RESEARCH PROJECT STUDY PLAN
CABINET-YAAK GRIZZLY BEAR DNA PROJECT
STUDY TITLE: Density and distribution of the Cabinet-Yaak grizzly bear population.
PRINCIPAL INVESTIGATOR: Katherine C. Kendall
BASIS+ PROJECT: Conservation requirements for wildlife of the Northern Rocky Mountains.
Project Number: XXX
BASIS+ Task: Cabinet-Yaak Grizzly Bear Project. Task Number: XXX
BACKGROUND AND JUSTIFICATION:
The grizzly bear (Ursus arctos) population in the Cabinet-Yaak Ecosystem (CYE) in
northwestern Montana and northern Idaho was designated Threatened under the Endangered
Species Act in 1975 and was found to be warranted for Endangered status in 1993 (US Fish and
Wildlife Seervice 1993). Information has been collected on this population since the late 1970's
(Kasworm et al. 2010). Demographic data available include the number, location and cause of
mortality, population trend, survival and reproductive rates, and minimum counts. Based on
accumulated knowledge from trapping, observations, radio telemetry, DNA sampling at hair trap
sites, and other sources, population size in the CYE Recovery Zone is estimated to be 30-40
(http://www.fws.gov/mountain-prairie/species/mammals/grizzly/cabinet.htm). However,
resources have not been available to sample the population intensively enough during one year to
produce a population-wide estimate of grizzly bear abundance with a measure of precision.
Agencies need a statistically rigorous baseline of population size and distribution to develop and
assess policies and practices designed to promote population recovery. The SelkirkCabinet/Yaak Subcommittee of the Interagency Grizzly Bear Committee has identified obtaining
a statistically valid population estimate in the CYE as a high research priority (P. Bradford,
personal communication).
During winter and spring 2011, several agencies took the lead in identifying and securing
funding sources for the project and requested the USGS to take the lead in designing and
implementing a DNA-based study of grizzly bear population size in the CYE. Funding became
available in spring 2011 to undertake initial work to prepare for sampling the population in 2012.
This USGS-led project is a cooperative effort involving federal, state, county, and tribal agencies
with a broad base of private industry, NGO, and public partners. An interagency study team has
been established to advise on study design and to promote communication about the project.
Members include representatives from all the involved agencies.
The apparent small size of the CYE population will make it challenging to obtain a reliable
estimate of abundance. John Boulanger, an expert in mark-recapture modeling, was contracted
to estimate the power of various sampling design alternatives. He conducted simulations to
explore predicted estimate precision associated with various sampling intensities and single
versus multiple data sources (Appendix A). His findings were used to guide the development of
the final study design. Field work in 2011 entailed preparations for field sampling in 2012,
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
including landowner contacts, identifying and setting up bear rub sampling sites, and preparing
scent lure.
OBJECTIVES:
The primary objective of the Cabinet-Yaak Grizzly Bear DNA Project is to develop an estimate
with statistical confidence limits of the size of the grizzly bear population in the Cabinet-Yaak
Ecosystem of northwestern Montana and northern Idaho. This information will provide the first
population estimate with a measure of precision and will provide information needed to evaluate
grizzly bear distribution and mortality rate.
STUDY AREA:
The study area includes 9,850 km2 (2,433,988 acres) of grizzly bear-occupied area and
encompasses the 6,765 km2 Cabinet-Yaak Recovery Zone plus 3,110 km2 of occupied range that
fall outside it (Fig. 1). The eastern bounds of the study area, in northwest Montana, extend from
the Canadian border south along the western shore of Lake Koocanusa to include the Purcell
Range surrounding the Yaak River drainage, trend west to Libby at the confluence of Barron
Creek, then southeast following the Cabinet Range's eastern valley shoulder to the study area's
southern bound 15 km west of Plains. The western boundary then extends northwest, following
the Lower Clark Fork River to the town of Trout Creek , trends west into Idaho to include the
northeast side of the Bitterroot Range, then follows a northerly direction along the western valley
shoulder of the Coeur d'Alene, Cabinet and Purcell Ranges to the study area's northern bound at
the Canadian border in northeast Idaho.
The study area contains parts of three national forests (Kootenai, Idaho Panhandle, and Lolo) that
include one wilderness area, Cabinet Mountains (382 km2), and one proposed wilderness area,
Scotchman Peaks (356 km2). There are 2,175 km of maintained trail, 6,530 km of open roads,
and 8,330 km of gated, bermed, closed, and private roads in the CYE study area. The study area
is a region of diverse land use with rugged mountains in the wilderness areas surrounded by
multiple-use forest lands. National forest and corporate timber lands have active timber harvest
and forest management programs. The study area also includes an active silver and copper mine,
two proposed silver and copper mines, and a closed vermiculite mine area (14.5 km2) that is not
available for sampling due to asbestos contamination. The major valley bottoms are primarily
private lands with a mix of forested and open parcels and variable density of towns, rural
residences, small farms and ranches.
With one year of mark-recapture sampling data, a closed model must be used to estimate
population size and density. A closed model assumes the study area is geographically closed to
bear movements during the mark-recapture sampling period, so the most robust estimate of
population size will be derived by sampling all of occupied grizzly bear habitat in the CYE. This
will strengthen the estimate precision by minimizing violation of the closure assumption made in
closed mark-recapture population models and by increasing sample size. Overall, the northern
study area boundary that follows the US-Canada border for 92 km (13% of study area perimeter)
has no geographic closure, and the remainder (87%) will be essentially closed to bear
movements during the 12-week hair sampling season.
2
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
In recent years, grizzly bears in the Northern Continental Divide Ecosystem (NCDE) in Montana
have begun to reoccupy the Salish Mountains, east of Lake Koocanusa at the western edge of the
NCDE. However, there is no evidence of natural grizzly bear movement from the NCDE into the
CYE. Excluding bears transplanted from the NCDE to the CYE that returned to the NCDE, only
one grizzly bear has been documented to have moved from the CYE into the NCDE. Therefore,
the eastern boundary of the study area can be considered closed to bear movements during
sampling. Likewise, no grizzly bear movement has been documented between the CYE and the
Selkirk population to the west. In addition, genetic studies have found extremely low levels of
grizzly bear movement across the Hwy 2 corridor between the Cabinet Mountains and the Yaak
River drainage and Purcell Mountains to the north. Therefore the Cabinet Mountains can be
considered closed to bear movements during sampling. However, continuity of occupied habitat
continues into Canada from the Yaak, and there is evidence of bear movement across the
U.S./Canadian border. Including Canadian occupied habitat would further strengthen study
results and management applicability for this small population, and initiation of a concurrent
study in British Columbia would be a welcome development.
PROCEDURES:
Population size will be estimated using mark – recapture methods. Multiple hair snagging
sessions coupled with microsatellite genotyping analysis to determine species, gender, and
individual identity of bear hair samples will provide the marks and recaptures.
Field sampling: Optimal sampling design was assessed based on results from earlier studies of
neighboring populations, capture rates from historic datasets, and simulations to estimate the
precision of a range of single and multiple sampling method designs for population sizes ranging
from 25-60 (Appendix A). The study will use two noninvasive hair sampling methods: 1) barbed
wire hair corrals distributed systematically throughout the study area, and 2) bear rub sampling
in which bear hair is collected from trees and other objects that bears naturally rub against with
barbed wire. Wire barbs define discrete samples and prevent collection of samples from more
than one individual. Mixed samples cannot be used to identify individuals because it is not
possible to separate the DNA from different bears. In previous studies, 2% of the hair samples at
hair corral sites and 3% from rub objects were from more than one individual black and/or
grizzly bear (K. Kendall, personal communication). In addition, hair samples collected by
barbed wire are larger, have more follicles, and require less time to collect than hair deposited on
the original rub surface. We will conduct concurrent sampling at hair corrals and bear rubs 5
times at 14-day intervals. Hair trap sites will be moved once during the sampling season. Rub
sampling will continue for 4 weeks beyond the hair corral sampling to target the time period in
which female detection rates peak at rub sites.
Sampling design:
Hair corrals – Hair corrals will be distributed on a 5 x 5 km grid overlaying the study area (Fig.
2) and will be revisited 14 days after scent-baiting to collect the hair caught on the barbed wire.
One hair corral will be established in each of 394 25-km2 cells during session 1 (early season),
rebaited in situ during session 2, moved to a new site within each cell during session 3 (midseason), and rebaited in situ during sessions 4 and 5. Measures that will be taken to maximize
detection probability at hair corrals include using a large quantity of lure, use of scent drag lines
3
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
to the sites, and a structured approach to selection of hair snag sites designed to consistently
identify optimum habitat and terrain throughout the study area.
Hair corrals consist of 30 m of lightweight four-pronged barbed wire stapled 50 cm above the
ground around 3 or more trees (Woods et al. 1999, Mowat and Strobeck 2000, Kendall et al.
2009). To attract bears, scent lure is poured on a pile of rotten wood and other forest debris that
is approximately 1 m high and located in the center of the barbed wire corral. Each hair corral
will be baited with 5 liters of liquid scent lure composed of a 2:1 mixture of aged cattle blood
and liquid from decayed fish. Blood will be treated to prevent clotting. Field crews will create
scent trails to the corrals by dragging conifer boughs splashed with lure along the ground from
corrals to natural and man-made travel routes in the vicinity. In addition to the main blood/fish
scent lure, a secondary lure will be applied to a cloth and hung over each hair corral during each
detection session. Skunk/wolverine and cherry scents will be applied during sessions 1 - 2 and
3-5, respectively.
Hair corral site selection – To provide consistency in selecting hair corral locations, core project
staff, subunit leaders, and other technical experts with knowledge of bear ecology and movement
patterns will identify hair corral sites. Hair corrals will be placed in the highest quality bear
habitat available during early and mid-summer, in travels routes, and/or in sites that promote
dispersion of attractant odor. Field crews will be carefully trained on what characteristics to look
for on the ground to promote consistency in trap efficacy between areas. Once in the vicinity of
a pre-selected site, field crews will evaluate microsite characteristics such as adequate terrain and
arrangement of trees, bear sign, and game trails, for actual station placement. All field and
supervisory personnel will receive the same training and information. Supervisors and subunit
leaders will stay in close contact with crews to provide feedback on performance of fieldwork.
For human safety and to minimize human disturbance of bears, hair corrals will be ≥500 m from
developed areas such as campgrounds and ≥100 m from roads and trails. For both early and
mid-season hair corral locations, a primary and alternate site will be identified for each cell.
Primary sites are where data and experience indicate the probability of catching bears is highest.
Alternative sites will be in the best habitat that is obtainable and may be used when the primary
site cannot be reached due to natural hazards or time constraints.
For grizzly bear studies that sample at ≤ 5 x 5 km grid, there is no evidence that moving sites
between sample occasions is required to ensure that all bears have the opportunity to encounter
at least 1 hair corral during each detection session. However, because access in early summer
will be limited by snow in high elevations, hair corral locations will confined to low and middle
elevations. Relocating the hair corrals once during the sample season will allow capitalization
on changing availability of grizzly bear habitat.
Where cattle are present, they often trample hair corrals, tearing down the wire or fouling it with
their hair. This results in sampling heterogeneity that is a serious concern. Because there are not
many cattle ranching operations in the study area, most hair corrals can be placed in good habitat
that does not overlap with cattle. If corrals must be placed in areas with cattle, three-strand
barbed wire exclusion fences will be erected around the sites, creating a buffer of approximately
100m2 that prevents livestock from contacting the corral wire. There is some chance that this
4
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
may discourage bears from investigating the site, but bears living where cattle occur are
accustomed to fences and may not be deterred. Bear hair will be collected from the cattle
exclusion fences as well as the hair corral within them.
GIS technology will be used to facilitate the selection of the hair corral sites. The primary data
layers to be used are:
 Grizzly bear habitat selection RSF model based on grizzly bear radio telemetry studies in
the Purcell, Cabinet and Selkirk Mountains (M. Proctor, unpublished data)
 DEM/hillshade
 Google Earth (satellite imagery)
 1994 and 2000 wildfire perimeters and severity
 Land ownership, cattle allotments
 Trails, roads (forest and highway)
 Backcountry cabins, outfitter camps
 Towns, rivers, lakes
 The multilayer version of TOPO! extension for ArcGIS, which has 1:100,000 or 1:24,000
topographic maps with hillshading
 5x5 km trap grid
The following steps will be used to select locations for hair corral sites:
1. Large-scale site selection
The RSF (grizzly bear habitat selection) model is based on a multi-variate combination of
attributes such as greenness, canopy openness, elevation, etc. It provides a relative measure of
habitat quality and will be used to identify the highest quality habitat in each cell. This will
provide consistent baseline information on habitat quality for various regions within the study
area. The land ownership layer will be used to estimate access private property. Fire perimeter
and severity layers will be used to identify recent fire activity and sites will not be located in
severely burned areas. Larger patches of high quality habitat will be chosen over smaller ones.
In addition, the following prioritized list of habitat features will be used to identify placement
targets for hair corrals in mountainous areas.
a. Avalanche paths: A complex of multiple snow slide systems is better than a single chute.
Locate the corral in the middle of an avalanche complex. Only target chutes with a lush
understory; not rock and bare ground.
b. Riparian areas: Edges of creeks, rivers, lakes, and wet meadows. Only select ephemeral
streams and marshes during wet period.
c. Wet patches of habitat within dry areas.
d. Mid- and high-elevation meadows.
Expert knowledge of bear activity patterns will be relied upon whenever it is available. Local
state, federal, , and private researchers and managers with expert knowledge of bear habitat use
and activity patterns will be assembled to work with project staff to select hair corral locations.
Personal knowledge of use patterns in a cell will be used to select among the locations of the
highest apparent habitat value to identify the primary hair corral site for early and mid-season.
Obstacles to reaching the primary site will be considered when choosing alternate corral sites. In
5
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
some cases, sites will be located outside of the best habitat as predicted by expert opinion. This
would be justified when: 1) the best habitat lies on private property and the owner has not
granted permission to access it, 2) the high habitat value locations receive heavy human off-trail
use, or 3) personal experience of bear habitat use strongly suggests that sites with higher
probability of bear activity occur outside areas that appear to be high quality habitat, or there is
expert knowledge that bears do not frequent some areas.
2. Small-scale site selection
Once preliminary site selections are made, the sites will be assessed in the context of surrounding
cells. Sites should be well distributed across the landscape to have the best chance of making a
corral within range of every bear during every sampling occasion. To ensure geographic
coverage of the sampling effort, there will be at least 1 km between hair corrals in neighboring
cells and between early and mid-season session locations. However, if high quality habitat is
limited, placing a station in good bear habitat or on bear travel routes is a higher priority than
following the rule for minimum distance between sites.
Specific site placement will be done using small-scale topography and aerial photography or
satellite imagery. Maps will be made with primary and alternate hair corral locations. When
field crews reach the vicinity of a pre-selected corral site, they will select a fine-scale corral
location based on the presence of natural travel routes, game trails, and bear sign, suitable
terrain, and location of trees.
Bear rubs - The majority of bear hair collection opportunities are found on trees but bears also
rub on a variety of other things, including sign posts, cabins, power poles, and fence posts and
gates (Kendall et al. 2009). Bear rub objects will be identified during surveys of trails, roads,
power lines, and fence lines throughout the study area during 2011 (Fig. 3) and winter/spring
2012. To facilitate hair collection, short pieces of barbed wire will be attached to the rub surface.
And because bears often habitually return to rub sites, no lure will be used.
Four 38 cm lengths of barbed wire will be attached to each rub object using 2-3 fencing staples
per length. Each piece of wire has three four-pronged barbs. The wire will be positioned to
cover as much of the rubbing surface as possible; a "Z" formation is effective for most rub
objects. A small yellow tag will be nailed to each side of the object at approximately 2.1 m (7
feet) from the base to help field crews visually locate them. Each object will be tagged with a
unique identification number nailed to its base, out of sight from the trail or road. All staples and
nails will be removed once surveys are completed.
In areas with horse and mule use, rub objects along trails can be bumped by riders, stock and
their packs. Because barbed wire would pose a risk if installed on these objects, double-stranded
barbless wire will be used as a stock-friendly alternative on these pack bump surfaces. Six 23
cm lengths of barbless wire, with the strands slightly separated on each end, will be mounted
~vertically on the rub surface. Hair collects on the ends and at the staples. If barbed wire is also
used, it will be installed only above and/or below the pack bump zones.
6
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
During session 1 in 2012, all bear rub stations will be cleared of hair that accumulated after they
were established in 2011 or early in the spring. Each rub will be visited at 14-day intervals
during sessions 2-8. All hair deposited on the barbs will be collected during each visit. Based on
other studies, we anticipate collecting ~3,600 bear hair samples at rubs, ~10% of which are
expected to be from grizzly bears.
2012 field season schedule - The goal of the field schedule is to start after spring bear hunting
season and as soon as all parts of the study area have snow-free areas, so that hair snagging is
finished before most of the high-elevation huckleberries ripen (to reduce competition for bear's
attention) and the fall hunting season opens. Hair snagging and collection will occur on the
following schedule:
Session
1
2
3
Date
Jun 7-15
Jun 21-29
Jul 5-13
4
Jul 19-27
5
Aug 2-10
6
Aug 16-24
7
8
Aug 30-Sep7
Sep 13-21
Hair Corrals
Set up session 1 corrals
1st hair collection & rebait corrals
2nd hair collection, take down
corral & relocate for session 3
3rd hair collection & rebait
corrals
4th hair collection & rebait
corrals
5th hair collection & remove
corrals
Bear Rubs
Clean rubs of hair
1st hair collection
2nd hair collection
3rd hair collection
4th hair collection
5th hair collection
6th hair collection
7th hair collection & remove wire
Sample handling and management - All hair from each set of barbs will be considered a separate
hair sample. Hair will be removed from each barb and placed in 2.5 in x 4.25 in paper coin
envelopes. Hair samples will be brought out of the field as soon as possible and stored in a
desiccant chamber containing silica gel beads. Samples will be protected from UV radiation and
freeze- thaw cycles to prevent DNA degradation.
The following data will be recorded on the envelope for each hair sample: names of collectors,
date of collection, hair corral grid cell or bear rub number and session number, and barb number
or height. A summary of hair samples collected at a corral or rub site during each visit will be
recorded on a site data form. Samples will be assigned a unique number that appears on twin bar
code labels attached to the hair collection envelope. One label stays attached to the envelope.
The second “piggyback” label can be peeled off and stuck to the site data form. The barcodes
are scanned into the database as the key reference field for hair sample information, hair corral
and bear rub summary data, and genetic data from the contract laboratory.
Based on previous bear hair snagging studies and anecdotal information regarding the relative
density of grizzly bears in the CYE, 20,000 bear hair samples are expected to be collected. The
peel and stick label means there is no need for field crews to copy the sample number onto the
data form, eliminating error caused by illegible handwriting. Scanning also eliminates
transcription error. We will use non-contact scanner wands that can read a barcode up to 12 in
7
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
from the label. The scanner input is registered as keystrokes to the computer; therefore the
scanning wand does not require any special software. Studies have shown that a proficient data
entry operator will make 3,000 errors per million keystrokes, whereas barcode data entry has an
error rate of about 1 in 3 million (source: http://www.kcsi.ca/barcoding_adv.html). The barcode
label system will maximize the speed and accuracy of data entry by eliminating over 250,000
data entry keystrokes. We will also use an auto-sense stand that provides hands-free scanning,
which will reduce repetitive movement strain. A label company will manufacture the barcode
labels, because laser printer ink quality is not adequate for field use. The label company will
create a proof of the barcodes to allow testing with the scanner wands purchased for the project
before the labels are manufactured. The labels will have the sample number printed underneath
the barcode so that the number is visually available in perpetuity.
Quality control during field activities - There will be differences in experience and training
among the personnel implementing the field sampling protocols. To mitigate this, the protocols
have been simplified as much as possible and have few special rules so it is easy to perform them
correctly. Field crews will be supplied with kits that contain everything needed for snag station
set up and sample and data collection to help keep every component consistent. All personnel
will receive the same technical training. Design assumptions will be explained in detail to enable
field personnel to make intelligent choices when unanticipated situations arise.
A system of formal quality checks will be instituted in the field. Due to the large size of the
study area and number of technicians involved in fieldwork, staff will be dedicated to fieldwork
inspections to ensure consistency in protocol execution. The study area will be divided into north
and south subunits. Subunit leaders will direct field crews and ensure that hair corrals are
operating in all cells and that corrals and bear rubs are visited every 14 days. To ensure that field
protocols are performed correctly and consistently, subunit leaders along with project core staff
will circulate among crews in the field to monitor work quality and provide timely feedback to
field technicians. Shoddy work or repeated failure to follow study protocols will be grounds for
dismissal.
To promote understanding and inspire confidence in the scientific methods among field
technicians, all project personnel will receive training on project methods. The training will
describe scrutiny that the methodology has undergone, e.g. quality controls built into the study
design, peer review of the field methods and data analysis, and blind testing of the genetic
analysis. Training also will explain how failure to follow study protocols reduces the credibility
of the science. Rumors will be taken seriously and dealt with by direct communication with
staff. Information about project activities and results will be disseminated to affected personnel
in partner agencies. The personnel conducting the genetic analysis will be given feedback about
fieldwork and results to promote lab personnel interest and sense of ownership in the study.
The recovery of grizzly bears in Montana is socially controversial, and pubic outreach is
essential. To aid our partners in this project, talks and materials explaining the project will be
presented to partner agency personnel and private organizations and landowners, to the general
public and to groups affected by grizzly bear management or project activities.
Genetic analysis:
8
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Microsatellite genotyping analysis will be used to determine species, sex, and individual identity
of the bear hair samples. Genetic analyses will be conducted using methods that have a wellestablished track record and have been published in a peer-reviewed technical journal, e.g.
Paetkau 2003. Peer-reviewed lab-specific error rates associated with those methods will be a
factor considered when proposals for conducting the genetic analysis are evaluated.
Electrophoretic scores will be examined by at least two technicians experienced with: 1)
microsatellite analysis, 2) analysis of small samples of non-invasively collected hair, and 3) ursid
genetic marker systems. The chief geneticist will make final decisions regarding divergent
scores. Microsatellite separations will be accomplished using laser-induced fluorescence
denaturing capillary electrophoresis. Automated processing will speed analysis and reduce the
risk of lane jumping and loading errors during electrophoresis. Allele scores will be calibrated to
published designations to facilitate comparison with other data sets. Samples will be kept at the
lab until the project is completely finished to facilitate re-running samples for error control. All
materials related to the project will remain the property of the U.S. Geological Survey, including
remaining hair samples and DNA extractions, raw data, and electropherograms.
Before DNA extraction, each hair sample will be examined using a light microscope to identify
samples with follicles. Samples without follicles will be excluded from further analysis. To
identify grizzly bear samples, 1 underfur or guard hair follicle will be clipped into a polymerase
chain reaction (PCR) premix for a mitochondrial DNA (mtDNA) species test. Black bear
samples will not be analyzed further. For each grizzly bear sample with ≥3 follicles,
approximately 1 cm will be clipped from the follicle root of up to 10 hairs. Hair follicles will be
placed in tubes for DNA extraction using Quiagen kits. The extracted genetic material will be
amplified using standard PCR methods. PCR and electrophoresis will be done in separate rooms
to prevent contamination of genomic DNA samples with amplified DNA.
A set of 8 variable microloci will be used to identify individual grizzly bears: G10B, G1D,
G10H, G10J, G10M, G10P, CXX110, and MU50. from each sample with amplified DNA.
Amelogenin will be used to identify sex. These data will be used to build detection histories for
each individual identified for use in mark-recapture modeling of population abundance. One
sample from each individual grizzly bear identified will undergo further analysis to determine
the degree of relatedness and amount of fragmentation occurring within the population and
facilitate comparison to other datasets. To accomplish this, a 21-locus genotype will be
determined using the following additional markers: A06, CPH9, CXX20, G1A, G10C, G10L,
G10U, G10X, MU23, Mu51, MU59, Msut2, and P07. The selected marker system has adequate
power to differentiate up to 100 individual grizzly bears in the CYE.
Based on previous bear hair snagging studies, it is expected that 70% of the hair samples will
contain enough genetic material to warrant extractions. There are several arguments for
analyzing as many samples as possible that are only offset by the cost. In general, as more
samples are genotyped, the number of unique grizzly bears identified increases. Additionally,
having duplicate high-quality genotypes of an individual means each genotype is more reliable.
The proposed budget includes the cost of genetic analysis of all samples with ≥ 3 follicles. If the
total number of samples collected exceeds 20,000, the 3-follicle threshold for analysis may be
raised due to funding constraints.
9
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Quality control during genetic analyses - To limit data entry errors, the genetics laboratory will
build their database around a copy of the hair collection field database. All hair samples will be
identified with a unique number that is embedded as a bar code on the hair collection envelope.
A bar code scanner will be used to link genetic results to associated field data via a relational
database system. The laboratory will be required to rigorously document all sample handling
protocols sufficient to meet chain-of-custody requirements. Upon completion of analysis,
merging datasets and further error screening will be a cooperative effort between field and lab
scientists to identify and address lab results that appear to be inconsistent with bear biology.
Given the variability of field conditions and wild populations, minimizing laboratory analysis
error is a primary concern in DNA-based population studies. The analytical laboratory will be
required to use stringent error prevention protocols, including 1) thorough training and close
supervision of technicians, 2) early culling of low quality samples, 3) confirmation of 1- and 2mismatch (MM) pairs of genotypes and 3-MM pairs consistent with allelic dropout, 4) repeated
analysis of samples from one individual detected long distances from each other, and 5)
confirmation of identified individuals with an independent set of 13 microsatellite markers.
Extraction, PCR, and electrophoresis blank and positive controls will be used at all times and
records of these controls will be maintained.
To help establish allelic dropout and false allele error rates and monitor the reproducibility of the
genotyping, blind testing of the laboratory results will be conducted throughout the course of
genetic analysis. Tests will consist of multiple submissions of a random subset of hair samples
from hair corral sites, as well as hair samples from known bears handled for research and
management purposes. Only samples with 20 or more follicles will be used. These duplicate
samples will be added to the field samples before being sent to the laboratory in a way that will
make them indistinguishable from original field samples to the laboratory staff. Only the project
leader and database manager will know the identity of the test samples and they will be well
documented to prevent inclusion in the data used for population estimates. Test samples will
include hair from grizzly bears and black bears, siblings, parents and offspring, and mixed hair
from more than one individual and/or species,. Gender data from known bears will be withheld
to monitor the accuracy of gender determinations.
Genotyping error rates will be estimated empirically from the frequency of 1, 2, and 3 pairs of
mismatched loci (Paetkau 2003). Patterns in the distribution of problem samples will be tracked
to check for potential laboratory processing problems and the samples involved will be flagged
for closer examination. For example, frequently incomplete genetic data may suggest that too
many low quality (error-prone) samples were retained for analysis during initial hair sample
vetting Initial lab observations, hair samples, DNA extracts, and capillary electrophoretic
results will be readily accessible to the project leader at all times during the analysis period.
The number of people authorized to handle samples and have access to data will be kept to a
minimum. To ensure project integrity, data, hair samples and extracted DNA will be kept in
secure facilities. Original data forms will be kept in locked file cabinets and hair samples will be
stored in a locked room in the project offices in Troy during the field season and in West Glacier
after August 2012. Samples will be hand delivered to the analytical laboratory by project staff, if
10
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
feasible, or sent by registered courier. As samples arrive at the laboratory, they will be logged
against a list of samples sent to the lab provided by the project office. From the time of receipt
of samples by the analytical laboratory, all hair samples and DNA extractions will be kept in
locked rooms when authorized personnel are not present.
Laboratory facilities will be protected by a security system that tracks access by individual and
prevents past employees from entering the facilities. Computers used to conduct analyses and
store results shall be protected from unauthorized access through password protections, stringent
firewall and antivirus software, and/or having no hard-wire connection to the Internet. Field and
genetic databases must be backed-up nightly with weekly backups stored off-premises.
Data analysis: A closed population estimation model will be used to estimate population
density. Demographic closure is a reasonable assumption since there will be no recruitment
during the sampling period and deaths are rare in grizzly bear populations in the study area.
Because the study area is large and includes virtually all occupied grizzly bear habitat in this
ecosystem except along the northern boundary, geographic closure violation should play a small
role in lowering recapture probability in this study. The density module in Program MARK will
be used to adjust density estimates for the lack of closure based on movement data (telemetry,
management captures, etc) for grizzly bears in the CYE and Purcell Mountains in southern
British Columbia.
Model selection – Data from our 2sampling methods (hair trap and bear rub) and from other
records (individual bears known to be present through other sampling methods(e.g. physical
capture, genotyped scats) obtained by other entities will be used to construct individual bear
encounter histories for use in Huggins closed mark–recapture models (Huggins 1991, White and
Burnham 1999, Boulanger et al. 2008, Kendall et al. 2008, 2009). Mark–recapture analyses will
be performed in Program MARK (White and Burnham 1999). We will use a stepwise a priori
approach to mark–recapture model development. To determine the best structure for each data
type, we will model hair corral and bear rub data separately. We will pool the other 2 data types
and use them as the first sample occasion for each exercise. For example, in the hair corral
models, we will combine bear rub and physical capture detections as the first sample session
followed by the 5 hair corral sessions. We will then combine the most supported hair corral and
bear rub models into a single analysis in which we construct encounter histories for each bear
detected during 11 sampling occasions as follows: other sampling methods(1), detection during 5
hair trap sessions (2–6), and detection during 5 bear rub survey sessions (7–11). Individual
grizzly bears detected in the CYE through scat analysis will be included in session 1 detection
histories if the scat was documented to have been deposited during Jun 7 - Aug 24, 2012 and the
genetic analysis is done by the same genetic lab conducting the hair analysis and is held to the
same error control standards.
Our candidate models will include gender, bear rub sampling effort for each bear, history of
previous live capture, and distance to edge covariates. Rub sampling effort is the number of days
since the last survey summed for all bear rubs within an idealized home range of each bear. A
bear will be considered to have a history of live capture if it had been captured or handled,
regardless of method, at any time before or during hair corral sampling. Distance to edge is the
distance of the average capture location of each bear from the open (northern) boundary.
11
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
The decision on whether it is appropriate to pool data for males and females in the population
estimate will be based on program MARK tests for difference in capture probabilities of sexes.
To test for the presence of genetic misidentification of individual bears, the estimate of
population size and the number of new individual detected will be plotted as a function of the
number of sampling occasions. Misidentification will show as an increase in both of these
quantities with increasing occasions. No misidentifications would result in these quantities being
flat lines. Lack of closure can cause misinterpretation of such plots, so care must be taken to
avoid this effect.
We will evaluate relative support for candidate models with the sample size-adjusted Akaike
Information Criterion for small sample sizes (AICc). We will obtain estimates of population size
as a derived parameter of Huggins closed mixture models in Program MARK (White and
Burnham 1999, White et al. 2001). Calculation of 95% log-based confidence intervals about
those estimates will incorporate the minimum number of bears known to be alive on the study
area (White et al. 2001). All unique grizzly bears identified at hair corrals, bear rubs, confirmed
sightings, reliable scat genotypes, and physical captures will be combined to derive a minimum
population count. Population estimates will be averaged based on their support in the data, as
indexed by AICc weights, to account for model selection uncertainty (Burnham and Anderson
2002).
Metadata:
Spatial data will be documented using ARC Catalog in ARC GIS v.8.3. Spatial Metadata
Manager System (SMMS) v. 3.2 (Intergraf Inc.) will be used to document all data collection
management, forms, sample handling, and genetic and data analysis. Field notes, initial lab
observations, hair samples, DNA extracts, and electrophoretic results will be archived in acidfree storage containers by Kate Kendall at the USGS Glacier Field Station, Science Center,
Glacier National Park, West Glacier, MT.
PERMITS and COMPLIANCE:
Permits to access lands under a variety of jurisdictions and conduct research activities pursuant
to this project will be obtained from a number of state, federal, and tribal agencies, private
timber, mining, and utility companies and cooperatives, as well as more individual private
landowners for a copy of the private landowner permission letter. US Fish and Wildlife Service
and US Forest Service requirements for compliance with the Endangered Species Act and
National Environmental Policy Act will be met.
EXPECTED PRODUCTS:
Papers in peer-reviewed scientific journals:
 Grizzly bear population density and distribution in the CYE
 Genetic structure of the CYE grizzly bear population
12
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Symposia, conference, workshop presentations:
 International Conference on Bear Research and Management: 2013
 The Wildlife Society, national meeting, NW Section, Montana Chapter: 2012, 2013, 2014
Progress and completion reports:
 Annual progress and final completion report to IGBC, CYE Subcommittee
 Annual progress and final completion report to all partner agencies and companies
Consultation and reports to partner agencies:
 Interagency Grizzly Bear Committee, CYE Managers Subcommittee in assistance to
grizzly bear recovery effort.
 Grizzly bear status for National Forest activities: Kootenai, Idaho Panhandle, and Lolo
National Forests.
 Baseline information for Montana Fish, Wildlife, and Parks and the US Fish and Wildlife
Service grizzly bear monitoring program and conservation strategy.
Data and maps:
 A variety of spatial formats of grizzly bear density and distribution information.
 Database of 21-loci genotypes of unique grizzly bears from hair sampling.
 Data on the distribution and characteristics of bear rub trees.
Website:
 Information on the CYE study background, objectives, study area, methods, results and
conclusions.
PUBLIC OUTREACH:
The CYE Grizzly Bear DNA Project will be active in giving presentations to public and private
groups, including the service organizations, local high schools, community groups, backcountry
horsemen, and other requesting information.
SCHEDULE:
_______________________________________________________________________
 Establish bear sign survey routes
May - Oct 2011
 Acquire scent lure components
Jun - Oct 2011
 Establish CYE Study Team and meet to lay ground rules
8 June 2011
 Draft study plan circulated for review
Nov 2011
 Report to CYE Subcom, Bonners Ferry, ID
9 Nov 2011
 Study plan finalized
Dec 2011
 Hair corral site selection
Dec 2011
 Report to IGBC, Missoula, MT
1 Dec 2011
 Advertise seasonal positions/update web site
15 Dec 2011
 Review and rate applicants/ call references/offer jobs
15 Jan - 28 Feb 2012
 Continue logistical planning/preparation
Jan 2012 – May 2012
Purchase equipment
13
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Prepare data forms, field protocols
Contract for genetic analysis
Bottle lure
 Conduct hair sampling field work:
Train field technicians
Conduct hair trapping
Conduct rub tree surveys biweekly
Conduct formal quality checks of fieldwork
Take down hair snags and rub tree markers
 Progress report to partners
 Genetic analyses:
Extraction, amplification,
Species, sex, and individual identity
 Progress report to partners
 Data entry, QA/QC
 Data analysis, population modeling
 Final report
May - Aug 2012
May, November 2012
1 Dec 2012 –Jun 2014
May, Nov 2012, 2013, 2014
June 2012 – March 2014
Summer 2014
December 2014
PERSONNEL
Principal Investigators:
Mike Mitchell, Montana Coop. Wildlife Research Unit, University of Montana, Missoula,
59812. Phone: 406.243.4390.
Email: mike.mitchell@umontana.edu
Katherine C. Kendall, Science Center, Glacier National Park, West Glacier, MT 59936-0128.
Phone: 406-888-7994.
Fax: 406-888 5835. Email: kkendall@usgs.gov
Research Assistants (University of Montana, Coop. Wildlife Research Unit employees):
Project Coordinator: Kris Boyd
Database and GIS Coordinator: Dan Kotter
Information and Technology Specialist: Marilyn Blair, USGS.
TRAINING
Training is critical to ensuring that field crew personnel correctly perform technical duties and
safely travel and work under remote and rugged conditions. Crew members will experience a
diverse range of field conditions and wide range of specific duties. As such, project training is
designed to address the full extent of what crews may encounter over the course of the field
season.
Prior to training, crewmembers will be given recommended reading lists and suggestions for how
to prepare for the upcoming season. Recommended readings include technical articles detailing
the methods employed by the project, popular materials about bears and the study area, and
safety-related articles about river crossings, diseases, and general first aid. Employees also
14
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
receive a list of ways that crewmembers can better prepare themselves for the physical demands
of the season. This list includes developing an exercise program, purchasing appropriate gear,
and becoming familiar with its proper use.
Training elements:
Training will provide crew members with the background on project methods and objectives
necessary to fully understand the scope and complexity of the project. Protocols, data forms, and
use of handheld Global Positioning System (GPS) units will be reviewed. Basic bear biology
will also be addressed. Technicians will be introduced to the significance of interacting with the
public and personnel from other agencies. Crews will be given adequate background on the
project to understand the importance of public perception and how to address questions that they
will likely encounter over the course of the field season. To reinforce this message,
representatives from most project cooperators will present the unique issues of their respective
agencies.
Wildlife encounters, and bear safety in particular, will be extensively discussed. This will
include presentations on how to work safely in bear country (a video sanctioned by the
International Association of Bear Research and Management) and indoor and hands-on training
in the proper use of bear spray.
Field training will reinforce many of the topics covered during classroom presentations with onthe-ground activities. Crews will be instructed in how to construct hair corral sites, complete
data forms, and properly collect samples. Crewmembers will be instructed in the use of
topographic maps, GPS units, and compasses in navigating to pre-selected hair corral sites.
Once the general location is reached, it will be the individual crew's decision as to where to
actually construct the hair snag station. The project leader will instruct crewmembers the on
"micro-site" characteristics that define a high quality hair corral location. Personnel will be
instructed in conducting rub tree surveys, which consist of hiking trail segments, locating
previously identified bear rubs, and collecting hair samples. Crewmembers will be trained to
identify and set up rub stations that may be bumped by riders and pack stock.
All field technicians will receive 2 days of wilderness first aid training. The focus will be to
teach crewmembers to avoid injuries by making good decisions. Specific topics include
dehydration, hypothermia, hyperthermia, anaphylaxis, scene assessment, snowfield and river
crossings, infections, wound dressing, and communications. Crews will be trained in how to
deal with emergency situations that may occur under extreme field conditions, and how to
effectively adapt and use the materials they will likely have available, such as backpacks and 2way radios.
Project training will include job hazard analysis and the importance of working safely, letting
people know where you are working each day, and communication protocols for radios. Course
work in defensive driving and what administrative procedures to follow if you are injured on the
job (reporting requirements for OWCP), conduct, and ethics will be covered, as will wilderness
15
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Leave No Trace ethics, fire safety, food storage rules, and what to do when various wildlife
species are encountered.
WORK AREAS
Office work primarily will be conducted at:
 USGS Glacier Field Station, Glacier National Park, West Glacier, MT
 Lincoln County offices, Troy, MT
Fieldwork will be conducted on- and off-trail throughout the study area.
SAFETY
A Safety Plan for the Northern Divide Grizzly Bear Project will be submitted to the Safety
Officer for the Northern Rocky Mountain Science Center. The plan contains the following
sections:
 Description of the office, laboratory, and field work including the potential for
encountering bears.
 Safety Officers for the USGS-NRMSC
 Hazards, hazard prevention and controls, including chemical hazards (bear scent lure,
indicator silica, bear pepper spray), biological hazards (mountain lions, bears, ticks,
mosquitoes, giardia, hantavirus, West Nile virus), physical (high altitude, UV light,
falling, driving), environmental (lightning, hypothermia, snowfield crossing, wind).
 Safety precautions (communication with backcountry crews, handling barbed wire,
transporting and handling lure, avoiding bear encounters, precautions used at snag sites).
 Fitness and health of Project employees (selection criteria).
 Training requirements (wilderness first aid, driver safety).
 Safety equipment (field gear, personal protective equipment).
 Communication plan.
 Emergency plan.
 Prevention of injury, illness, and property damage.
16
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
HAZARD ASSESSMENT
Workplace Assessed: USGS Glacier Field Station, West Glacier, MT; Lincoln County office,
Troy, MT; Cabinet-Yaak Ecosystem, northwestern MT and northeastern ID.
Assessed By: Katherine Kendall
Date of Assessment: 10/11/2011
Northern Divide Grizzly Bear Project Job Tasks Assessed:
 Office Work
 Field Work
 Lure Bottling and Wire Cutting
 Travel
OFFICE WORK
HAZARDS
CONTROLS
Eye Strain
Ensure proper lighting. Ensure computer monitor and document copy
stand are at approximately the same height and distance. Reduce computer
screen strain by using flat screen monitors.
Wrist Strain
Ensure computer keyboards are adjusted so that the elbows are at a 90degree angle and arms and hands are parallel to the floor. Use wrist rests
or other supports so that wrists are maintained in a neutral position.
Neck/Shoulder
Fatigue
Ensure monitors are properly adjusted so that the top of the screen is
slightly below eye level and the screen is between 18 and 28 inches away.
Slips/Trips/Falls
Lifting
Electrical Shock
Walking
Falling off
Furniture
Cutting tools
File
Cabinets/Shelves
PPE
Required
Use good housekeeping practices. Secure tripping hazards (cords) to the
floor. Do not leave file drawers open when unattended.
Use proper lifting techniques. Get assistance when necessary. When
lifting, keep the load close to the body and lift with the legs.
Ensure equipment is properly maintained and grounded. Protect electrical
cords from damage by using cord covers. Do not overload outlets.
Be alert of walking surface. Wear flat shoes with a non-skid sole.
Use a step stool. Do not climb on furniture.
Cut in the direction away from hands and body.
To avoid tipping, fill the bottom file first. Do not open more than one
drawer at a time. Place heavy objects in the bottom shelves/drawers.
17
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
FIELD WORK
TASK: Fieldwork includes hiking on maintained trails and hiking off-trail through dense forest,
constructing hair snag stations, and pouring bear scent lure.
HAZARDS
Exposure to the
elements
Snake Bites
Animal
Attacks/Bites
CONTROLS
PPE
Required
Wear proper clothing. Be aware of exposure duration and limit duration
if necessary. Be knowledgeable of the symptoms of exposure related
illnesses; e.g. hypothermia, hyperthermia. Keep hydrated. Always carry
rain gear and warm layers when in backcountry
Wear proper field boots or snake chaps. Do not harass/kill snakes.
Make noise while hiking to warn bears and other wildlife of your
presence. Be especially careful when visibility is limited, it is windy, or
there is background noise (e.g. waterfall).
Bear Pepper
Spray, First
Aid Kit, Radio
Do not approach animals. Use caution and composure when encountering
animals.
Carry pepper spray at all times. Approach lure sites with plenty of noise.
If evidence of bear in lure site area, leave and come back later in the day.
Place warning signs near the station site and at one or two locations on
the approach to the site. Sign any other areas that seem appropriate to
warn people that may use the area.
Insect bites and
stings
Noise
Ankle Injuries
Knowledge of and avoidance of insects that bite and sting. Knowledge of
diseases that can be caused by insect bites (e.g. West Nile, Rocky
Mountain Spotted Fever, etc.) Knowledge of any allergies to bites or
stings. Do not wear perfume or cologne. Know where to obtain first aid.
Wear proper hearing protection devices.
Wear proper field boots with ankle support.
Insect sting
reliever,
personal bee
sting
medications (if
allergic)
Ear Plugs
Appropriate
Field Boots
Eye Injuries
Wear appropriate eye protection as necessary.
Safety Glasses
Injuries from
Barbed Wire
Take your time when rolling barbed wire out, using staples long enough
to insure wire is tight on trees. When taking the barbed wire down to
move to the next site, roll into a small, tight roll. Use foam to wrap wire
in and place bundle on pack.
Double-palmed
Gloves
Hatchet/Fencing
Tool Injuries
Use extra caution when using hatchet to trim off lower limbs on trees. Be
sure lower legs and feet are out of the path of the hatchet swing.
Double-palmed
Gloves
18
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Slips/Trips/Falls
due to unstable
footing, steep
terrain, and dense
vegetation, etc.
Be observant of walking/working surfaces. Take your time when
climbing/descending mountain trails. Use handholds when possible on
steep ground. Avoid slick/wet surfaces such as logs, rocks, or moss. Take
time to select best route if possible. Do not run downhill. When ground
surface is hidden due to dense vegetation, place feet deliberately or
slowly before putting full weight down.
Liquid scent lure
bear attractant
Make sure lids are tight on lure containers and exterior of bottles are
clean. When transporting in truck, use carrying case (supplied). When
pouring lure on woodpile don’t stand directly over the jar, hold at arms
length, and pour slowly, being careful not to splash on clothes or boots.
Latex gloves
When hanging scent in tree avoid getting lure on clothes/body. If lure
gets on clothes/body wash immediately, hang affected clothes, as you
would with bear attractants when camping.
Head Injuries
When using objects to place lure string over tree limbs, be aware where
flight of object will be on the downfall.
Always be cautious of falling trees/snags, especially when hiking in
burned areas and in high wind.
Lightning /
Storms
Know weather forecast before going into the field.
Plan to be at a lower elevation later in the day to avoid late afternoon
mountain storms. If caught in a storm descend to a safe area and wait out
the storm before continuing planned hike. Avoid trees, rock caves, and
exposed areas.
Other injuries
Two person teams are preferred for all aspects of the fieldwork for this
project. If a crew person has to miss a field day, a substitute, if available,
will be assigned to work with the other crew member. It is acceptable for
personnel with experience working / traveling in grizzly bear country and
mountainous terrain to work / travel alone if equipped with radio, and
other required safety equipment.
All crews are
given first aid
kits and
communication
equipment
(radio or cell
phone)
Crews are assigned GPS, compass, and maps. Crews are trained in
orienteering, first aid, river crossing, and snow crossing,
19
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
LURE BOTTLING
HAZARDS
CONTROLS
PPE
Required
Liquid scent lure
bear attractant
Ensure area is well ventilated. Wear protective clothing and thick rubber
gloves. Be cautious when reaching into or transferring material from
barrels due to threat of puncture wounds from fish bones. Be careful not to
contaminate any open sores. Be careful not to splash lure components into
face. Wash clothes and body thoroughly when work is completed each
day. Scrubbing with disinfectant solution is advised.
Thick rubber
gloves,
protective
clothing
(coveralls,
boot covers),
water
available to
rinse eyes,
first aid kits,
and phone to
call
Emergency
Medical
Services if
required.
CUTTING WIRE
HAZARDS
CONTROLS
Eye Injuries
Wear appropriate eye protection as necessary.
Injuries from
Barbed Wire
Take your time when cutting and rolling barbed wire. Use only sharpened
tools. Use appropriate gloves when handling barbed wire.
PPE
Required
Safety
Glasses
Doublepalmed
Gloves, long
pants, thick
soled shoes
20
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
TRAVEL
HAZARDS
CONTROLS
PPE
Required
Secondary Roads, Forest Service Roads, ATV trails, Bicycle trails
Bicycle and ATV
Travel
Obey traffic and trail laws. Adjust speed to road, trail, and weather
conditions.
Helmet
designed for
the activity
(i.e. biking or
ATV riding)
City, Highway, Secondary Roads, Forest Service Roads
Motor Vehicle
Accidents
Obey traffic laws. Adjust vehicle operation to road and weather
conditions. Employ defensive driving techniques.
Drive slow – maximum speed limit on forest roads is 25 mph
Uneven Surfaces
Deer and other
wildlife
Dust
Reduced Visibility
Slick, Snowy, or
Icy Roads
Reduce speed appropriately
Stay alert, use caution, and drive defensively.
Drive with windows closed. Reduce speed.
Ensure windows/mirrors are free from snow and ice. Drive with headlights
on. Reduce speed appropriately.
Use studded or chained tires, reduce speed, and increase following
distances.
21
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
LITERATURE CITED:
Boulanger, J., K. C. Kendall, J. B. Stetz, D. A. Roon, L. P. Waits, and D. Paetkau. 2008.
Multiple data sources improve DNA-based mark-recapture population estimates of grizzly bears.
Ecological Applications 18:577-589.
Burnham, K. P., and D. R. Anderson. 2002. Model selection and inference: a practical
information theoretic approach. Springer, New York, New York, USA.
Huggins, R. M. 1991. Some practical aspects of a conditional likelihood approach to capture
experiments. Biometrics 47:725-732.
Kasworm, W. F., H. Carriles, T. G. Radandt, M. Proctor, and C. Servheen. 2010. Cabinet-Yaak
grizzly bear recovery area 2009 research and monitoring progress report. U.S. Fish and Wildlife
Service, Missoula, Montana. 78 pp.
Kendall, K. C., J. B. Stetz, J. Boulanger, A. C. Macleod, D. Paetkau, and G. C. White. 2009.
Demography and genetic structure of a recovering grizzly bear population. Journal of Wildlife
Management 73:3–17.
Kendall, K. C., J. B. Stetz, D. A. Roon, L. P. Waits, J. B. Boulanger, and D. Paetkau. 2008.
Grizzly Bear Density in Glacier National Park, Montana. Journal of Wildlife Management
72:1693–1705.
Mowat, G. and C. Strobeck. 2000. Estimating population size of grizzly bears using hair capture,
DNA profiling, and mark-recapture analysis. J. Wildl. Management 64(1): 183-193.
Paetkau, D. 2003. An empirical exploration of data quality in DNA-based population inventories.
Molecular Ecology 12:1375-1387.
US Fish and Wildlife Service. 1993. Grizzly bear recovery plan. Missoula, MT,. 181 pp.
White, G.C. and K.P. Burnham. 1999. Program MARK: Survival estimation from populations of
marked animals. Bird Study Supplement 46:120-138.
White, G.C., K. P. Burnham, and D. R. Anderson. 2001. Advanced features of Program Mark.
Pages 368–377 in R. Fields, R. J. Warren, H. Okarma, and P. R. Seivert, editors. Wildlife, land,
and people: prioritites for the 21st century. Proceedings of the second international wildlife
management congress. The Wildlife Society, Bethesda, Maryland, USA.
Woods, J.G., D. Paetkau, D. Lewis, B. McLellan, M. Proctor, and C. Strobeck. 1999. Genetic
tagging of free-ranging black and brown bears. Wildl. Society Bulletin 27(3):616-627.
CLEARANCES:
Submitted by: ____________________________________________________
Principal Investigator
Date
22
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Center approval: __________________________________________________
Center Director
Date
23
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Figure 1. Cabinet-Yaak Grizzly Bear DNA Project study area in northwestern Montana and
northern Idaho.
24
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Figure 2. Cabinet-Yaak Grizzly Bear DNA Project study area with 5 x5 km hair snag sampling
grid.
25
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Figure 3. Bear rubs identified in the CYE and set up for monitoring as of Sep 28, 2011.
26
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
APPENDIX A: Optimal design for the Cabinet-Yak grizzly bear DNA mark-
recapture study
September 27, 2011
John Boulanger, Integrated Ecological Research, 924 Innes, Nelson, BC , V1L 5T2,
boulange@ecological.bc.ca, www.ecological.bc.ca
Introduction
This report details the investigation of optimal sampling designs for the Cabinet-Yak grizzly bear
population. This population has potentially low numbers and therefore a pertinent question is
the sampling intensity needed to obtain reliable population estimates. I use data from previous
hair snag projects to asses a hair-snag alone design and then use simulations to explore gains
from adding rub tree and management bears as supplemental data sources.
Methods
Two approaches were used to assess optimal sampling designs.
Analysis of historic DNA mark-recapture data sets
The results of analyses of historic hair snag (HS) data sets were used as a first step in
determining the feasibility of a hair-snag alone sampling design with likely population sizes in
the Cabinet Yak Study area. As part of results presented in Proctor et al. (2010), I conducted an
analysis of historic DNA-based population estimates that attempted to predict coefficient of
variation as a function of estimated population size and hair-snag detection probabilities. When
considered as quadratic terms, both detection probability and estimated population size were
significant predictors of the coefficient of variation (at =0.1) (Proctor et al. 2010). Of the
historic designs used for DNA sampling, the most applicable design is the 5x5 km cell size with
sites not moved between sessions. This design has been used to obtain reasonably precise
population size estimates for the Jumbo Pass, South Selkirk, and Highway 3 DNA markrecapture projects (Proctor et al. 2007, Proctor et al. 2010). Therefore the results of these
projects were a focus of the analysis and detection rates from these projects were used to inform
simulations.
Population sizes of grizzly bears in the Cabinet-Yak probably range from 25 to 60 individuals
(Kate Kendall, per. Comm) and therefore results were evaluated across this range of population
size.
Simulations to further explore precision of estimates with HS alone
and multiple data source sampling designs
Previous research (Boulanger et al. 2008, Gervasi et al. 2008, Kendall et al. 2008, Kendall et al.
2009, Gervasi et al. 2010) has shown that use of rub tree and management bear data as separate
sampling sessions can improve estimate precision and robustness of estimates. I used simulation
modeling to predict gains in estimate precision by the addition of rub tree and management bear
data.
27
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
The multiple data source designs that were considered were based upon likely budgetary
constraints and limits on field logistics (Table 1). As a baseline for simulations a hair snag
alone design with 4 and 5 sessions (HS4 and HS5) was considered. For each baseline HS design,
2 or 5 RT sessions were added, as well as 1 session for management bears on the grid during
sampling. Hair snag probabilities were based upon the mean detection rate for historic designs
that used the 5x5 km cell size (Proctor et al. 2007). Rub tree detection probabilities were based
upon mean detection rates in Glacier National Park (Kendall et al. 2008) and the NDP project
(Kendall et al. 2009, Stetz et al. 2010). The detection rate of management bears assumed that 7
collared bears would be present during sampling, and the detection rate was simply the number
of management bears (7) divided by the simulated population size. The estimated proportion
of the population sampled was estimated by the formula p*=1-[(1-pHS)HS(1-pRT)RT(1-pMG)]
where pHS, pRT, pMG where detection probabilities from hair snags, rub trees, and management
bears and HS, RT, were the number of HS and RT sessions. This is a deterministic estimate of
the proportion of the population sampled with each design. It can be seen that the HS4 samples
72-79% of the population whereas the more intensive HS5 RT5 MG1 would sample 89-96% of
the population. These figures should be interpreted in a relative rather than absolute fashion
since it is possible that each of the sampling methods targets a different segment of the
population which would bias these estimates.
Design
Table 1: Sampling designs used in simulations.
Sessions
Detection probabilities
Number P*(design)
HS RT MG HSHS-F
RTRTMGA
M
F
M
M
F
4
0
0
0.32
0.27
0
0
7
0.79 0.72
4
2
1
0.32
0.27
0.19 0.05
7
0.90 0.82
HS4
HS4 RT2
MG1
HS4 RT5
4
5
1
0.32
0.27
0.19 0.05
7
0.95 0.84
MG1
HS5
5
0
0
0.32
0.27
0
0
7
0.86 0.80
HS5 RT2
5
2
1
0.32
0.27
0.19 0.05
7
0.93 0.87
MG1
HS5 RT5
5
5
1
0.32
0.27
0.19 0.05
7
0.96 0.89
MG1
A
The detection rate simulated for management bears was calculated as the number of
management bears (7) divided by the simulated population size.
Simulations were run across a range of 25 to 60 grizzly bears with an assumed even sex ratio.
The simulation generation model had sex-specific hair snag and rub tree detection rates with
management bear detection rates pooled for sex as illustrated in Table 1. The estimation model
for simulations was the same as the generating model since the prime objective of simulations
was to evaluate relative precision as opposed to overall estimator performance. The Huggins
(Huggins 1991) closed population estimator in program MARK (White and Burnham 1999) was
used for estimation as well as simulations. Five hundred simulations were conducted in the
MARK simulation module for each population size and design combination.
The resulting population estimates were evaluated in terms of the average coefficient of variation
for pooled sex estimates which is simply the standard error divided by the population estimate
28
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
averaged across all simulations. In some cases the Huggins model did not converge resulting in
improbable (high estimates) and these results were deleted.
Sampling designs were considered optimal if an average CV of 10% was obtained in
simulations. Given the relative simplicity of simulations, the results were interpreted in terms of
relative gains in precision compared to a HS only design, rather than an absolute prediction of
estimate precision.
Results
Analysis of historic hair-snag DNA mark-recapture data sets.
I plotted the results of the analysis in (Proctor et al. 2010) to focus more on smaller population
sizes of bears (Figure 1). In addition, the results of the 5x5 km cells size/4 sampling session
projects were superimposed on the plot to assess how these results compared to analysis
estimates. It can be seen that when population size is 25, HS detection rates had to be at least 0.3
to obtain an estimate with a coefficient of variation (CV) of 20%. If population size was 60 then
it was marginally possible to obtain a precise estimate when HS detection rates were 0.25. HS
detection probabilities from the field projects ranged from 0.26-0.47 and in all but one case CV’s
were less than 0.2. The one case in which CV>0.2 occurred when the population size was 15
which was the reason for the low precision of this estimate. The higher detection rate of 0.47
was for the South Selkirk project that had a high degree of population closure and as a result, a
higher overall HS detection rate (Proctor et al. 2007, Proctor et al. 2010).
Coefficient of variation
0.40
0.35
0.30
0.25
0.20
0.3
0.15
0.10
0.31
0.05
0.47
0.26
0.00
10 20 30 40 50 60 70 80 90 100
Population size
HS detection p: . 0.10 0.15
0.25 0.30
0.20
0.40
Figure 1: Predicted relationship between population sizes, HS detection probabilities, and
coefficient of variations of population size estimates (Proctor et al. 2010)
29
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Simulation results
The average coefficient of variation for designs varied by design intensity and assumed
population size (Figure 2). The lowest levels of precision were for the HS only 4 session design.
In this case, CV’s were above 0.2 unless population size was greater than 30. The addition of 2
or 5 rub tree sessions resulted in CV’s of close to 0.1 at all population sizes. The HS only 5
session design also resulted in increased precision but CV was still above 0.1 unless population
size was 50 or greater. As with the 4 session design, precision was increased for the base 5
session HS design if RT and management bear data sources were added, however, the gain was
not as great as with the 4 session HS designs.
Figure 2: The mean coefficient of variation from simulations as a function of population size and
sampling design simulated.
If a criteria for the optimal sampling design is a mean CV of 0.1 then the HS4 and the HS5
design (when N<=40) , and HS4 RT2 MG1 (when N<=30) are not optimal. However, the HS4
RT2 MG1 design did display mean CV’s that were very close to 0.1 (Table 2).
Table 2: Mean coefficients of variation for each sampling design as a function of population size
simulated. Mean coefficients of variation of >10% are shaded suggesting that these designs are not
optimal.
Design
HS4
HS4 RT2
MG1
HS4 RT5
Population size
25
0.233
0.116
30
0.205
0.109
40
0.171
0.100
50
0.147
0.093
60
0.126
0.086
0.095
0.089
0.080
0.075
0.069
30
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
MG1
HS5
HS5 RT2
MG1
HS5 RT5
MG1
0.159
0.087
0.138
0.085
0.114
0.075
0.100
0.070
0.089
0.065
0.074
0.072
0.064
0.057
0.053
Another way to interpret simulation results is the proportion of simulations in which the CV was
less than 0.1. This interpretation better accounts for the variation in estimates generated in the
simulations. In this context, at least 50% of the simulations resulted in CV’s of less than 0.1 for
the designs judged to be optimal in Table 2. These results can be interpreted in terms of risk
management in that they give the relative probability of obtaining a CV of less than 0.1 in the
designs proposed.
Table 3: Proportion of simulations where the coefficient of variation was less than 0.1 as a function
of population size. Cells that are shaded grey correspond to Table 2 and simulations in which
<50% of the simulations had a CV of <0.1.
Design
HS4
HS4 RT2
MG1
HS4 RT5
MG1
HS5
HS5 RT2
MG1
HS5 RT5
MG1
Population size
25
2.0%
41.2%
30
3.2%
47.5%
40
6.4%
55.8%
50
11.2%
69.5%
60
18.6%
81.4%
64.8%
75.2%
86.3%
89.3%
95.4%
14.2%
75.6%
25.3%
78.4%
40.6%
91.5%
59.4%
95.2%
78.4%
97.8%
88.1%
89.5%
95.2%
99.6%
99.6%
Another way to consider simulation results is the risk of obtaining an imprecise (CV>0.2)
estimate with the designs proposed (Table 4). It can be seen that the 4 session HS only design
has a relatively high risk (large percentage of simulations with a CV>0.2) of an imprecise
estimate when population sizes are less than 40. All of the multiple data source designs showed
resulted in the probability of obtaining an estimate with CV<0.2 was 0.95 or greater.
Table 4: Proportion of simulations where the coefficient of variation was less than 0.2 as a function
of population size
Design
Population size
25
30
40
50
60
49.0%
61.2%
75.4%
88.6%
97.6%
HS4
95.1%
98.0%
99.6%
99.8%
99.8%
HS4 RT2 MG1
98.4%
99.2%
99.6%
100.0%
100.0%
HS4 RT5 MG1
82.0%
89.8%
97.0%
99.0%
99.6%
HS5
99.6%
99.6%
100.0%
100.0%
100.0%
HS5 RT2 MG1
99.8%
100.0%
99.8%
100.0%
100.0%
HS5 RT5 MG1
31
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Discussion
Results of empirical analyses and simulations suggest that additional sampling effort beyond the
traditional 4 session HS design is needed to ensure precise population estimates. The
simulations used for this study provide a way to evaluate the relative gain in precision of various
sampling designs in comparison to the baseline 4 session HS design. In this context, simulation
results suggest that the greatest gain in terms of precision is through the addition of multiple rub
tree sessions. The addition of an additional hair snag session will not guarantee a precise
estimate, especially if population of bears is lower. The HS4 RT2 MG1 design that utilizes 2 rub
tree sessions demonstrates reasonable precision unless the population size is below 30
individuals whereas the HS4 RT5 MG1 design has optimal performance across all population
sizes.
The simulation study did not try to address variance caused by undefined heterogeneity or model
selection uncertainty. Therefore estimates of precision may be optimistic. However, if it
assumed that the degree of heterogeneity is similar for each design, then results still should allow
valid comparison among different designs. Precision from simulated 4 session designs was
similar to that observed in the field and therefore it is likely that the levels of precision from
simulation “are in the ballpark” of those observed in the field.
One component not considered in this study is the relative cost of the various sampling designs.
Relative costs could be added to the simulation results to allow a better determination of optimal
designs. For example the ratio of mean CV from simulations to relative cost might allow a better
idea of the best gain per unit of field/financial effort. The allocation of sampling effort in terms
of sampling design has been developed further as discussed in Williams et al. (2002).
Should the hair snag sites be moved?
One question with the 5x5 fixed design is whether HS sites should be moved to ensure adequate
coverage and minimize potential for HS site habituation. I used a spatial simulation model to
explore the effect of moving sites on trap encounter and found that a cell size of 5x5 km was
needed to ensure the entire population was targeted by HS sampling. The simulation model and
estimation models in these simulations were simple and therefore these results should be
interpreted cautiously.
32
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Figure 3: Results of spatial simulation model investigation of moving sites versus not moving sites
and the effect of grid cell sizes (J. Boulanger, unpublished data)
Further empirical investigation of the effect of moving sites and not moving sites for a 7x7 km
cell size suggested slightly lower population size estimates when sites were not moved. Lower
estimates were presumably due to lower encounter rates of sites by females at the 7x7 km cell
size. These results would not necessarily apply to the smaller 5x5 km cell size. A slight
behavioural response was also detected for male bears when sites were not moved in this study
(Boulanger et al. 2006).
One potential way to allow better placement of sites due to seasonal difference (i.e. the snow
level is higher in later sessions and therefore moving sites to better habitat is possible in later
sessions) is to move all the sites after the 2nd session. This would allow better site placement and
mitigate any potential issues with behavioural response. It would make sense to move all sites
to ensure that the effect caused by moving sites would be a temporal change in detection rates
(that would affect all bears equally and is relatively easy to model) rather than heterogeneity (if
only some sites were moved therefore potentially causing only some of the bears detection rates
to increase).
Other ways to improve precision and estimate robustness.
The use of meta-analysis strategies that jointly model several data sets has been used to increase
precision for the other 5x5 km designs in British Columbia (Proctor et al. 2010) as well as other
project in British Columbia (Boulanger et al. 2002) and Alberta (Grizzly Bear Inventory Team
2008). This approach could be useful for the Cabinet Yak especially since the Highway 3 and
South Selkirk studies are in relatively close proximity to the Cabinet Yak area (Proctor et al.
2007). It is possible to use meta-analysis method to just model detection probabilities for HS
data even if rub tree and management bear data was not used for the other projects (by fixing RT
and MG detection probabilities to 0 for projects that did not use RT and MG data). The use of a
meta-analysis strategy along with multiple data sources should be strongly considered given that
lower sample sizes of detected bear would challenge the modeling of heterogeneity variation
using the HS data alone from just the Cabinet-Yak study.
33
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Literature cited
Boulanger, J., K. C. Kendall, J. B. Stetz, D. A. Roon, L. P. Waits, and D. Paetkau. 2008.
Multiple data sources improve DNA-based mark-recapture population estimates of
grizzly bears. Ecological Applications 18:577-589.
Boulanger, J., M. Proctor, S. Himmer, G. Stenhouse, D. Paetkau, and J. Cranston. 2006. An
empirical test of DNA mark-recapture sampling strategies for grizzly bears. Ursus
17:149-158.
Boulanger, J., G. C. White, B. N. McLellan, J. G. Woods, M. F. Proctor, and S. Himmer. 2002.
A meta-analysis of grizzly bear DNA mark-recapture projects in British Columbia. Ursus
13:137-152.
Gervasi, V., P. Ciucci, J. Boulanger, M. Posillico, C. Sulli, S. Focardi, E. Randi, and L. Boitani.
2008. A preliminary estimate of Apennine brown bear population size based on hair-snag
sampling and multiple data source mark-recapture Huggins models. Ursus 19:105-121.
Gervasi, V., P. Ciucci, F. Davoli, J. Boulanger, L. Boitani, and E. Randi. 2010. Addressing
challenges in non invasive capture recapture based estimates of small populations: a pilot
study on the Apennine brown bear. Conservation Genetics Online.
Grizzly Bear Inventory Team. 2008. Grizzly Bear Population and Density Estimates for Alberta
Bear Management Unit 6 and British Columbia Management Units 4-1, 4-2, and 4-23
(2007) Alberta Sustainable Resource Development, Fish and Wildlife Division, British
Columbia Ministry of Forests and Range, British Columbia Ministry of Environment, and
Parks Canada, Hinton Alberta <
Huggins, R. M. 1991. Some practical aspects of a conditional likelihood approach to capture
experiments. Biometrics 47:725-732.
Kendall, K. C., J. B. Stetz, J. Boulanger, A. C. Macleoud, D. Paetkau, and G. C. White. 2009.
Demography and genetic structure of a recovering grizzly bear population. Journal of
Wildlife Management 73:3-17.
Kendall, K. C., J. B. Stetz, D. A. Roon, L. P. Waits, J. Boulanger, and D. Paetkau. 2008. Grizzly
Bear Density in Glacier National Park, Montana. Journal of Wildlife Management
72:1693-1705.
Proctor, M., J. Boulanger, S. E. Nielsen, C. Serveen, W. F. Kasworm, T. Radandt, and D.
Paetkau. 2007. Abundance and density of Central Purcell, South Purcell, Yahk, and south
Selkirk Grizzly Bear Population Units in southeast British Columbia. BC Ministry of
Environment, Nelson and Victoria, British Columbia, Canada <
Proctor, M., B. N. McLellan, J. Boulanger, C. D. Apps, G. Stenhouse, D. Paetkau, and G.
Mowat. 2010. Ecological investigations of grizzly bears in Canada using DNA from hair:
1995-2005: a review of methods and progress. Ursus 21:169-188.
Stetz, J. B., K. C. Kendall, and C. Servheen. 2010. Evaluation of bear rub surveys to monitor
grizzly bear population trends. Journal of Wildlife Management 74:860-870.
White, G. C., and K. P. Burnham. 1999. Program MARK: Survival estimation from populations
of marked animals. Bird Study Supplement 46:120-138.
34
Cabinet Yaak Grizzly Bear DNA Project Study Plan: vJan 2012
Williams, B. K., J. D. Nichols, and M. J. Conroy. 2002. Analysis and management of animal
populations. Academic Press, San Diego.
35