times and examined the mean, 95% conﬁdence intervals, and
estimate bias across all simulations. To examine bias, we calculated the percent relative
bias (PRB; White et al. 1982):

(1.302;) - N)

PBR = x100

where E (N ) is the mean estimated population size across all replications and N is the true

population size, which in our simulation was 2,000.

102

Sub-sampling Simulation

To simulate the affects of sub-sampling, we ﬁrst had to examine the incidence of
multiple bears visiting a hair snare during a single checking interval. To estimate this we
randomly choose 49 checking occasions where multiple hair samples were available from
our hair snaring efforts in 2002 and 2003. We genotyped all hair samples (N=210)
collected from the 49 occasions. The total number of hair samples genotyped for both
years was 210. We then derived the probability of multiple bears visiting a site given the
number of hair samples genotyped at that site (Table 4.2).

We summarized the frequency of the number of hair samples collected from our
2003 hair sample collection data by snare location (N=239) and the number of hair
samples collected on each occasion (1-5 hair sampling occasions were sampled in 2003).
From these data, we derived the distribution of hair samples collected across all snaring
checks and locations (Table 4.2).
Sub-sampling Hair Samples Simulation Model

To examine the affects of sub-sampling hair samples we used SAS (SAS Inc.
2002) to write a bootstrapping simulation that ﬁrst simulated the deposition of hair
samples on a predetermined number of snaring locations and number of checking
occasions, second allowed for the speciﬁcation of the number of hair samples to
genetically identify, and third estimated population size for each scenario of the number
of hair samples genotyped.

To simulate the number of hair samples deposited on our virtual number of snares
we used the frequency of the number of hair samples collected on each occasion (Table

4.2; collected from our 2003 ﬁeld data) in a bootstrapping routine for 239 snares and 5

103

 

 

 

1. _ we: 3. .1 l. 1. l l. 259—.
_ meee e v e e We e We +e.
meee mNeee m N e e . e e e
mNeee mmeee v _ e e e e _ w
_ewee meee e v mNe mN.e e We e N.
Game Nveee m N e We We e e e
eenee 886 m _ n e Nv_.e emNe m3 .e eNve m
eeee Nv_e.e E v - e We We e v
memee emee me e - - mmme mmme mmme m
wage Nemee ee w - - - mNWe mnme N
emcee emnee we _ _ - - - - _ _
meme mane ewe e - - .. .. - e
mung—.9...— »unoseoﬁ $85 .3 67: £9..— ..c..— m 9 m N g 8.55% .55
252......5 gun—om zones—5.....— 8=m .3 .eZ 2.23.2. 2.3.5 .8 62 a... .3 595 he 52

 

.38....er ”Eu—92.9 2......

3. :8...— 8353 h...— e—N 33.33% :8...— eoZBe 83:55:95 A2333: 959...... 93.9.3.— mEJV 2.32.3: “.5953.
m 398.. 2.3.3... 9:35 :2. eMN 5...... 688:8 8.956 .3...— uc .225... a... he b.2555 a... E... 3.? we .35.... :02”

o... .5. 69:93.. 8.953 .5... he .35.... a... .8 12:53.8. £25.... 2.3.... .3 .395... 2.. be ex; Ezﬁanehm .N... 025.

104

sampling occasions. We randomly assigned each hair sample an identiﬁcation number
from a virtual population of 2,000 individuals using the probability of multiple bears
visiting a site given the number of hair samples genotyped (Table 4.2). We then could
specify how many different hair samples were genotyped (e.g. if we speciﬁed that 1 hair
sample was genotyped, then one individual per site was randomly selected, even if
multiple individuals were available to be sampled). We then matched all identiﬁcation
numbers across the 5 occasions and 239 snaring locations to form capture histories for the

purposes of population estimation. We used closed capture models written in SAS to

estimate population size (P. Lukacs, person. comm. 2004). We used the M, model

because our probability of capture changed through time as a result of different numbers
of individuals being randomly sampled (Otis et al. 1978). Each selection of hair samples
(1, 2, 3, 4, 5, 6, and all hair samples) was simulated 1,000 times and we examined the
mean and 95% conﬁdence intervals for each scenario. In addition we plotted the
standard error of the estimates to examine how precision of the estimate changes with the

number of hair samples selected.

RESULTS

Genotyping Error Simulations

Mean estimated population size (E =1 ,992) was slightly less than the true
population size of 2,000 when genotyping error was 0.0/allele (Figure 4.] and Table 4.3).
Mean estimated population size was positively biased with increasing error rate. With an
error rate of 0.029/allele, the simulated estimate was more than 50% larger (N=3,214)

than the true population size (N=2,000; Figure 4.] and Table 4.3). We found that the

105

conﬁdence intervals around the simulated population estimates were small and attributed
to the high probability of capture (p=0.10) used in the simulation (Figure 4.1). The
standard error of the estimates increased as a result of an increase in genotyping error

rate.

 

3500

 

3000

-— — N N

0 VI 0 U1

0 O O O

O o O O
l 1

Mean Population Estimate

(I:
O
O

 

 

 

0.0/Allele 0.008/Allele 0.027/Allele

Genotyping Error Rate per Allele

 

 

 

Figure 4.1. Mean estimated population size and associated 95% conﬁdence
intervals across 1,000 simulations for genotyping error rates, 0.0/allele, 0.008/allele,
and 0.029/allele.

Table 4.3. Mean population size estimate, standard error of the estimates, and
percent relative bias for each genotyping error rate across 1,000 simulations.

 

 

Genotyping Mean Population Estimate

Error Rate Estimate Standard Error Percent Relative Bias
0.0/Allele 1992 3.795 -0.39%

0.008/Allele 2269 4.820 13.46%

0.027/Allel_e 3214 8.615 60.70%

 

106

Sub-sampling simulations

Standard error of the estimate decreased with increasing number of hair samples
genotyped (Figure 4.2 and Table 4.4). The standard error of the estimates decreased by
approximately 50% by the selection of 3 hair samples (SE=12.724) over the selection of
1 hair sample (SE=30.405; Table 4.4). The selection of more than 3 hair samples did not
substantially decrease the standard error of the estimates. For example, the standard error
with the selection of 3 hair samples was 12.724 and the standard error with the selection

of all hair samples was 12.256 (Table 4.4).

 

 

35

30

25

20

15

 

Standard Error

10 _ c. 7 - ...._- z___ _ _ .. . .__. ._. ._.- ..__ -_..._._-._ 2-..-.. ﬂ

 

 

 

1 Hair 2 Hair 3 Hair 4 Hair 5 Hat 6 Hair All Hair
Number of Samples Geneotyped per Site

 

 

 

Figure 4.2. Standard error of the estimate resulting from the number of individual
hair samples chosen to be genotyped across 1000 simulated data sets.

107

Table 4.4. Number of hair samples genotyped per site, the mean of the 1,000
population estimate replicates and the resulting standard error of the estimates.

 

No. of Samples Population Estimate

 

 

Genoyped Mean Standard Error
1 2209 30.405
2 2101 1 5.436
3 2060 12.724
4 2050 12.492
5 2060 12.636
6 2055 12.225
All 2055 12.256
DISCUSSION
Genotyping Error

Genotyping error has caused concern about the efﬁcacy of non-invasive methods
because it has the potential to bias population estimates (Creel et al. 2003). This bias is
caused by genotyping errors which add new genotypes that do not exist in the population.
Creel et al. (2003) found that genotyping errors overestimated wolf population size in
Yellowstone by 5.5 times when using feces as a DNA source. Waits and Leberg (2000)
found through simulation that population estimates were > 2 times larger when error rates
were 0.05/ locus. Genetic marks differ in some aspects to that of more traditional
marking techniques. One difference is in error identiﬁably. For example, if marks are
placed and recorded on animals in the population and recaptures do not match marks that
were not put into the population, the identiﬁcation of an error is made. However, a
genotype with an error can be extremely difﬁcult to identify because a flawed genotype
could be a possible genotype in the population.

Others have proposed methods to limit the amount of genotyping error. Taberlet

et al. (1996) proposed the multi-tubes approach that involves the dilution and genetic

108

analysis of a DNA in multiple “tubes” to obtain reliable genotype scores. One problem
with the multi—tube approach identiﬁed by Paetkau (2003) is that it can lead to increased
cost to conduct such an analysis. Additionally, we found that the limited amount of DNA
extracted from DNA sources such as hair, feathers or feces could also be a limiting factor
when using multiple loci.

The basis behind the quality control protocols proposed by Paetkau (2003) is that
not all genotypes need to be examined for potential errors. There may be genotypes that
are drastically different from all other genotypes and the probability of error at multiple
loci is low, therefore even if an error occurred in one of these genotypes it would have
little consequence to the population estimate. Using these protocols, including sample
culling and the identiﬁcation of mismatching samples, we were able to drastically reduce
the amount of genotyping error within our data. The overall affect of this reduction was a
decrease in the bias of our population estimate as demonstrated through simulation.

However, we did not decrease the amount of genotyping error to zero and it
maybe unnecessary to do so. Numerous statistical models have been developed to deal
with violation of assumptions in capture-recapture estimation models (Otis et al. 1978;
Huggins 1991; Pledger 2000). Currently, program CAPTURE and the closed capture
models in program MARK require the assumptions that animals are identiﬁed correctly
and they do not loose their marks (White et al. 1982; White and Bumham 1999). Others
have been interested in tag-misreads in the application to marine mammals (Stevick et al.
2001). Lukacs and Bumham (in press) propose closed capture models that estimate and
account for misidentiﬁcation of individuals within the estimation models. The

effectiveness and efﬁciency of non-invasive methods to collect large sample sizes of

109

secretive animals will ensure that such methods will increase in importance as a wildlife
management tool. Therefore, we must ﬁnd ways to account for potential bias caused by
genotyping error. Quality control protocols proposed by Paetkau (2003) can decrease the
amount of genotyping error. The estimation models by Lukacs and Bumham (in press)
can provide unbiased estimates of population abundance accounting for genotyping error
within the estimate.

An additional source of bias can exist if too few microsatillite markers are used to
resolve individuals, as found with our simulations. We randomly created a population of
2,000 genotypes with allele frequencies of the 5 loci that we empirically estimated in our

population estimation research in the NLP of Michigan (Table 4.1). We calculated

PIsibs single locus across all loci with our estimation data to be 0.00865 (n = 544). After

we created the random genotypes, we found 10 of the 2,000 genotypes matched for all 5
loci. The overall affect of these matching genotypes within the population is the .
downward bias in population estimates as demonstrated through our simulations. These
ﬁndings are consistent with those found by Mills et al. (2000), where the downward bias
increases with increasing duplication of genotypes. The overall bias caused by matching
genotypes is less than the bias caused by genotyping errors and can be resolved by

including additional loci in the genetic analysis.

Sub-sampling
As wildlife research budgets become increasingly limited, resource allocation
becomes an important issue. The precision of a population estimate using a capture-

recapture methodology is a function of the number of individuals marked in the

110

population reflected in the probability of capture (White et al. 1982). Given a limited
research budget to genetically analyze all hair samples collected, decisions about how
many hair samples to genotype should be based on the identiﬁcation of unique
individuals and an attempt to limit the number of duplicate genotypes. Boerson et al.
(2003) collected 448 hair samples with > 10 hair follicles and choose 116 samples to
derive a population estimate. Additionally, when genotyping error is present in the
genetic analysis, the analysis of additional samples can increase the overall bias in the
estimate. Waits and Leberg (2000) found through simulation that increasing sampling
intensity resulted in increasingly biased estimates even with low rates of genotyping
errors.

We propose using a simulation model to add rigor in making decisions about the
number of hair samples to genotype. We found that the genetic analysis of 3 hair
samples, when multiple hair samples were present, signiﬁcantly decreased the standard
error of the estimates. However, genotyping more than three hair samples did not
signiﬁcantly decrease the standard error of the estimates. These ﬁndings may be
consistent with other non-invasive sampling methodologies where multiple samples are
collected from the same sampling location and suggest that not all samples must be
genetically analyzed.

In our simulations, we consistently overestimated the true population size. We
attribute this positive bias to having very low probability of capture values. As the
probability of capture values increased by sampling additional hair samples, this bias also

decreased (Table 4.4). However, the general conclusions of the simulation still hold. Not

111

all samples must be genetically analyzed from a collection site to obtain a stabilization of
estimate standard error.

Precision and bias are two very important aspects of a capture-recapture
experiment (White et al. 1982). Concerns about bias attributed to genotyping error are
extremely relevant and should be considered. Quality control protocols and new models
that consider misidentiﬁcation show great promise in obtaining reliable estimate of
wildlife population through non-invasive methods (Paetkau 2003; Lukacs and Bumham
.in press). Non-invasive methods continue to be effective at sampling populations to a
point where the researcher is required to make decisions about sub-sampling data. We
offer insight on these difﬁcult decisions and propose that precise estimates do not require

that all DNA samples be genotyped.

112

LITERATURE CITED

Boerson, M. R., J. D. Clark, and T. L. King. 2003. Estimating black bear population
density and genetic diversity at Tensas River, Louisiana using microsatellite DNA
markers. Wildlife Society Bulletin 31(1): 197—207.

Creel, S., G. Spong, J. L. Sands, J. Rotella, J. Zeigle, L. Joe, K. M. Murphy, and D.
Smith. 2003. Population size estimation in Yellowstone wolves with error-prone
noninvasive microsatellite genotypes. Molecular Ecology 12: 2003—2009.

F rawley, B. J. 2002. 2001 Michigan black bear hunter survey. Wildlife Division Report
3364. Michigan Department of Natural Resources, Lansing, Michigan, USA.

Gagneux, 1., C. Boesch, and D. S. Woodruff. 1997. Microsatellite scoring errors
associated with noninvasive genotyping based on nuclear DNA ampliﬁed from
shed hair. Molecular Ecology 6: 861-868.

Huggins, R. M. 1991. Some practical aspects of a conditional likelihood approach to
capture experiments. Biometrics 47: 725—732.

Lukacs, P. M. and K. P. Bumham. in press. Estimating population size from DNA based
closed capture-recapture data incorporating genotyping error. Colorado
Cooperative Fish and Wildlife Research Unit, Colorado State University, Fort
Collins, Colorado, USA.

Mills, L. S., J. J. Citta, K. P. Lair, M. K. Schwartz, and D. A. Tallmon. 2000. Estimating
animal abundance using noninvasive DNA sampling: promise and pitfalls.
Ecological Applications 10(1): 283—294.

Mowat, G. and C. Strobeck. 2000. Estimating population size of grizzly bears using hair
capture, DNA proﬁling, and mark-recapture analysis. Journal of Wildlife
Management 64(1): 183—193.

Otis, D. L., K. P. Bumham, G. C. White, and D. R. Anderson. 1978. Statistical inference
from capture data on closed animal populations. Wildlife Monograph 62: 1-135.

Paetkau, D. 2003. An empirical exploration of data quality in DNA-based population
inventories. Molecular Ecology 12: 1375—1387.

113

Paetkau, D., W. Calvert, I. Stirling, and C. Strobeck. 1995. Microsatellite analysis of
population structure in Canadian polar bears. Molecular Ecology 4: 347—354.

Palsboll, P. J., J. Allen, M. Berube, P. J. Clapham, T. P. Feddersen, P. S. Hammond, R.
R. Hudson, H. Jorgensen, S. Katona, A. H. Larsen, F. Larsen, J. Lien, D. K.
Mattila, J. Sigurjonsson, R. Sears, T. Smith, R. Sponer, P. Stevick, and N. Olen.
1997. Genetic tagging of humpback whales. Nature 388: 767—769.

Pledger, S. 2000. Uniﬁed maximum likelihood estimates for closed capture-recapture
models using mixtures. Biometrics 56: 434—442.

Poole, K. G., G. Mowat, D. A. Fear. 2001. DNA-based population estimate for grizzly
bears (Ursus arctos) in northeastern British Columbia, Canada. Wildlife Biology
7: 105—115.

Rexstad, E., and KP. Bumham. 1992. User’s guide for interactive program CAPTURE.
Colorado Cooperative Fish and Wildlife Research Unit, Colorado State
University, Fort Collins, USA.

SAS Institute. 2002. SAS/STAT user’s guide. Version 8. SAS Institute, Cary, North
Carolina, USA.

Sloane, M. A., P. Sunnucks, D. Alpers, B. Beheregaray, and A. C. Taylor. 2000. Highly
reliable genetic identiﬁcation of individual northern hairy-nosed wombats from
single remotely collected hairs: a feasible censusing method. Molecular Ecology
9: 1233—1240.

Stevick, P. T., P. J. Palsboll, T. D. Smith, M. V. Bravington, and P. S. Hammond. 2001.
Errors in identiﬁcation using natural markings: rates, sources, and effects on
capture-recapture estimates of abundance. Canadian Journal of Fisheries and
Aquatic Sciences 58: 1861—1870.

Taberlet, P., J. J. Camarra, S. Grifﬁn, E. Uhres, O. Hanotte, L. P. Waits, C. Dubois-
Paganon, T. Burke, and J. Bouvet. 1997. Noninvasive tracking of the endangered
Pyrenean brown bear population. Molecular Ecology 6: 869—876.

Taberlet, P., S. Grifﬁn, B. Goossens, S. Questiau, V. Manceau, N. Escaravage, L. P.
Waits, and J. Bouvet. 1996. Reliable genotyping of samples with very low DNA
quantities using PCR. Nucleic Acids Research 24(16): 3189—3194.

114

Taberlet, P. and G. Luikart. 1999. Non-invasive genetic sampling and individual
identiﬁcation. Biological Journal of the Linnean Society 68: 41—55.

Waits, J. L. and P. L. Leberg. 2000. Biases associated with population estimation using
molecular tagging. Animal Conservation 3: 191—199.

White, G. C., D. R. Anderson, K. P. Bumham, and D. L. Otis. 1982. Capture-recapture
and removal methods for sampling closed populations. Los Alamos National
Laboratory, LA 8787-NERP, Los Alamos, New Mexico, USA.

White, G. C. and K. P. Bumham. 1999. Program MARK: Survival estimation from
populations of marked animals. Bird Study Supplement 46: 120-138.

Woods, J. G., D. Paetkau, D. Lewis, B. N. McLellan, M. Proctor, and C. Strobeck. 1999.
Genetic tagging of free-ranging black and brown bears. Wildlife Society Bulletin
27(3): 616—627.

115

CHAPTER 5 — RECOMMENDATIONS TO THE MDNR

The overall goal of our research was to obtain a population size estimate (with
conﬁdence intervals) of black bears in the Northern Lower Peninsula (N LP) of Michigan.
This information would assist the MDNR in the management of the species. We were

successful in achieving an estimate, but it was not without struggle because of the large
study area (36,848 kmz), logistics of organizing ﬁeld staff, and the recent development of

non-invasive techniques. We learned a great many things while conducting this research,
which can contribute to the ﬁiture use of this technique. The objective of this chapter is
to provide the MDNR with recommendations and estimates of resources needed to

successfully execute this technique based on our observations and experiences.

Hair Snare Placement

Based on the large study area with mixed land ownership and heterogeneous bear
habitat, the random stratiﬁed design was most suitable for hair snare placement (Chapter
3). We identiﬁed 239 hair snare locations using this design (Appendix 2.4). These
locations were chosen at random and from the MDNR bait index locations. Because the
frequency of snares was based on frequency of harvest grouped by township over a 5-
year interval, it is important that the harvest ﬁequency coverage be updated prior to the
evaluation of snare locations to identify new townships where harvest was occurring.

One of the strata in our stratiﬁed random design included potential bear habitat.
An evaluation of the overlap between the vegetation types and radio-telemetry data
collected in the months of June and July by the MDNR (Etter et al. 2002; Chapter 2)

showed that the northern hardwood vegetation class had a large number of locations

116

(17.17%). The northern hardwood vegetation class was not included in our stratiﬁcation,
but it should be considered in future stratiﬁcations.

In some situations, after all the stratiﬁcations were made, there was no sampling
area leﬁ in townships where snares were to be placed. In some instances this was
because of the lack of public land within the area. In particular, this occurred in the high
harvest area of northern Alcona county and southern Alpena county (i.e. “club country”).
In these instances, it will be important to make contact with private landowners to obtain

permission for access.

Hair Snaring

Because of the large study area and the large number of snares required to obtain
a population estimate, this methodology required a large number of personnel. We met
these personnel needs through a multi-agency and organization effort (Appendix 2.3 and
2.4). We suggest a training session be held prior to the ﬁeld collection of hair samples.
This will the ensure consistency of the technique in situations where personnel turnover
is occurring. These training sessions should include an overview of the protocols for
both snare set-up, checking of snares, and the collection and storage of hair samples
(Refer to Appendix 5.1 and Appendix 5.2 for a copy of protocols and a list of supplies
necessary for hair sample collection).

We found that the number of snares that a two-person crew could set in a day
depended on the distance and travel time between snares. Generally, 6—10 snares could
be set in 8-hours. We recommend that detailed information about the location of the

snare and driving directions be documented on a standardized data collection form (refer

117

to Appendix 5.3 for a copy of our set-up form). Obtaining a Global Positioning System
(GPS) location for each snare is also recommended.

We used a variety of baits during the 2002 ﬁeld season including combinations of
liquid scents, chicken, sausage, and bacon. Based on our ﬁndings we recommend that
baits consist of l-pound bacon and anise extract contained in a ﬁlm canister. We
recommend that baits be changed every two-weeks if the bait was not taken by a bear.
Replenishing the bait helps to add fresh scent.

It was critical that collection envelopes were pre-numbered with unique numbers
to ensure that samples were not confused. To expedite the sorting of hair samples in the
lab, it was critical that the quantiﬁcation of the number of hairs present be made at the
collection. The categories that we used in 2003 included: 1-4, 5-10, 11-15 and > 15 hairs
(refer to Appendix 5.4 for an example of the collection envelope). In addition, hair
juxtaposition on the wire should be documented to assist in calculating the probability of
hair samples collected on adjacent barbs being from the same bear. If these probabilities
are of little value, this information could be deleted from the form (refer to Appendix 5.5
for a copy of the checking form).

In 2002 we set hair snares with a single strand of barbed wire and in 2003 we set
snares with 2 strands of barbed wire. The double-wire hair snare was used in 2003 to try
to obtain better hair samples (i.e. more hair follicles). It was difﬁcult to determine if the
single or the double wire was more successful at collecting better quality hair samples
because we did not set this up as an experiment. In general, ﬁeld staff showed preference
towards a single wire because some believed that two wires alarmed bears resulting in a

non-visitation. Future research might consist of setting two snares in close proximity,

118

one with a single wire, and a second with two wires. Comparisons could then be made
between the effectiveness, reﬂected in hair quality, of one and two stands of barbed wire.

The number of sampling occasions was an important aspect of this research
because it relates both to the precision of the estimate and also to the valuable time of
ﬁeld personnel. To gain insight into the number of sampling occasions, we examined the
occurrence of new genotypes across the 5 hair snaring occasions in 2003. On average,
we identiﬁed approximately 33 new individuals per week of sampling (range = 30—37).
The largest number of new individuals was 37 identiﬁed in the ﬁfth week of sampling in
2003 (Table 5.1).

Table 5.1. Number of new bears and number of previously identiﬁed bears for each
sampling occasion conducted in the summer of 2003.

 

Sampling No. of New
Occasion Bears Identiﬁed

l 34
2 30
3 31
4 31
5 37

 

These data would suggest that increasing the number of sampling occasions resulted in
the increased identiﬁcation of new individual bears. Identifying more bears increases the
precision of the estimate as demonstrated from our simulations conducted in Chapter 4.
If the MDNR desires to decrease the number of sampling occasions (e.g. from 5
occasions to 4 occasions), further work should be conducted to examine the impact of

such a decision with regards to population estimate precision.

119

Hunter Harvested Hair and Tissue Collection

In both 2002 and 2003 we had great success collecting muscle tissue samples
from harvested bears. We attribute this success in part, to holding training sessions
where collection protocols (Appendix 5.6) were reviewed. We recommend that these
training sessions continue. The MDNR has established protocols for collecting premolars
from all harvested bears for cementum ageing. Cementum ageing is conducted at Rose
Lake Wildlife Disease Laboratory. In the process of collecting all bear registration
materials for the MDNR, we experimented with extracting DNA from tooth samples.
Extracted teeth typically contain small portions of tissue, blood and tooth pulp. We had
great success (99.3%) in obtaining DNA from tooth samples and based on this success, it
may not be necessary to collect muscle tissue samples from harvested bears. However, in
the cases where tooth samples cannot be obtained, it will still be necessary to collect
muscle tissue samples as a source of DNA. The hair sample collected from each
harvested bear was used to access genotyping error and should continue to be collected.
It was critical that all collection envelopes be pre-numbered individually with a consistent

number for each collection set.

Sample Selection

We had greater than anticipated success at collecting hair samples from hair
snares in both years (Chapter 2). This success required the development of sub-sampling
protocols because of our limited budget to genotype hair samples (Chapter 3). We
examined the affect of this sub-sampling on population estimates and determined that the

standard error of the estimates can be drastically reduced if a maximum of three hair

120

samples (each individual barb equals one sample) are genotyped at a location (Chapter 4).
Therefore, we recommend that future sub-sampling protocols include randomly selecting
only three hair samples when 3 or more hair samples are present at a snare.

In both years, we genotyped all harvested bear muscle tissue samples. One reason
that we sub-sampled hair samples was because there was a given probability that a single
bear deposited hair on multiple barbs and the genetic analysis of all the hair samples
would duplicate genotypes (Chapter 4). This duplication would be redundant for
estimation purposes and thus waste resources that could be used to identify new
genotypes. All harvested bears were assumed to be unique. Therefore, duplication was
not an issue. In addition, the sample size of harvested bears was much greater than the
number of different bear identiﬁed through the summer collected hair samples. We

recommend that all the harvested bear samples continue to be genotyped in the future.

Genetic Analysis

We changed our genetic analysis methods between 2002 and 2003. The result of
these changes was an increase in genotype ampliﬁcation success from DNA samples
collected in 2003 as compared to samples collected in 2002 (Table 5.2). We believe the
difference in ampliﬁcation success was related to three major changes. First, we
extracted hair samples within approximately 7 days after their collection. Samples
collected in 2002 were extracted as many as 6 months after their collection. Roon et al.
(2003) demonstrated that ampliﬁcation success was related to the time between collection
and extraction, with longer time periods decreasing ampliﬁcation success. We suggest

that the short duration (7 days) between collection and extraction continue.

121

 

 

 

 

 

 

..\..S.~w when .32: 33,—. ..\e 5.3 emve en: 33,—.
fee _ .3 we: Nmem 33225 X.» _ .ee mum. NNE me _ O
$8.3 Ewe— Nmem emDESD $3.3 eam. NNE 4e _ 0
$2 .em Eve— NmeN 4e _ 0 games ed: Ng— Qe _ O
$8.va em: Nmem QeTU $2.8 em: 32 220
$2.; 82 Nmem X20 oewﬁwe NNS Ng— X20
69,830 8_o=< 13.830 8_o=< 01:8...— 334 @3830 8_o=< "8:930 8_o=< 03.85— maes—
338.— .3 .32 .3 62 33.3.— .3 .oz .3 62
nee" 28> Nee" 30>

 

 

.mﬁbans 33:0» 2.8.: was ._3: ween can Sea 2: 5 com: 282 .38 .3.
.833 6383: 085 3 «uni—3.8.. 2: can .833 coinage .3 8:53. .8_o=a 05.82. .3 8.3::— o... .3 b.2555 .N.m 035.

122

Second, we only extracted hair samples with _>_ 5 hair follicles in 2003. In 2002
we extracted any hair sample, regardless of the number of hair follicles. Goosens et al.
(1998) found that genotyping error was a function of the number of hair samples in the
extraction. Error rates were greater with fewer follicles. Further work needs to be done
to examine whether even better genotypes could be obtained by choosing hair samples
with greater numbers of follicles (e.g. only choose samples with 2 10 follicles) and the
impact of such protocol changes on the number of bears detected.

Third, we used quality control protocols proposed by Paetkau (2003) to increase
genotyping success and reduce error (Chapter 3). We explored the affect of these
protocols on estimate bias and discovered that the mismatching procedures drastically
decreased estimate bias. We recommend these protocols be used in the ﬁJture, but more
work must be conducted to examine if a more stringent culling protocol will further
reduce genotyping error.

In 2003 we found that 4 of the harvested bear genotypes matched. This ﬁnding
may suggest that we used too few loci to resolve individuals because all harvested
individuals are assumed to be unique. Future work may consider adding additional loci,
or at least the analysis of all matching genotypes with an additional locus. The overall
bias of matching genotypes of unique individuals within the population estimate was
negative and we found through simulation that this bias was very slight (Chapter 4, it

underestimated the population on average by 8 individuals).

123

Population Estimate

The effect of genotyping error on population estimates was very dramatic with
positively biased population estimates (Chapter 4). Even with quality control protocols
proposed by Paetkau (2003), we still had a non-zero error rate within our genetic data.
Therefore, we recommend the estimation models proposed by Lukacs and Bumham (in
press) to account for any remaining genotyping error. It is possible that these models will
be incorporated in program MARK in the near future and thus easily accessed (P. Lukacs,
Colorado State University, person. comm.).

Closed capture models in program MARK or the Lukacs and Bumham (in press)
models allow the user to build models using a design matrix (White and Bumham 1999).
The design matrix allows the user to account for a behavioral response, a time response,
or heterogeneity in capture, for each capture occasion within the data. We found the
ﬂexibility of the design matrix to be critical when using harvest as a different capture
method than that of summer collected hair samples. Any number of models can be
developed, but a reasonable set of four models should be created a priori based on the
biology and ﬁeld knowledge of bears in the NLP. We found a strong behavioral response
in our data, which might be expected when we use bait that offers a food reward. We did
not ﬁnd as strong a response related to time, but this factor should always be considered
in building models in the future. Another advantage of program MARK is the ability to
use individual covariates (White and Bumham 1999). We used a sexing primer for the
DNA samples collected in 2003. We used sex as a covariate in the estimation models

and found that these models were not substantially better than the models without sex. In

124

the future, we recommend that a sexing primer continue to be used because they can

possibly contribute to the estimation models.

Personnel and Cost Summary

The amount of personnel hours and cost of this project is important for assessing
its utility. Therefore, we summarized cost and personnel hours of the project for the
MDNR’s consideration. We conducted work in both the summers of 2002 and 2003. We
used the summer of 2002 as a learning year to improve our methods and better
understand what such research would require. We used the data collected summer/fall of
2003 to generate our population estimate using what we had learned. Therefore, our cost
and personnel summary reﬂects the summer/fall of 2003 because it better represents what
the MDNR will need to invest to reproduce this methodology.
Personnel Summary

We divided the project into a series of tasks and sub-tasks to examine the
personnel hours (Appendix 5.6). The task that required the greatest number of personnel
hours was the set-up and checking of the hair snares, with 2,800 estimated hours
(Appendix 5.6). The task that required the least amount of time was the identiﬁcation of
hair snaring locations, estimated at 40 hours (Appendix 5.6). Overall, the estimated
personnel hours were approximately 5,1 10 hours (Table 5.3). This summary includes the
hours of personnel from a variety of agencies and organizations.
Cost Summary

In addition to the time of personnel, there are a variety of monetary costs required

to execute this methodology. We summarized these expenses as if they were above and

125

Table 5.3. Summary of the personnel hours to perform the necessary tasks as

reﬂected in the portion conducted in 2003/2004 of this project.

 

Task and sub-task (indented portion) to be performed Time Required (hours)

Snare location identiﬁcation
Summarize frequency of harvest by township
Update or change stratiﬁcations
Choose snaring locations
Organize ﬁeld supplies and distribute
Stamp envelopes
Purchase supplies
Make copies of ﬁeld forms
Organize supplies
Hold training session
Set-up and check hair snares
Set-up snares
Check snares
Collect hair samples from all routes and take to MSU
Collection of hair, tooth, and tissue from harvested bears
Sort and extract DNA samples for genetic analysis
Sort hair samples by number of follicles
Extract DNA from hair follicles
Extract DNA from tissue samples
Genetic analysis of DNA samples
Organization of samples
Preparation of database
Initial genetic analysis
Re-analysis to ﬁll non-ampliﬁcations
Veriﬁcation of genotype scores
Identiﬁcation and re-analysis of mismatched samples
Data analysis and population estimate derivation
Identify the genotyping error rate
Summarize data
Create capture histories
Generate population estimate
Total

126

40

80

2,800

100
70
320

1 ,600

100

5,110

beyond the cost that the MDNR would incur, if they duplicated the research themselves
and did not use the services of a graduate student. Therefore, the labor associated with
the MDNR doing the work was not included in this summary. We divided these costs
into a number of expenses including, ﬁeld supplies, harvest collection supplies, and the
labor and supplies for the genetic analysis. The largest expense was the labor (estimated
time of 1,600 hours) for the genetic analysis estimated at $23,205 (Table 5.4). The
estimated overall total cost of supplies and genetic analysis was $38,945 (Table 5.4).

Table 5.4. Approximate cost for each expense for the 2003/2004 portion of the
project.

 

 

Expense Cost

Field Supplies

Barbed wire, staples, tools, nails, ﬂagging, etc. $500.00

Bait (Bacon and Anise) $1,372.00
Harvest Collection Supplies $100.00
Genetic Analysis

Labor $23,205.00

Laboratory supplies $15,740.00
Total $40,917.00

 

To place this cost in perspective we examined the laboratory cost (supplies and
labor) to identify a single bear based on our results from 2003. We identiﬁed 544
different bears from the genetic analysis of 1,026 DNA samples. Using the number of
bears identiﬁed and the laboratory cost, we estimated the genetic cost to identify a single
bear in 2003 to be $71 .59/bear identiﬁed.
Comparison to Current Methods and Further Work

Currently, the MDNR makes decisions about bear management in the NLP based

on a population model that is veriﬁed by a bait index. The bait index does not provide an

127

estimate of population size, but provides a trend in the population. One of the major
advantages of the bait index is that it is inexpensive to operate. However, population
models can only be as accurate as the information used in their creation and no previous
estimates have been derived in the NLP of Michigan. Therefore, the bait index is an
inexpensive methodology, but it did not address the needs of the population model to
have an initial population size estimate.

One limiting factor of the non-invasive methodology was the cost and personnel
requirements. In the future the MDNR may consider using the non-invasive hair snares
as a bait index. Having this methodology established would allow the MDNR to collect
hair samples on a given interval to estimate abundance. This estimate would verify the
index and the population model and provide an estimate of abundance. For example, the
MDNR could continue to operate the 239 hair snares across the NLP and document the
number of hair samples on the wire across a series of checking occasions. Instead of
genotyping the hair samples, the number of hair samples could be used as an index using
our data about the number of unique bears found when we genotyped all hair samples
from a snare (Table 4.2). On a three-year interval, the hair samples could be collected
and genetically analyzed to estimate population size. Further work would need to be
conducted to establish the hair snares as a bait index, but the basic methodology has

already been identiﬁed through our research.

128

LITERATURE CITED

Etter, D. R., L. G. Visser, C. M. Schumacher, E. Calrson, T. Reis, and D. Rabe. 2002.
Black bear population management techniques. Final Report. Federal Aid in
Wildlife Restoration Project W-127-R-17, 18, 19. Michigan Department of
Natural Resources, Lansing, Michigan, USA.

Goosens, B., L. P. Waits, and P. Taberlet. 1998. Plucked hair samples as a source of
DNA: reliability of dinucleotide microsatellite genotyping. Molecular Ecology 7:
1237—1241.

Lukacs, P. M. and K. P. Bumham. in press. Estimating population size from DNA based
closed capture-recapture data incorporating genotyping error. Journal of Wildlife
Management.

Paetkau, D. 2003. An empirical exploration of data quality in DNA-based population
inventories. Molecular Ecology 12: 1375—1387.

Roon, D. A., L. P. Waits, and K. C. Kendall. 2003. A quantitative evaluation of two
methods for preserving hair samples. Molecular Ecology Notes 3: 163—166.

White, G. C. and K. P. Bumham. 1999. Program MARK: Survival estimation from
populations of marked animals. Bird Study Supplement 46: 120-138.

129

APPENDICES

I30

.3080 0:: .3 A§m~Av 880: 8:05:90
8 m: :000: 8 0388 08:3 :8 .:000: .0388 8: .0388 88:0 - 808888: 08 8
0888088 8:00am w:_3o=o.: 0:: 08:3 ”$2180— 8 08020 :308 ::_3 :88: 8835

8.8.0 802200: 82:: 058% 0:08 8:3 .88 3:: =8: :0: 80:0 ”.8980 0::
08880:: 8808 80:200.: 08:3 ”gee :80: 8 088—0 :308 ::_3 :88: 82:5

.280: :_ :00: em 8008 :0: on
388:8 8:: .m:oo:m 88: 8880. ::_3 388:8 .:o_8:0w0> .8003 3: 8888085

8:8,:
8 88:: 002-808 .8088 A: 8888088 :o_8:0w0> 888:8: 8832:0802

8888:; 088 8 :03 8 88:08 038 88 80:0 88:0:—

2880 00:: 0888 53 8:033:20

8:828: ..w.0 - bow080 8:8 0:: :_ 88:08 083 8820 8:88.: 28 8:8

:3: .3 3:080:88: 0:: 8808 80?. .:o_8:0w0> 80.8888: 080m :8800 88
88 :_ 22888 088.88 .3 830 0:03:88: 230 o\oem 8:: m8. ::_3 808 8:5

88:: 8:82 08:80::- :0 0388:8883 m_
08:88 0:: 8:: cm :o_8:0w0> .3 80.0 380853 83-3-83.— 8 ::_3 88:80 83:8:
8388: 8.8: :08 £8888 088-88 .3 830 0:03:88: 230 o\oem 8:: 880:0

80338: 80882

:88“: 803203: 8805-805 8:5

22:28

88330 0:008:82

888:: 2:08:80

8:5 .0888: 30:

8:5 388::— :8:

 

355.39 0530.50 :55

.CowSaU 0053030 6:81—

 

.80583: 1:: 8:3ﬁ30 00: c5:— \.:0>e0 .8:— 5888305 .881— :U nee: .:.~ 0580.53:

131

.>:0:00
:83: 0:: 888800 0:3 0.00m 08:3 ”$0.180. 8 0800.0 :308 :83 :83: 0835

03080
:83: 0:: 888800 0:3 00: 0:0:3 ”$3. :80. 8 088.0 :308 ::.3 :83: 0835

.a:0:00
:83: 0:: 888800 0:3 08:3 0:0:3 ”$3. :80. 8 0800.0 :308 :83 :83: 0835

.088 8:08:38 083: 088% 038 8:3 88 0:8 .8: :0: 800 ”.3080
0:: 088800 8.00% 80:0:300 08:3 :33. :80. 8 0800.0 :308 :83 :83: 083 5

888.0 :308 :0 Son A 30:0:800 0:3 0.8.: :83 :8880008 0. 80 08
.083 0.00% ::.3 008888 0. 80 0:0:3 2x09. :82 8 088.0 :308 :83 :83: 0835

.xq0:00
:83: 0:: 088800 8.00% 080 0:0:3 “$00 :80. 8 088.0 :308 ::.3 :83: 083 5

8.00% 80 :0 0:3
::.3 008888 :08: 0:83 3:08 8%: 800.08 m8>00 :83: 0:: 088800 8.00%
008.0: 08 .:0:.: 0:83 80%: 0:0:3 ”foo: :80. 8 0800.0 :308 ::.3 :83: 0835

.08: 00: :0 08:3
:83 000308: 0883: :0 0.00.80: :83 000308: 80::3: ..w.0 - 08:300 08% :83
008888 :83: 000308: :8::3: ”$00 :80. 8 0800.0 :308 ::.3 :83: 0835

08: 0.00.. 0835

08.. 00M.

08: 08:3

:83 n. 808:.30 8:5
08.. 0.08.380

80

:0:.m\:0%<

88390003081 :8::32

 

0088.85 00580.30 0::-—

3.8030 0058030 0::-—

 

0.2.00 ._.N 80:09.3.

132

50800.0: won 0:0
0007:0502 00 00:0 020000 00.50. :00:w:0>0 000070005 03 00005800 ”0:020:00
.03 .00 0300205 0:00 00: :20? 0:0 .:0_:000w0> 2030000»: ._0 000000 wEtoaqsm
.00 030000 00 0: 500:0 m:0_ 000.050 0:0. 05 2600 ._0 .000: .00 00:03 505 000 :<

.0030 0:0 30:05 00 :03 020000 025 00002000 an 00005800 ”0:020:00
:03 .00 0300205 0:00 00: 5023 0:0 .:0_0000w0> 0&0006»; :0 0:00.00 30.80000

00 030000 00 00 @005 m:0_ 000.050 0:0— 05 0>000 :0 .00: J0 00:03 505 000 :<

.0003 00.20. 0:036. 058% 00080050 ~30 SE :00 :0: 0000 ”gov

:05 000. 00>00 00: 0:0 :8 000‘: on :05 002 :00000w0> x0003 3 00005800 ”0:230:00

:03 .00 020205 0:00 00: :02? 0:0 .:00000m0> 23:00:03 :0 000000 wEton—asm
00 030000 00 00 @005 w:0_ 0000.50. 0:0. 05 0>000 ._0 .000: .00 00.03 50$ 0000 :<

._mom :0 00:03 05 .00 000050
05 0>000 m:_0:Sm 3:03 000000000: 0:0:m_m._00-:0: 0:0 0:806:00 .23 ”0:000:00
:03 00 0>_:00_0:_ 0:00 00: :02? 0:0 .:0_00:0w0> 052000»: ._0 030000 mitoaazm
00 030000 00 0: $005 w:0_ 000.003 0:0_ 05 0>0n0 :0 .000: .00 00:03 505 00.0 :<

.2050 000000 05 0:05:80 .6 800.00 0:0 603%
0::? .00000 0::? 00 in: 0 00055 ”$00 0000— :0 0500—0 :308 £05 3000.: 0:033

seam :000w:0>m 0300170005 0:0—304

020m m:0=0_00Q 0300170005 0:00:04

aim 0:233 .200

30000.2 0030:0003 0:0w00Em

:E\00:.am\:0000

 

55.509 0020030 0:01—

b0w0000 003.360 00.3

 

.0080 .3 5050.3

133

$0080 0:: 00:05:80 :0000 05:3 ::0:::0: 0:0:3 ”0:550:00
:03 00 05:00:05 0:00 00: :0_:3 0:0 .:5:0:0m0> 0:500:03 :0 05:00:00 w5::000=0
00 030000 0: 0: :w00:0 w:0_ 000080 0:0: 0:: 0>0:0 :0 .:00: .:0 :0:03 :53 00:0 :<

$0050 0:: 0805800 02:00 5003 0:0:3 ”0:550:00
:03 00 05:00:05 0500 00: :053 0:0 .:0_:0:0m0> 003020»: :0 05:00:00 w5::000:0
.8 030000 0: 0: :w00:0 m8: 000080 0:0_ 0:: 08:0 :0 .:00: .:0 :0:03 :53 00:0 :<

0:00:00 0:: 00:05:80 050 500.0 0:0:3 ”0:550:00
:03 00 05:00:05 0500 00: :053 0:0 .:0_:0:0w0> 0:200:05 :0 5:080 w5::000:0
.8 030000 0: 0: :w00:0 w:0_ 000080 0:0. 0:: 0>0:0 :0 .:00: .:0 :0:03 :53 00:0 :<

0:00:00 0:: 0:05:80 0:

0:558 8800 500.0: 0:0 0:008 .:00 .50 00 :000 008000 .8 05:: 0 0:0:3 ”0:550:00
:03 00 05:00:05 0500 00: :053 0:0 .:0::0:0w0> 0:500:03 :0 050000 w5::000:0

00 030000 0: 0: :m:0:0 w:0_ 000080 0:0. 0:: 0>0:0 :0 .:00: .:0 :0:03 :53 00:0 :<

.0008 0:0—:03 00:00:00 2.50000 08::

8:8 0:0 0:5 :00 :0: 0000 0:000 0 :000. :0 00 550: 0:30:: 0 05:000: 000:: 0:0 .0300
:0:: :0:00:w 00 :0>00 000:00 0 :53 .0808 _0_::0:00 00003 .3 0305800 ”0:550:00
:03 00 05:00:05 0:00 00: :053 0:0 .:0_:0:0w0> 05:58:00: :0 05:00:00 w5::000:0
.8 030000 0: 0: :m00:0 w:0_ 000080 0:0_ 0:: 0>0:0 :0 .:00: .:0 :0:03 :53 00:0 :<

:0000 05: 3 E0582

00::0m 5005

050 5000 0:030:

 

00:55.00: 0023080 0:04

000308: 0:0—30: 000:2
0:000? 00:00:00 :0::O
f0wWEU 0033080 0:01—

.0.::0U .—.~ 50:00.3:

134

 

0:250:00 050 0: 0:0 :00: cm :0:: 000. .:0 ::w_0:

0 0: 03:00 .:_::0:0E:00 0:0 :0:: 0050000 00:: 0:805:00 :: 00:0:5600 ”0:250:00
:03 .:0 03:00:05 0500 00: :033 0:0 .:0_:0:0w0> 0::.::00:0.:: :0 05:0000 w:_::000:0
.:0 030000 0: 0: :w:0:0 w:0_ 000:::0 0:0_ 0:: 0>0:0 :0 .:00: .:0 :0:03 :53 00:0 :<

0:30. 0w:0_ :0 .000::00 0.0m 05:00 00 :000 .000:: :0
03:30 :53 0:30. :00 3000—0 .:0 000:0 :: 0050:: .000:0 :0::: :00: :0 :5 00000505

00:00 .0:0>_: .0.:0>:000: .0052 ..w.0 - ::000:0 :0_:0:0w0> 0: :53 :0:0300 000:<

000.0 0:0: :0::0m 0:: 0:0 000:: .:0 :0:50 :00 030805000:
0:008 00:0 :0 : .:5:0_:0 050: 00000.0 0:0—:03 0:0 _0:::_005:w< .500: 0000008 .:00
0:0: .0:00 ..m.0 .. :0>00 :0::0 :0 :0_:0:0w0> 00: 00:0 0:: 00 m: :0:: 000. :033 :5 0:0:

.:00:00 0:: 8:03:50 00030:0: 0:0 3:300 0:030. .:0 0::: 0 0:0:3 ”0:350:00
:03 .:0 0>_:00_0:5 0:00 00: :033 0:0 .:0_:0:0w0> 055::020? :0 05:00.00 w:5::00000
00 030000 0: 0: :m00:0 w:0_ 000050 0:0. 0:: 0>0:0 :0 .:00: .:0 :0:03 :53 00:0 :<

:0::m :00:w:0>m 00>00:-0_000Z 0:0—30:

0:0—000:0 :0:::

:0:03

_ucmd cot—um

0003002000300 0:230: 000::):

 

00:50:09 00:30.60 0:0:—

.:.:0w0:0U 0090:2000 0:0..—

 

.0.::0U ._.N 50:09:.

135

00:0 0:0: 0:: 00 $3 :0::

0:05 0:0>00 :0>00 :00030 :00000 00:: 0:0 030::0 .:0003 00 80033500 0:0. 000: ,0. .:5000Q 30:30::m 0:0.03
00000.50 3500200: :00 00:50::502 000500 0303000
000:: 00:00:30 :00 :30:w :0: 000:: :0003 .::00::Z\00:0:0:_>\00:0:0:O

0:::300w0 00000035: 00=0-:0: :0 05000 :00
000: 0:0_ 00020:: 00_< 6:0: 00.003 .000 00000020 :00000 :00 000: :000:0m0> 0:330:00 00000035: 00=0-:0Z 000:0 0w0:00

0:003:00 .E00 .000 030: :5 08:20 000:0 32:0 00 0:050:00 :000:0m0> 000:0 300
00:0:0w0> :0:0::00 $3 :0:: 000_ :53 :000:00:0 00:0 :00 00:0 00:0 0:0: 0:03:00 00:0:0w0>..:0Z

.0:0_ m:_0_:00 0:0 0000: _0>0:w 0:0 00>00
w:_0:_0:_ 00:0:0:::0 000E-:0E $3 :0:: 000_ 0:0 .000. :0:: :0:00:m 05 00:0 0:04 :0::: .:50:0::0 304

0:0_ w:_0_:00 :0 .0000: 0:00:50 :0:: :0::0 0:000:06

 

000E-:0E E0:0 000:: :0>00 0003:0085 0000 £03 :0:: :0:00:w 00:0 0:0: c035 35008:— :05:
.0:0_ 0:0—:00 :0 00000 :04 0500000000
.0.:03::: 05020:: 005.80 00050 :353 0:0. 0003:0080 _ 0::00:_<

:00300: 0030:0300 0:04 .:.:0w0:00 009.5300 0:04

 

00003000 0:0 002%0200 00: 0:04 0:0>00 0:0. 0300:030:m 0:04 040.9 02K .N.~ 000000000

136

00:0 0:0_ 00 .0000. 0000008 :0:03 :000 00 :00:000:0

£000.08 9 $00.38 52: 0005:
0:0: :00 0: 0:0 5:00 05.00080 ..::0w0:00 :0 :0 ::0 0::: m5 0 :0::00:0 00:2
0. 0 0 0 . : . .= 0 0 0 .

.:00:00 0:: 00 .0000 0000008 000:: 00800300 00 :00:000:0

:00:00 0:: 00 .0000 0000000 0:005:00 0:0—0: 030.5: 00 :00:000:0
:00:00 0:: 00 $00 000088 00:50 00 :00:000:0

:00:00 0:: 00 $00 0000000 000:: 000005000 00 :00:000:0

:00:00 0:: 00 0000 0000000 0050000 0_w:_0 :0::0 :00 00 :00:000:0
:00:00 0:: 00 $00 0000000 :000< 00 :00:000:0

:00:00 0:: 00 .0000 0000008 0:00 00 :00:000:0

:00:00 0:: 00 .0000 0000008
:85 30:0 > .:00:“: .:0< 05:3 .0003000m .:000m .02002 00 :00003500

:0>00 :0003 00 0:050:00 00:0 0:0. 00 $3 :0:: 0004

:0:03

00:00 00002 0:203
20:80 050.0: 8::
0:005:00 0:0.0D :0::0
00:50

000005000 0:033 0000::
000005000 0:203 :0::0
0000500000. :000<

:00050000< :00

:00050000< 000300: E0:::02

0:03000 00000030:

 

0000:0009 0030:9000 0:04

.::0w0:0U 00:30.60 0:04

 

.0.::0U .N.~ 50:09::

137

 

8080 0:800:00:

0000 0:00 .00 02000.0 0_ 00>00 0004

:00 0000 00 0000 .00 0:000:00 0000000 0_ 00>00 0001—

000000000 00000 000 000_ 0::—0.0 000 0000003 000008002

00>00 0003.000 .00 :8 0000000 000000w0> 000w00000 .00 000000000
00>00 0003.000 .00 $00 0000000 00000 000.32 .00 000000000

00>00 0003.000 .00 $00 0000000 000000m0> 000000 w00000 .00 000000000

0000.8 2 0.08000 0.20 8052
00002000 00 0000000 .00 000000000 .b0w0000 00000 000 000_ w00=0¢0 000 0000.0 00002

000000 000 .00 300 0000000 00000 0000000000 .00 000000000

000000 0.0 ,00 goo 0000000 0000 000000000 .00 000000000

00070 022

0.000 000000m

:0m\000m

000003 000000-002 00002
00000 >9 000000Em

020m 000.304

000030~ w0000E

000000 000:2 0000504
0000,.— 000000000 0000504

0000000 000000000 0:0—301—

 

0000003— 00900060 000‘—

000ﬂ0009 00909009 0004

 

0.0000 .N.N 00000000..

138

 

 

 

139

 

2 _N 0.080

50082.0 o 0 0 .880 NES 0. NSN 8

500820 0 v 00280 30000 0. NSN N0

5003200 N N 30080 08000 00 NOON a
0050 o 00 NN _ 350 03000 00 NSN 0<
000.50 0. 0 00:30 £800 0. NSN 0<
00.50 0 N 00330 000 .00 00 NSN 2
000.00 0 N 000N000 080% 0: NOON 0<
00.0.0 0 N 000030 0:30 00 88 2
00:00 N N 00:30 000000 00 NSN 0%
000.0 N _ 08300 5200 00 NSN 2
02:50 N _ 00280 :0 :0 00 NOON E
0000.00 00.00.30 00.00.00 0:05 .02 $000002 w0000m 000N 000> 0:002 005 000.5

000: .02 0000004 42.54

 

 

0000004 03000 ._0 0000mm 0m 00.002000. 0.00404
.000000 000 M00002? 0000000090 000000u00000 0000 00.0 000000. 00.5 00000 0000 000 0000 000000 000 000
00000=00 00.00000 .:0.— .00 0000000 .0000.— 00 000000000, .00 0000000 .000000. 42.0.: .0000 .0500 00000 .00.. .00 0004 .m.~ 000000040

 

 

 

140

 

20 mm 0.0.0.0.
0000?. 2 m 02.3.00 on 2 03. 2 88 2000—0
0000_< 00 0 00m _ m3 m0 _ con 2 meow 30000—0
0000?. mm 0 30030 Somwh 0. Sam 30000—0
000014. mm 0. $8030 N0. _ men E 83 30000.0
0000_< _ _ mmw _ 30 0330 0. Saw 3008.0
0000_< o o 000N300 0000mm 2 Noon 0 0000—0
0000_< o o 030000 $8? 0. Noon _0000_0
000000 c o 0000.000 0 _ 83. 0 _ Noam 00000000I0000_0
000000 c o _ 2.030 0390 0: Sam m0000000I0000_0
000000 m _ m m _ Snow 0083. 0. N03 m0000000l00020
000000 o c 80030 $0900. 3 Noon _0000000I0000_0
000000 0 _ 3.92% 2.003. S NOON o _ 000000900020
000070 o o 000300 mmommu 0 _ moom _ 0000000I0000_0
000000 c c 00330 ohommw S Noom o _ <000000I0000_0
0000.00 00002—00 00.050@ 0005 .02 hat—0.002 M0000m BEN 0000» 0802 3% 000.5

0003 .02 0000004 420.:

 

 

00504000...— 0042 00000003— .00502 .00 00005.00 00
.0080 .0.N 005000.

 

 

 

141

 

00. 00 0.0.00
050003 0 0 0000000 000000 0. 0000 000000003
080003 0_ 0 0000000 00020 0_ 0000 00200003
0.0.0003 00 0 0000000 00:0 0. 0000 00.00003
0.00003 0 0 00_ _000 0000 _ 0 0_ 0000 00.00003
0.50003 0 0 000000 000000 0_ 0000 _ .0505...
08.0003 00 0 0000000 000000 0_ 0000 0050003
0000000 0 0 0000000 000000 0_ 0000 00
9.0.0000. 0 0 0 .00000 000000 0_ 0000 00.
8.08000 0 0 000000 000.00 0_ 0000 00
00000.00 0 _ 0000000 000000 0_ 0000 00
0003.00 00 0 0000000 _00000 0_ 0000 :0
8000002 0 0 0000000 000000 0_ 0000 00
8000002 0_ 0 000000 00:00 0. 0000 00
08.00002 0 0 .0000 00000 0. 0000 00
00000002 0 0 0000000 02 _00 0_ 0000 00
8000002 _0 0 00 _0000 000000 0. 0000 00
8000002 0 _ 0000000 000000 0_ 0000 00
8000002 0 0 200000 000000 0_ 0000 _0
50:00 00:00:00 00.00.00 0:05 .02 £20.82 003000 000N 00> 0502 000 0.0000
00: .02 000300 200

 

0050 0.000 005000 000.5003— _0.:=0Z .00 000.020 00
.00000 .00 55000

 

 

 

 

 

 

 

 

 

 

0. m 0.000...
30.00400 0 0 0000.00 0.00. .0. 0. 800 00
000000003. 0 0 0000000 . .03.. 0. 0000 my.
00.050003. 0 0 000 0000 00 m 000 0. 0000 my.
000000003. 0. m 0000 . 00 00000.. 0. 0000 .~.
50:00 0800:00 00.00.00 00.0.0» .07. $00207. ”0.00.0.0 000N 00> 00.07. 00.0 0.0000
0.0: .07. 000000.. 5...:
00.....0 0.0.... 00.03-009.303. .000007. .0 000.0000 0..
0 m 0.000...
0.3.0.0 0 0 0000000 0 . 0m .0. 0. 0000 000
0.3.0.0 0 0 0000.000 .000... 0. 0000 .x
000.0 o 0 0000000 000000 0. 0000 00000.0
000.0 0 0 0000000 . . 00 00 0. 0000 0.00.0.0
0.0000 00002.00 00.00.00 00.05 .07. 40.00007. 00.0000 000N ._00> 00.02 00.0 00000
0.0: .07. 00.00001. 2...:

 

00.....0 00.0..— HBL00=0 00000000.”. .3302 ..0 000002000:
0.0.00 .00 0.08000

142

 

 

 

I43

 

20 a. 0.88..
282 0 0 80000.. 000000 0. 88 33 20.;
882 0 0 0808.. 0000.0 2 800 :33 2.8,.
885 00 0 :80.... 0000.0 0. .800 28.. 2.50
8.880 00 0 00.20.. 0000.0 2 080 0038.. 2.80
08852.82 NM 0 000.5. 000 :0 2 800 033 2:3.
0808.552 00 0 5.000.. 00080 0. 800 038.. 2:3.
0805.582 o 0 .0009... 000. .0 2 800 0 _ 0380m8
0805.582 S N 0838 0.8.0 0. 800 000385 8
08225.52 0 0 0.0.8.. 80000 3 88 2038818
0805.552 o o 3038 00800 0. 800 8038218
0822582 o 0 00080.. 000000 0. 800 «$038.38
08205.82 2 0 0.0800 0.0000 0. 88 08.38030
0822582 N _ 000080 00000 0. 800 82385.8
0822582 o 0 000800 000000 2 88 59385.8
20 8820 o 0 30200 000000 0. 800 80382.8
20. 882.. o 0 08200 000:0 0. 800 50385.8
882 o 0 $0009. 8080 2 88 0 £00032.
882 o 0 $38.. 0880. 0. 800 8002320
88:. mm 0 80000.. 8.: 00 0. 800 8002820
882 R 0 _ _ 0. 800 0308820
0.0000 00.02.00 00.?500 0..0.> .07. M00337. wag. 0000 .00.» 00.07. 0.5 0.000
.80 .02 2.88.. 25

 

 

00.....0 E0... 0.00..< -000..0000~. .0.0.07. .0 .000....0 00.
0.050 .3 0002......

 

 

 

 

 

 

 

 

144

 

N m 258.
88%; 2:20 o o o3 _ 3v nmmvom 2 moom coho
8.38% 2:20 _ _ 33¢? 32 G 2 Sea nab
3.522% 2.80 o o omvomav Romeo 2 Sam v8.0
8822... “EEO _ _ ammommv 53$ 3 Sam nor—b
8825‘ 2.80 o o 2.0mm? mmowmo 2 Sam NBC
833:. 9.80 o o c333 owmumo 2 meow .80
.3550 638.30 8.95m 8%; .oz $.2th wazmam «SN .:3> 0:32 85 9:25
ha: 52 .8339— SE
SEC 33,.— _U oﬁo>2h .moogscmom .9532 he 2.25.3 on
m m .255.
coEEoomom o o «823 owmwoh o _ Noon 0 _momohlom
coEEoomom o o no _ m _ av 33% E moon o _ D8218
soEEoomom _ _ E. _ m _ ow 33$ 2 Sam monomohlom
=oEEoomom o o «33$ 83% 3 Sam wooomohlom
58688“ c o 0383 S _ mac 0 _ meow mooomohlom
58888”. o o own 3: 33.3 2 Noom oomomohlom
:oEEoomom v _ Rm 3: 855$ E meow o _ <omohlom
08.335 o o momma? onowce 2 moon mo<omoulom
3550 138—30 moi—cam 33> 62 $3th $33”,— oncN .:3> 0:52 35 9.2m
.:am 62 253.5 ZED

 

 

850 22,.— 8:3 :3.— .5: inoamum .9532 .3 .582: a:
6.2.5 .3 £52.?

 

 

 

 

m. o £83.
8.3 o o 2.?ch $83 o. Noam n.m.v.<u.
8.3 N. m ”was”: . .38 e. Noom vM.m.v.<3
8.3 o o 03.3.? vovwmm 0. Son mm.mv.<3
8.3 o o 3.3me 338 o. Noom cm.mv.<1.
8.3 o o m3 .8? 9.8% 0. mean m.mv.<3
8.3 m N muses...“ n . .33 w. meow N. .4521.
8.3 m . 532.. 33% c. Noon m . .mv.<3
.5250 188.30 8.95% mam; 67. 3.5.87. @523 33 28> 0:52 85 2.3m
ha: 57. .835qu- ZED

 

 

 

coach—:9..— EBEam -mougcmom .3557. .3 2.25.5 on
.335 .3 55...?

145

 

 

 

 

mm 2 2:8,:
885 o o 2 .88 m _ N8: 8 88 $8859:
885 2 m 588: :82 2 88 <eﬂo::.na:
882.80 N N 8828 8: R 2 88 $38.55:
885 h m 888: 82:: 2 88 5.8.85.3:
8.85 o o 888: :22 2 88 <eﬁ8513:
888:6 _ _ 88:8 $82 2 88 9385.3:
88:86 o o 88 _ 8 x. 3% 2 88 8.8.8513:
886 _ _ 98:8 382 2 SS 8.3.82.3:
88:85 N N 38:8 .88: 2 88 388:3»:
88:85 N _ _ 88:8 3 _ o: 2 88 mo c8859.:

5550 6300—30 moan—am mzmtr .OZ "En—«.52 wﬂmumam— OEON 53> 0:32 35 ohuﬂm

Law: .62 :0:5001— EH:

 

350 20.... v.8. «0 $8383. .9537. .3 E253 on
3:5 .3 8:89?

146

 

 

 

 

cm 0 m_SP—.
ooxzmmﬂz 3 N o2 398 3.38 o _ Noom «Om—80.13:
ouxswmﬂﬁ o o Nummmov omwhoo 2 Noom _Omomo.I3=
08.33%,— w N 33 g av $38 2 Noom o E8813:
08.33%; o o o _ oomov 88$ 2 moom _Oﬂomo..I>E
08.33;; m _ N. _ gnaw 338 S Noam o _DomEIB:
coEEoomom _ _ hammmov ouwib 3 Noam mOUomEI>E
coEEoomom o o v _ hmmav w _ oawc 2 Sam c0<82l3=
coEEoomom o o 8% _ av omcowo o _ Noom 36:80:13:
3550 638.30 3.95am 33> .cZ ﬁat—“.82 anam— oaoN 23> 0:32 8% «Sam
ham .cZ 5:384 ZED

 

 

 

3E0 Hor— ucﬁﬁcomcm -moohmcmom .9532 ._c .:0—5.3 on
3:5 .3 5:33.

147

 

 

 

 

 

 

 

148

 

.:m O. 3.5....
03.3. O O N . 33v NNNVOm O. NOON N .mm
0.5.5 O O 333.: .vaNm O. NOON . .mm
D_N...om. v . 5218.: 3.3% O. NOON O.m.m.
oﬁcom m N $33.: 33% O. NOON OOmm
0.5.0.. . . .mmommv w MONOm O. NOON wOmm.
0.5.3. ON O NO . mmov OmmNOm O. NOON NOmm
3.5.5 688.30 8......am 33> 67. wﬂgtez Ema... oaoN 30> 2.37. 25 9:25
._...ﬂ 67. 3.23:1. 5......
23...... «>5 .:0 ._.... «35.0 .c 6...... 8.3.2... .529 2....
S: O 25c...
283.20 mm O wOthov SVNON O. NOON No 5.59.312...
8.530 O O OONvMov O. .m I. O. NOON mm =o_.8mm..mslo_E
8.580 . . ONONvav mails O. NOON OV couﬁmﬂmzlozc
.3080 O O mmOhvov VmOOOh O. NOON Om couﬁmﬂmsleﬁ
c.8380 .. N @233: wovvon O. NOON w. cozﬂmﬂmslog
@550 188.30 8......am 3...; 67. £5.37. 9.33m. onoN 53> 2.37. «mm 98.5

 

:8: .:z 2.88. SE.

 

5.23m Saga 2:. .. 3.?5m 69.8... 8.55 3......
3:5 .3 8.52.3

 

 

 

 

N . £53...

88.5.2 O O OOOOOOV .O . Rm O. NOON m . wco5EA.5:.
83.5.). N . Ova . OOv . MOEO O. NOON O . w=0mmEIm.5.=
83.53. O O OOSOO.V mm . mum O. NOON mow..0mmEIm.5E
86.53. O O OONNOOV .‘OmmOm O. NOON MOm20mmEIm.5E
86.5.). O O Nunmowv 392% O. NOON mN.v..o5...Im.5...
85.5.). O O .83: .ONRO O. NOON <N . vco5...lm.5E
88.5.). O O O.. . ma...V .2.me O. NOON O . v..o5...lm.5:.

8.... O O . . .nwwv wOmmwm O. NOON wOchmmEIm.58
83.53. O O O . Nmowv OMOONO O. NOON NOvco5...lm.5E
5.5.5 128—BU 8....5m 3......» 57. “._...th wacmam oucN 50> 0.57. 85 85.5

._.... 67. 5.28.5 2...:

 

m. 5...... «3.3.0 ._.. an .35 0.3.4

.525 .3 55...?

149

 

150

 

22 3.22. N . OO.OOOO ENMN. O. NOON N0
22 2.22.. O . O... .OO NEON O. NOON 20
22 2.22. .. . ON.ONOO .OOOON O. NOON :0
22 2.22.. O . EN. 5. $82 O. NOON . . 0
2m. 2.22.. O O OOO..OO ORRN O. NOON 20
22 3.22,. O O .OO. 5. 2.2.8 O. NOON . .0
22 2.22. N . $88. 23.. O. NOON .0
22.2 .. O OmOSOO .OMOwN O. NOON 22...
.22.... O. . OmOOOOO .232 O. NOON <22...
O2.2.2.26 O O NaOmOO OwOOO. O. NOON Om
5......2.0 N N OOOOMOO mOOOO. O. NOON 8
5.2.2.0 O O EOROO N58. O. NOON E
5.2.2.0 O O OOEOO ONNSO O. NOON Om
5.2.2.0 O m :53. NO. . E O. NOON 2.
28.2.2.0 N. . EEO. OQOO. O. NOON Om
28.2.2.0 . N 232% OOONOO O. NOON . . m
5......2.0 3 O .ON:OO .OOOwO O. NOON E m
5.2.2.0 . . EMNO. .OOOOO O. NOON m . m
=OO.OOO..0 N . MOENOO NOEOO O. NOON N...
22.8.2.0 . . NONOMOO :82 O. NOON O. m
5.500 60.00.30 8......am m..m.> .cZ an...th MEERH 0.8N 50> 0.57. 2% 0......m
.....= .07. 5.2.0.5 5...:

 

3.2224... 2.8 ._....EOE.
.3200 .ON 2.22....

 

15]

 

3:088:53. N . OOm . OOO NMOONO O . NOON O0
5:225:05. . m m OOOMOO.V OmmVON N . NOON <OO
5:088:82 O. N N.O.OO.v NOOOON N. NOON <30
5:08.553. O N $33.. OOOO mm O. NOON <m0
5:088:85. O N OOONOO.V NOOONO O . NOON <N0
5:08.582 O. O .OOOOOO OOnN: O. NOON .0
5080 . . NOOKOO .NOOOO O. NOON On.
5030 O O N . .EOO OvaOO O. NOON v”.
5030 O O .OOOOOO .OONON O. NOON m“.
5080 ON m SO33 NOmNOO O. NOON N.
5030 O O OONNOOO NOOOOO O. NOON . n.
2.30.5.0 O m OONOOOO 33$ O. NOON Nd
2.30.5.0 O O OgiOm MOOONO O. NOON ND
2.96.5.0 .N O OREOO ONOOOO O. NOON .Q
0.m. 0380.2. O N ONOO . Om OSONN O. NOON o0
0?. 0:30... O O OOSNOO ONmNmO O. NOON 20
0?. 00.2.0.2. m . mO. mmOm NNmONN O. NOON 50
0.m. 0:30... ON N .mmNmOO OmOEN O. NOON <8
07.. 0:30... m . OOmNmOm OOmmmN O. NOON .0
0.m. 0.68.2. O O «ONOvOO OOOMMN O. NOON m0
3580 60000.30 8.95% 8.0...» .07. “0.5.87. 953... 0=0N 30> 0507. 0.6 000.5
0.0: .07. 00.3004 ZED

 

3.23:5 calla 5.0.5.2
.9250 .O.N 2.2.2.3.

 

152

 

2.82 O. O OOOOOOO OSOON t NOON 2....
2.82 O O OOOOOOO OOOOON O. NOON .z
2.82 O. O OONOOOO ONOOON : NOON <Oz
2.82 O O .OO .OOO OOOO . O : NOON Oz
22.20 . N OO .OOOO EON. O. NOON O...
2.220 . . OOOOOOO O. O. .. O. NOON N...
22.20 O O OOOOOOO _ N.NOO. O. NOON :2
22320 O O OO..OOO O.NNO. O. NOON N.
22320 t O .O .OOOO OEOO O. NOON 3

2.2.2 NO . OOO..OOO OOOOON : NOON .5222.
2.2.2 O O .OOOOOO OOOOOO : NOON O:
2.2.2 O O OOONOOO O.OOOO : NOON O:
22.2 O . OgOOOO OOO.OO : NOON E
2.2.2 O N OEOOOO :OOON t NOON 2O:
2.2.2 O. O OO.OOOO ...O.ON .. NOON O:
2.2.2 ON O O.OOOOO NN.O..N : NOON O:
2.2.2 O O OSOSO OOOO . O t NOON O...

3:20.552 . . OO.NOOO OOOOO. O. NOON O0

35.05202 O O OONNOOO OOOONO O. NOON O0

.8235: t O EOSO OOOON. O. NOON .0
55.00 00.00.30 8.955 m..m.> .07. 9.3.207. wﬂaam— 0:0N 30> 0807. 8% 0.3.5

0.0! .07. 00.3001. EEO

 

0.22.2: 2'5 ._. 52.2.2
.380 .O.N 2.2.252

 

 

 

 

mm m—NHOOF
82.. O O OOOONOO .OOOON O. NOON O2
82.. O O OOONNOO NOOOON N. NOON NO
82.. O O OONOOOO OOOOON N. NOON .2
EEOC N . OOOONOO OOOOON O. NOON <OO
2.225 O. O OOOONOO :ONON N. NOON NO
Eeowo O N OO . OOOO OOOOON N. NOON .O
2.82 O O OOONOOO OONOOO N. NOON O2
.6580 c800=OU 3.95am 83> .OZ MOE—tel mazmam 0:0N .30..» 0:32 8% 0.35
222.25 Eb

 

5.0.5.5 035 :a 50:2

 

.9250 .O.N 2.2.2.3.

153

 

 

 

 

Nm 0. 0.0.0h.
0:002 N . $530.. 0... . Own 0. MOON m0=00.0
0:00.< o 0 «mm . 30 m0 . own 0. mooN 0.0.80.0
0:00.< w m wvamov .OONOO 0. MOON 09:00.0
0:002 0 m 08030 Nb . new 0 . mooN Nm0=00.0
0:00.... Nm m NNw . mow «memos 0. mooN 30:00.0
0:00.< o 0 $9000 30%.. 0. mocN . .0000...
0:00.< v N 33000 Soc 3. 0 . mooN .0002.0
0.0030 o o 08?? con . .OO 0 . MOON 000000O0I0000...
0.0080 o o $0030 03 . E. 0 . SON m00000O0I0000.0
00080 o o Nwommav 0.. . N: 0 . mooN m000080l0000...
00080 N. m 30.030 2.33. 0 . mooN . 00008060020
00080 . . 30060.. ”.033. 0 . mooN o . m.0000m0l0:00.0
0:00.< m N N0 . mm? 20th 0. mooN . m0.000O0I0..00..0
00080 o o mm .03.. SON: 0. mooN o .0000w0l0:00.0
3.500 00.02.00 0.0—.:00m 3.... .02 $3.202 Maﬁa. 0=0N 200 > 0.002 0.5 020.5

2.0: .02 03.0004 2H3

 

00.....0 0.0.... 0.5. - 00020000“ .02...02 .0 .00....20 0:
6.020.; 0... ”5.02000 03.00.0090 .:0E0u000... ..000 20.. 00.000. 0.0.: 020.... 2.0.. 0... .0... 5.5.00 0... .50
00.02.00 8......00. 2.0.. ..0 200.00.. 2.200.. .3 3030......» ..0 200.5... 50.2000. SS .200.» .050: 02000 2.0.. ..0 .O: .v.N 5000...;

 

154

 

 

 

.ON OO 225
2.2.2 .O O OONOOOO OOOOOO O. OOON N08012:...
2.2.2 O O OONOOOO ONOOOO O. OOON . .81....23
2.8.0 .O O OOOOOOO .OOOOO O. OOON 2.3.2.2;
22.20 O O NOOOOOO .OONOO O. OOON 8.3.2.2...
22.2258... ON O NOO .OOO OOO.: O. OOON 8.3.2.2...
28.2.2.8... OO O OOOOOOO ONOOON O. OOON 3.3.2.82
22.22282 O O OOOOOOO OON. .2 O. OOON 209.82....
22.28282 O O OONOOOO OONO .2 O. OOON 8038......
22.22582 O O OOO.OOO OOONNO O. OOON O. 0282.2
22.22.25: O O OOONOOO ONOONN O. OOON 309.82....
22.22582 O O OO.OOOO ORONO O. OOON 322.82....
22882.8... O O OOOOOOO NNOONN O. OOON 30.2.22...2
2822.22.02 O O OOOOOOO OONONO O. OOON 80.2.22...2
22.28282 O O NO . OOOO NNOONO O. OOON 30.28....2
2.. 2.8.. N . OO.O.OO OOOONO O. OOON 82.2.82...
2.. 2.8.. O _ NNOO.OO OOON: O. OOON 32.2.82...
2.2.2 O O ONOOOOO OOO.O2 O. OOON OO.OONOO...
2.2.2 O. O OOOONOO OOO. OO O. OOON 80022.2.
2.2.2 N N OOO.OOO OOOOOO O. OOON SS. .22...
2.2.2 OO O OOO . OOO ONOOOO O. OOON 8.083....
.Oﬂ—SU ﬁance—BU mains—Em at: .07. £5:th wig—«H GEON 23> 2:07. 35 9:25
.0... .Oz 8.82.. 2.5

 

 

00....O 0.0.... 0.00..< - m002...0m0~. .020.02 0.. .000..20 0:
0.02.0 .O.N 2.2.2.22

 

 

155

 

 

 

 

O . .28..
8.2.2.: O O N.OOO.O O.O..OO O. OOON £32.22...

8228.3. O O OO. . N.O 888 O. OOON 20.8....
82288. O . OON . NOO 80.8 O. OOON 80.8....
8228.3. O O OON....O OOOOOO O. OOON 80.8....
8228.8. O O .ONOOOO .OOOOO O. OOON 80.8....
82288.. O O OO.OOOO OOOOOO O. OOON 80.8....
82288.. O O OOOOOOO .OOOOO O. OOON 80.8....
82288.. O O O.OO.OO .OORO O. OOON O . 2.8....
82288.. O O OO. OOOO OOO :O O. OOON NO<.8....

.mE—uanv §0«00=¢U mO—Qaﬂm 3:.— .OZ ”Hz—:Oz “ﬁtmﬂm OGON hﬂo> OERZ 03m yuan—W

.:a: .02 BOP—GOO}— 2P:

 

 

359 in 8.24 :0:. =0: - 80.583— .2507. 0.. .:0—5.... on
._....80 .O.N 2.2.22.2

156

 

 

 

 

 

 

 

 

 

 

 

 

. . .28...
823.2 O O OO.OOOO OOOOOO O. OOON 22823...
823.2 O O .OOOOOO . .88 O. OOON 808232.
823.2 O O N.OONOO O2O2OO O. OOON 8823...
823.2 O O OO2OOOO OOONOO O. OOON 88232.
823.2 O O OOO.OOO OOOOOO O. OOON 2823...
823.2 O O OOOOOOO OOOOOO O. OOON 88232.
2.3 . . ONO2OOO O2OOOO O. OOON ON8232.
5.500 0802.00 moi—02m Oz: .02 $5002 ﬁrm-Om 000N 03> 2002 8% 9:25
.:0—.— .02 0020004 2.:—
ﬂcmaaam .5 ea“ EBB-«m— - ou_>.—om “no.8“— mgaum 303::
O. 2 .282
2.2. O O OONOOOO 2OO8O O. OOON 2.002..
9.2. O O OO.O2OO NOOOOO O. OOON 8.002.
2.2. O O OOO2OOO . .OOOO O. OOON 8.002.
8.2. . . O.OOOOO OOOOOO O. OOON 8.0020
8.2. O O OOOOOOO ONOOOO O. OOON 8.02.
e... O O O22O2OO 2.88 O. OOON 2.0.2.
2.3 O O .OON2OO 2.88 O. OOON O . .002.
30000 0802—00 .2905 3:.— .02 mat—v.02 M533— 000N 03> 2002 «am 9:25
.:0: .02 00:000.— ES

 

350 Eur.— EBEam - 80010082 .3502 ._0 .:0—5.; a:

 

 

8.8.0 .O.N 2.2.22.2

157

 

 

 

158

 

mm . on £80...
22...... O N OOONOOO 0.0000 O. OOON 8222.3
22...... O . O2000OO OOO0.0 O. OOON 8222.3
22...... O. O OOOOOOO OO22 .O O. OOON 8.22.3
22...... . . O NOOOOOO O . O2.O O. OOON 2223
22...... 2 O OOOOOOO OO22.O O. OOON N222...
22...... O O OOO.OOO ONO2.O O. OOON 2222.3
22...... N. O OOONOOO OO2ONO O. OOON . .2283
22...... OO O OOONOOO NON0.0 O. OOON 2222.3
8.8.22 N . O. OOOOO OOOOOO O. OOON 2o
828...... 2N O OOOO.OO .O.2OO O. OOON O0
828...: O. N .O . O.OO OOO . OO O. OOON O0
828...... . . 0.00NOO OO0.00 O. OOON O0
828...... O N O2.ONOO 2OOOOO O. OOON O0
82...... . . OO.2NOO OOOOOO O. OOON N0
828...... O O .O2ONOO OOOOOO O. OOON .0
O28...... O O OOOOOOO 2.0NOO O. OOON O2
2.82.... 2 N O.OOOOO 2OOOOO O. OOON O2
2.82.3. O O OOOOOOO .O. .OO O. OOON O.
2.82.3. O N OOOOOOO NOOOOO O. OOON NV.
2.82... O O OONOOOO OOOOOO O. OOON .2
@0000 0802.00 8.90% 3.: .02 3.5.02 #032...— 000N 03> 0.022 8% 9.20m
ha: .07. 03230..— SE

 

 

SEC Eu... 00....30 - 89.00.82 .2522 .0 .025... on
.2280 .O.N 202.22....

 

 

 

159

 

O N ._....O
88.0 O O 22OOOOO O2OOO2 O. OOON 22882.3:
28.2.2.0 O . OOO2.OO OO. . .2 O. OOON 02882.3:
8.20 O O OOOOOOO OO2OO2 O. OOON 20882.3:
8.20 . . ONOOOOO 2.O.O2 O. OOON 22882.3:
28.8.2.0 O O OOOO.OO O2NNO2 O. OOON 02882.3:
82.2.2.0 O O OOON.OO OOO2OO O. OOON 02882.3:
8.20 O O NOOOOOO OOOOO2 O. OOON 02882.3:
80.2.2.0 O O .OOO.OO .OO2O2 O. OOON 00882.3:
8.22.2.0 O O OOOO.OO NOOOO2 O. OOON 0 0882.3:

Nan—=00 5302—00 m0-qaam 3.: .OZ Mama—«.37. atmanm OEON 53> 0507. 3mm aha—.m

.:0—.— 57. :Otﬁocw— 2S

 

0050 0.0...— ..:0.2mmv .. muggy. .0..me .0 .0oEt0mon
8......0 .O.O 2.2.2.2

 

 

 

 

0 0 8.50...
08.2.8.3. . . 0m. .32 03200 0. m00~ 20m80. 3:
08.2.8.3. 0 0 ~33: 000500 0. m00m .Om80...3=
08.2.8.3. m . 0200.02 $0000 0. m00~ 03.80.13:
08.088.). 0 0 0.00miq NNO0B0 0. moon 5080.13:
08.2.8.3. 0 0 233.2 200000 0. m00N 0 .O80.I3=
:0:—.5083. . . 30mm? 03.80 0. m00m mOU80.I3=
0082.083. . . 2.530.. m .0000 0. m00~ 0O<80...B:
002.8083. m . ~03 . 02 000000 0. m00~ 0 . <80.IB:

3.500 0802.00 80.95% 3.: .07. 3.5.57. main... BEN 23> 2:27. 3% 0.32m

._.“: .07. 20.0301— EH:

 

 

 

0050 0.0.... :0:—8083— - 8003083— 12527. .0 .:0—5.8 on

2.2.0 .22 2.2.2.2

 

160

 

 

 

 

161

02 0. 0.0.0.0
0.800. 00 0 0000002 002200 0. 0000 0.m.m.
0.0000. 0 0 0. .0202 020000 0. 0000 . .mm
0.0000. 0 0 0000002 02. 000 0. 0000 0 . mm
0.N:0m 0 0 0000002 000000 0. 0000 000.0.
0.000.. 0 0 0000002 200000 0. 0000 00mm.
0.0000. 0 0 00. 0 002 000000 0. 0000 00mm
8.020.... 000.0 0 0 00.0002 020000 0. 0000 00.00
8.02.... 000.0 . . 0000202 .00000 0. 0000 00.00
8.0.6.... 000.0 0 0 0020002 000000 0. 0000 20.00
8.0>0.... 000.0 0 0 0000002 .00000 0. 0000 00.00
8.0.6.... 000.0 0 0 0000202 000000 0. 0000 .0...O

30:00 00.02.00 00.02.00 030- .07. Mair—07. 00.8000 0:0N .00> 02.07. 00.0 0.000

.:0: .07. 00.00001— 2E

 

.0:— 030 .00 000 03020 .0 00:00. 00. 0.830.... 000.0 00.0
0050 0.0.... 3.0 8.03%.0 - 0.00.0853— .0.::0Z .00 2.0.5.0 00—

2.2.0 .O.N 2.2.2.2

 

 

 

 

 

 

20 0. 8.0.0...

8220 O O KNOB. .MOWOO O. .OO. .02

222.. O O 0.82. NOOQO O. MOON 82220
28.2....0 O O N223. .OEO O. MOON 8
28.2.2.0 . . m N828 2.82 O. mOON N0
28.2.2.0 O O NOOmOOm ONO2.O O. MOON .0

2220 O O 382% 828 O. .OON .2

222.. O O 8.28 :88 O. .OON .2

822.. N . 8.2.8 .228 O. mOON .2

822m. . . 2 20.8. .288 O. MOON O2

222.. 8 N OO....O. .0..OO O. .OO. .2

822.. O O 82.8... OO22.O O. MOON .2

822.. O O 822% N82NOO O. MOON M2

:22... NN N :88. NERO O. MOON .2

0.0000 00.02.00 00.02.00 8.: .07. 00.5.07. 00.80...— 0000 000.0 0507. 00.0 0.000

._.0: .07. 00.00004 2...:

 

 

2.0.0... 03000 .0 80:00 00 00.0.03 0.....—
....2.0 .2.N 2.2.2.2

162

 

 

 

 

0 0 0.0.0.0
008.002 2 0 000.002 00 0000 0. 0000 00....0.0000E..0.00E
008.00.). 0 0 . 000002 020000 0. 0000 00.20.0000EI0E0E
008.00.). 0 0 02. 0002 000000 0. 0000 .I.>.m.00000.I0.00E
008.00.). 0 0 2002002 000000 0. 0000 .3050000812000.
008.00.). 0 0 0.0.002 00.000 0. 0000 .0Im0000EI0.000.
008.00.). 0 0 0 . 0. .02 .0 . 000 0. 0000 0.)...0.0000.0I0.000.
008.00.). 0 0 0 . 00002 000200 0 . 0000 <0...m.0000EI0.000.
008.00.). . . 2000002 0. 0000 0. 0000 <0 00.0.0000EI0.000.
008.00.). 0 0 0000002 00. 000 0. 0000 < . 20.0.0000...I0.00E
008.00.). 0 0 0 002002 00020 0 0 . 0000 <0 . 2Im0000EI0.00E
00.0.. 0 0 0000002 002.00 0. 0000 0 . 0 . Im0000EI0.00E
0000.). 2 . 000.002 020 000 0. 0000 0. . .Im0000EI0.00E
008.00.). 0 0 0000002 0. 0000 0. 0000 <00Imm0000EI0.00E
0.0—.00 00.02.00 00.00.00 0.... .07. 00.5.07. 00.80..— 0000 00> 0.007. 00.0 0.000
.:0: .07. 00.00004 S00:

 

2.2.2. .3020 .2 .2... 23.0 ._....

2.2.0 .2.N 2.2.2.2

163

 

 

20 2.320 2 v 32 an 0032 t 88 m _ 0
20 83.20 0 q 8% am 30th : SON N00
20 2.820 m _ 0:: an :EsN : SON _ _0
20 2.320 c o 088% 28: 2 88 _0
282 o o 083% 8&2 : SON N600
2.82 m N 3N8? 02ch t 88 :20
500820 0 N 02080 :02: 0. SON 00
50052.0 o o @280 $32 2 SON 00
$00520 N _ 8380 N32: 0. SON E
500520 o 0 838m 0830 E SoN 00
500820 m _ 8230 :2 K 2 SON mm
50052.0 o 0 $330 2%.: 2 SON E
500806 o o 8380 ONE: 0. MSN NNm
£03020 0 o 8:08 5N8 2 SON .8
50052.0 v _ on. 8% $82 2 SON 80
500820 m N 802% 003% 2 88 m _ 0
$0386 0 o 003% £080 2 SON z 0
$0386 2 .2 ENNNS 88$ 2 88 20
50385 N _ _NENom 8200 0. 88 N_ m
503020 mN 0 $880 58. 2 SON 20
0.0000 00.02.00 80.00.00 8.... .07. 00.0207. 00.80.. 000N 000.0 0.007. 3.0 0.000
0.0: .07. 02.0001. 2...:

 

32900: elem 0000.02
0.050 iN 008...?

164

 

 

165

ow0m~0 O O OO _ NOOm ONOOOO o _ mOON _ “—
E_b:< O O OO _ Ono: 3:3 3 mOON mm
E_.=:< m _ Oovgav OOOOVO O _ mOON Nm
EtE< O O mwNnOOv N _ NOmO O. MOON _m
5051sz m N m _ MOOOV Own _ we 2 mOON VD
0:05:20 O O OthOOm NOV—S E mOON mo
0:05:20 O O MN _ v _ Om m moose 2 mOON ND
0:05:95 m N hvaOm $38 2 mOON _D
000.0 O O Show: wOwNmO O. MOON 3000.0
0.8—0 O O OOON.OO: mmocmo 2 mOON 3000—0
0E 0:305 E m :3 _ Om OOOONN. 2 mOON O0
01.3 0:305 O O OOBVNOm OVmwa E mOON mu
07,— 0:305 O O m2 mmOm QwaNb 2 mOON mu
02 0:305 ON m wmcomOm NOOVMN. S mOON m0
03 0:305 O O «MuhmOm Oncmmn 2 mOON v0
0.3 0:305 _ _ NMOOmOm 33$ 3 MOON mu
02 0:305 O O OOVOMOm OthNh O. MOON ND
03 0:305 O N OmthOm ONVOON C MOON 30
0E 0:305 v _ mmmw Sn 2ng 2 mOON 20
0E 0:305 N _ ON _ ONOm VSVON S mOON E U
#2500 330—30 3.5.—Em at: .02 “£5.82 ﬁtmam— BEN 30> 0802 03m 0.3—i
.:0: .92 55304 SH:

 

£30245 gala .3033:
6.2.5 .3 £089?

 

 

::0O_< _ _ M _OOOOn ONENN 2 MOON MI
«:0:? O _ N OvawOv OOMOON 2 MOON 3 z
«:0:? MN v M _ NOOOO m MOOON E MOON M _ I
O:0:? O O NONOOOO SEEN N. O MOON N _ I
O:0:? N _ ONO _ 8: b _ MM _ M E MOON O _ I
5:088:02 _ _ OOONOOOV OO_ M M: O _ MOON O0
5:058:82 O O SONS: mmmmNN O_ MOON O0
5:08.882 O _ M m NNmNOv VOMONN O _ MOON NO
5:088:02 O O OON.»? OONONN O_ MOON O0
5:088:02 O_ N MOOMOOV NvaON : MOON no
5:088:82 _ _ _ v _ O _ OOOq MNOOON N _ MOON v0
5:088:82 O O 3833 _ :3 MN O _ MOON M O
5:08.582 O O M _ ONOOOv N _ VON: O _ MOON NO
2880 E N OEOOOV vaMNN O _ MOON _ _O
2880 N _ OOOOOOOv Ohm _ MN. O _ MOON O _ O
5:225:22 N_ M OmOOOOw MNONK O_ MOON _o
ow080 : _ N283: NMOOOO O _ MOON Mr,—
ow0mOO O O OEE‘OV OMVOOO OO MOON vu—
ow0mOO O O OOVOOOO OMOMON O_ MOON Mm
ow0$0 _ _ OOONOOOu OOMOOO O _ MOON N5
b::OD 380.80 8.585 :3: .OZ wEOIOZ £583— 0:¢N 80> 0:82 03m 085
:3: .82 .8339— SF:

 

3.3.5 2.3 .5 :22
.380 .3 55:3

166

 

 

888 O. .. ONONOOO NEON O. NOON ON 8:308:22

888 N . NONE... OOOOON O. NOON ON 8:323:22
22320 O . 68.8.. 832 O. NOON N. 8:328:22

0088 O O 888.. ONNONO O. NOON 22
22320 O O N33... O.NNNO O. NOON N2

8820 N N NONE... OONNON O. NOON Os.
.8820 O O .NNNNOO .OOO.OO O. NOON N:
22320 O O 282... NONOON O. NOON «2

2.86 O O NN .OOOO NNNNNN O. NOON N2

888 N N .88.... 8... .N O. NOON N2

888 O O NNNNNOO NNN...N O. NOON OE.

888 O O NNONOOO .NNNON O. NOON :2
22320 N N 8:2... .ONNON O. NOON N:
22320 O N NN.NNO.. NNNNNO O. NOON 3

882.. N. N OOONOOO NONNNN N. NOON 302.8:

88:2 O O NOOONOO NNONON N. NOON N:

82.2 N . ONONNOO NEON N. NOON N:

82.2 N. N NOOOOOO NON.ON N. NOON N:

«.82.. O N ONNNOOO NNOOON N. NOON N:

«8:... O O ONNOOOO ONNaN N. NOON O:

@580 3.00.80 8.58% Oz: .02 wEOIOZ mﬁOmam 0::N 80> 0:82 85 0.8.5

.:0: .02 833.5 25.3

 

3.82:: «8'3 80.8.:
._...80 .O.N 2:88....

167

 

 

220 O O NO. 3.... NNNNOO O. NOON NO
88883. O O .NNNONO OO.ONO O. NOON ON
82283: O. N ONNNONO ONNNNO O. NOON ON

8230 O. N ONNOSO . .N. .N O. NOON ON
8.88 O O 2.88.. OOOOON N. NOON NO
8230 O O .NNNNOO NNONON N. NOON NO
888 O O NONNNOO ONOOON N. NOON .N
885 N N NONONOO OOONON N. NOON N83.
888 OO O .NNNNOO OONONN N. NOON N28:
885 O O 2.88.. NONNON N. NOON .83.
85.883. O. O ONOOOOO ONOOON O. NOON N:
85583: . . 2.8.3 .OONON O. NOON .2
2.8.... O . NNONNOO .NNOON N. NOON Oz
82: N . ONNNNOO NNOOON N. NOON Oz
382 O O ONNNNOO NONONN N. NOON N2
2.8... O. N . .NONOO ONNNON N. NOON Oz
382 O O OON .OOO .NNN . N N. NOON Oz
2.8... ON N NNNOOOO NNOOON N. NOON N. 2
22320 O O :NNNOO NNNNON O. NOON NO 8:233:22
888 O O NONONOO N.ON.N O. NOON ON 88.3.32...
3.500 6080.80 .:0—9:56 3:.— .OZ wintcz wEOmam 0:¢N 80> 0882 02m 085
8: .Oz 8.38.. 2.5

 

3.39245 3.3 5.0.222
._...80 .O.N 2:89....

168

 

 

 

 

NON NN. 2.89
538.0 . . NNOOOOV NOVM .N. O. MOON NX
53.8.0 O N OOONOOO MO. OON. O. MOON .x
.8230 O O NONEOO NNNNNN O. NOON ON
>880wO O M 30. MOv OmMMMO O. MOON m...
::oo.< O O MONOMOv O0.00N O. MOON v...
88. O O ONOVNOV OOOOON O. MOON M...
008. N . VOVONOV .NOMMN N. . MOON N...
008. O N OONOMOv OOOOON O. MOON E.

5:80 3.00.80 8...:8m 3.: .02 8.582 manna 0:ON 80> 0:82 85 0.8:m

:8! .02 8.884 2.:—

.82.... 8.3 8 22:.
._...80 .O.N 2.32.8

 

169

Appendix 5.1. Hair snare set-up and checking procedures
Michigan Blacﬂear Hair Snaring Research Project- Procedures

1. Setup Procedures
a. Trap Number and Location
i. All hair snare traps have an identifying number and GPS location.
This number and location should be noted on the top of the data
sheet along with the date in which the trap is set, and the
individuals who are setting the traps.
b. Site Selection
i. Although the traps have a speciﬁc location (noted by the GPS
location) the placement of the traps does not have to be in that
exact location, but should be reasonably close (within a ‘A of a
mile). Traps should be placed where they have the greatest
probability of being visited. These would be in the following
locations:
1. Upland/lowland edges of the following habitat types:
a. Aspen
b. Northern Hardwoods (esp. cut or thinned)
c. Upland brush/ opening
d. Swamps
2. Features to look for that would channel bear travel:
Streams or rivers
Edge of lakes
Old logging roads
Power line or pipeline right-of-ways
Well used deer runways
Upland ridges through swamps

”9999‘!”

c. Tree Conﬁguration
i. Once a trap area has been selected a grouping of trees must be
found.
1. A suitable grouping of trees should meet the following
criteria:

a. Look for at least three trees that could be used to
form an enclosure that is at least 10 feet on each
side.

b. Trees should optimally have smooth bark so that
bear sign can be easily detected.

c. The ground should be as level as possible to insure
that the height of the wire remains consistent.

d. Trees should also be chosen that contain many tree
branches from which the bait can be suspended.

e. If a site is suitable but does not have a suitable
arrangement of trees, then metal posts that have
been provided can be used.

(1. Barbed Wire Set-up

170

Appendix 5.1. Cont’d.

i. Lower Wire

1. Starting at one of the selected trees, the barbed wire should
be fastened to the tree using a metal fencing staple. The
height of the wire should be 8 inches above the ground.

2. Wire should then be hand-stretched to the next tree and a
measurement of 8 inches should be taken from the wire to
the ground. The wire should then be attached to the second
tree using a fencing staple.

3. Continue to attach the wire to all of the trees, forming an
enclosure.

4. Inconsistencies in the ground such as low spots can be
ﬁlled in with logs or other woody material to insure that the
height of the wire remains 8 inches above the ground.

ii. Upper Wire

1. Using the same method, another strand of barbed wire
should be fastened to the tree at 20 inches above the
ground.

e. Warning Signs and Flagging
i. Once the wire has been stretched to form the enclosure:

1. Florescent flagging material should be tied on the wire in at
least 3 locations to warn any human individuals of its
presence.

2. Warning signs should be posted on 2 trees surrounding the
site to warn people of the barbed wires’ presence.

f. Bait placement
i. Using the bacon bait provided:

1. Attach string to one end of the bait.

2. Throw the bait over a tree limb at least 10 feet in the air.

3. Attach a second string to the bacon bait.

4. Attach ﬁlm canister to string beside bacon bait and put
anise scent on cotton balls in canister.

5. Pull the bait into the air so that it is at least 8 feet above the
ground. Tie off each end.

6. The bait should be at least 15-18 inches ﬁ'om the trunk and
24-36 inches from the nearest limb.

g. Filling out the data sheet
i. Provide the trap name and date that the trap was set.
ii. Provide the coordinate locations of the trap using a GPS.

1. These coordinates can be either in lat/long or, preferably,
UTM.

2. If you do not have a GPS then notify the MSU crew and
they will help you out.

iii. Provide the names of the individuals that set up the trap.

17]

 

 

Appendix 5.1. Cont’d.
iv. Provide general directions to the site to assist any person that may
be checking the trap, but has not previously been to the site.
v. Give the time in which the trap setup began and the time in which
the trap has been totally set and baited.
vi. Provide any general comments that may be useful, or are pertinent
(an example may be the presence of bear sign).
vii. Give an above view sketch of how the barbed wire is conﬁgured.
Supplies list for set-up:

Hammex
Bar.bed.wir¢..
Eliéréft'géiuwnrs
fansingstaples
Japemeasurc

Data sheets
Flagging material _
Warning signs
Nails for signs
Metal posts

Bait string . , . V
Knife; 9t: scnssors
Bacon bait

Film canisters
Anise scent
(Shading denotes supplies that are not provided by MSU)

II. Checking Procedures
3. Checking intervals
i. Hair snaring will occur in one single period throughout the
summer.
ii. The period consists of 5 checks that are separated by 5-8 day
intervals.
iii. The MSU crew will assist in a checking schedule for all DNR bait
routes once the baits have been set, if needed.
(Refer to attached sample calendar for example)
b. Checking procedures
i. Begin by making a sketch of the barbed wire trap in the space
provided on the checking sheet.
ii. Checking the barbs
1. Hair should only be collected off of the upper wire. Any
hair on the lower wire should be removed and ﬂamed only.

172

Appendix 5.1. Cont’d.

2. Using a white piece of paper or an index card, start at the
farthest north post or tree. Check each barb for the
presence of hair by placing the white piece of paper behind
each barb and proceed around the trap in a clockwise
direction.

3. The barbs should be counted in the same clockwise
manner.

4. The ﬁrst hair sample that is encountered will be sample one
and its general location on the barbed wire should be noted
on the sketch and its barb number should be recorded on
the data sheet.

5. Each barb is considered to be an independent sample and
should be place in separate envelopes.

6. Once a hair sample is placed in an envelope, the number
that is preprinted on the envelope should be recorded on the
data sheet.

7. Proceed checking each barb until you reach the starting

point, making sure that each barb is independent.
iii. Measuring the height of the hair sample

1. Using a tape measure, indicate the distance between the
hair sample and ground (Although the wire was set at 20
inches, irregularities in the ground make the wire slightly
different heights at different locations).

2. This should be done for each independent hair sample.

iv. Removing the hair

1. Hair should be removed using forceps, or hand pulled and
placed into the pre-numbered envelopes and the
approximate number of hairs collected should be circled on
the envelope.

2. The most important part of the hair is the white tip of the
large guard hairs. It is extremely important to take good
care of these hairs because they contain the most DNA
material.

3. Hair samples should be kept in a dry location and if the hair
sample is wet when collected it must be dried by placing
the envelopes separately in a warm dry location.

4. The barb should be ﬂamed with a lighter to prevent cross-
contamination.

v. Other notes:

1. Note any evidence of bears at the bait site by looking for
claw marks on trees or the absence of bait in the general
comments of the data sheet.

173

Appendix 5.1. Cont’d.

2. Note any evidence of a non-target species being present at
the bait site in the comment section of the data sheet.

3. Of particular interest is if there is evidence of black bears at
the bait site, but there was no deposition of hair on any
barbs.

c. Filling out the data sheet
i. Begin by indicating the trap site name, the date of the visitation
and the time of the visit.
ii. Circle the visitation number that is appropriate and indicate the
name of the person collecting the sample.

iii. Fill in the table for hair samples, making sure to write down the
envelope number, barb number, and number of hairs on the data
sheet

iv. Make general comments about the condition of the hair samples.

v. Make any general comments that might be pertinent to the study
and ﬁll them in on the comments section of the data sheet.
d. Bait replenishment

i. If the bait is absent from the site, hang a new bait and indicate the
presence/absence of the bait on the data sheet.
ii. If bait is still present at the site, you do not have to replenish the
bait.
e. Trap removal
i. At the conclusion of the sixth check, all of the barbed wire and any
indication of our presence should be removed from the site.
ii. This includes all of the baiting materials and any ﬂagging that
marks the directions to the bait site.
iii. Items that should be saved include the warning signs and the
barbed wire.
iv. All other items should be disposed of properly.
Checking Supplies List: Data sheets
Hair envelopes
Eorcepﬁ
Bait
Note cards

.....

Hammer

Nails

Staples

Film canisters

Anise scent

(Shading denotes supplies that are not provided by MSU)

l74

Appendix 5.2. List of supplies needed for each task to be performed in the collection
of hair samples and hunter harvested bear hair and muscle tissue samples.

 

 

Task to be Performed Supply Needed
Equipment needed to ﬁnd sites Global positioning system
Topographic maps
Hair Snare Equipment for one snare 60 feet of barbed wire

Checking one snare for one week

Harvested Bear Hair and Tissue Collection

175

10 fencing nails

2-4 10 penny nails
1-2 warning signs
Tape measure
Hammer

Fencing tool
Flagging

Five gallon bucket
One pound of bacon
20-40 feet of twine/string
Anise

Film canister

Latex gloves
Envelopes

Pencil

Data sheet
Clipboard
Envelopes

Forceps

Tooth extractors

10 mL vials

Stamps for envelopes
Buffer solution
Protocols sheet

Appendix 5.3. Formed that was ﬁlled out upon the set-up of a hair snare.

Hair Snare Set-up Data Sheet - 2003

Trap Site #: Date: UTM: E
Date Baited: N

 

Set-up Crew Names:

General Directions to Site:

 

 

 

 

Setup Information
Start Time: End Time:

General comments:

 

 

 

 

Sketch of @it site:

 

 

 

I76

 

Appendix 5.4. Copy of the stamp that was placed on all the hair collection envelopes.

Bear Hair
Date: Envelope #:
Trap Site #:
Visitation #2 l 2 3 4 5 6 Barb#:
NumberofHairs: 14 5-10 ll-lS >15
Collector:
Comments:

 

 

 

 

l77

 

Appendix 5.5. Data collection form used when checking hair snares for the
presence or absence of hair samples.

Hair Snare Checking Data Sheet - 2003

Trap Site #2 Date: Time:
Visitation #2 l 2 3 4 5 6 Collector:
Total Number of Hair Samples Collected:

 

 

 

Hair Samples:

Envelope Height at Barb
# Number Number Comments/ condition of

1
2
3
4
5
6
7
8
9

 

10
General comments:

 

 

Sketch of bait site, barb number, and sample number at approximate location of hair
samples: (Start collecting samples at the northernmost post or tree and proceed
clockwise)

T

N

 

 

 

 

 

 

 

 

178

 

Appendix 5.6. Protocols for the collection of hair, muscle tissue, and tooth samples.

Sampling Protocols for Black Bear Tissue Samples.

Items needed:

In the materials provided by MSU you should ﬁnd:

Four envelopes. Inside of the larger envelope marked "Bear Samples", are three
smaller envelopes marked "Bear Tooth", "Bear Hair" and "Bear Tissue".

A 1.5 mL snap-top plastic tube ﬁlled with buffer. This is pre-numbered and is
inside of the envelope marked "Bear Tissue".

Scissors for cutting a small portion of clean muscle tissue.

Tweezers/forceps to hold the muscle tissue when you cut it and put it into the
tube.

To take aLtissue sarmle:

I.

Cut off a small piece of muscle tissue. Do not use soft tissues such as heart, liver,
or kidneys. A piece of fresh clean muscle from the lining of the abdominal cavity,
legs, or other easily accessible area from the ﬁeld-dressed carcass can be used.
You can also use tongue.

The cut piece of tissue should be longer than it is wide. A piece about as long as
your thumb nail and no bigger around than a pencil will work nicely. Do not
overﬁll the tube with tissue, the buffer doesn't work well if there is too much
tissue in the tube. A general rule is that there should be 2 parts buffer for 1 part
tissue.

Aﬁer you put the tissue in the tube, please make sure the top "snaps" closed. Turn
the tube upside down a few times to make sure the tissue is covered with the
buffer.

Because we are collecting muscle tissue samples for genetic analysis, it is
necessary to clean off the tweezers/forceps and scissors between bear samples.
They don't have to be spotless, you can simply rinse them with water.

Label the envelope marked "Bear Tissue" with the Seal Number and hunter's
name.

Place the tube inside the "Bear Tissue" envelope.

The tube does not need to be refrigerated or frozen. It can be stored at room
temperature for several months.

Note: The buffer is made up of Tris, EDTA, UREA, Sarcocine, NaCL, and Water. It is
not in any way hazardous but please don't drink it.

179

Appendix 5.6. Cont’d.

Sampling Protocol for Black Bear Hair Samples
1. Pull a few strands of hair from the bear — be sure to get the follicle (roots).
2. Label the envelope marked "Bear Hair" with the Seal Number and hunter's name.

3. Place the hair in the envelope marked "Bear Hair".

Sampling Protocol for Black Bear Hair Samples
1. Follow the instructions provided by the MDNR to collect a tooth.
Label the envelope marked "Bear Tooth" with the Seal Number and hunter's name.
Place the tooth in the envelope marked "Bear Tooth".

PLEASE BE AWARE THAT YOU ONLY NEED TO COLLECT l TOOTH. YOU

UP?!“

NOT NEED A TOOTH FOR THE MDNR AND ANOTHER TOOTH FOR MSU.

Final Protocol for Black Bear Samples
1. Label the envelope marked "Bear Samples" with the Seal Number and hunter's name.
2. Put the three smaller envelopes ("Bear Tissue", "Bear Hair", and "Bear Tooth")
inside
the envelope marked "Bear Samples".

3. Store the "Bear Samples" envelope in a dry spot.

4. Michigan State University will contact you about picking up your samples.

180

   

Illlrjlll‘lggijlll(alum