ELK RESPONSES TO RECREATIONAL USE AND HABITAT POTENTIAL
IN MICHIGAN
By
Chad Ryan Williamson
A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Fisheries and Wildlife – Doctor of Philosophy
2021
ABSTRACT
ELK RESPONSES TO RECREATIONAL USE AND HABITAT POTENTIAL IN MICHIGAN
By
Chad Ryan Williamson
The growing use of public lands for nature-based recreation has prompted a demand for
research evaluating recreational use and its direct and indirect effects on wildlife populations and
their habitat. Although a growing body of research has reported numerous negative effects that
recreational use can have on wildlife resources, recent research has demonstrated that suitable
habitat may mitigate the effects of human-wildlife interactions. In Michigan, the Pigeon River
Country (PRC) and Atlanta (ASF) State Forests serve as the core range of Michigan’s elk
(Cervus elaphus nelsoni) herd. The PRC is a Special Management Unit that limits certain trail-
based recreation types (e.g., equestrian use, mountain biking) to designated trails and prohibits
some motorized vehicles (e.g., ORVs). Our primary goals were to examine the interactions
among elk space-use and resource selection patterns, habitat suitability and potential, and
summer trail-based recreation on public lands in the Michigan elk range. For our first objective,
we developed habitat suitability index (HSI) and habitat potential models for elk within the
Michigan elk range. Our HSI models indicated areas of high habitat suitability and potential for
winter thermal cover, winter food, and spring food throughout the elk range. For our second
objective, we quantified and compared the intensities and group sizes of common summer trail-
based recreation types (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) at
different temporal scales (i.e., year, month, day, hour) in the PRC and ASF. Recreation was
monitored using trail cameras and we captured 11,412 recreation events during 263,664 hours of
monitoring in the PRC, and 5,034 events during 266,184 hours in the ASF from May–October,
2016–2018. Greater recreational intensity was detected for all recreation types in both regions
during September, weekends, and mid-day (11:00–16:59). The most frequently detected types of
recreation were equestrian use (58.8% of events) in the PRC and ORV use (51.8% of events) in
the ASF. Our third objective was to evaluate and compare space-use and resource selection
patterns for Michigan elk in response to habitat suitability and the intensity of summer equestrian
use, hiking, mountain biking, and ORV use at different temporal periods. Global positioning
system (GPS) collars were placed on 27 cow and 26 bull elk from 2016–2018. Dynamic
Brownian bridge movement models were used to quantify elk space-use patterns, and elk
resource selection was modeled at landscape- and home range-scales. Elk home range sizes in
May were 1.3–2.0 times greater (P < 0.05) than in June–September. Weekends accounted for
36% of the greatest daily elk movement distances. Elk demonstrated changes in the proportional
use of cover types within home ranges during peak periods of recreational intensity. For our
fourth objective, we evaluated the behavioral responses of elk to experimental recreational
events. During September 2018, we monitored 69 equestrian use and 3 mountain biking events
using handheld GPS receivers. We evaluated elk responses to encounters with recreation events
that occurred within 2 times the average documented flight distance (60 m) for elk in Michigan.
We recorded 4 encounters with the same cow elk during our events, and found no responses or
changes in habitat use from encounters with recreation events. Our results highlight the need to
consider the varying effects of different types of recreation on wildlife populations and the
amount and quality of habitat components that may mitigate negative effects of interactions
between wildlife and recreational users. Achieving a balance of interactions among wildlife,
wildlife habitat, and recreational users is essential for ensuring long-term sustainability of
wildlife populations, habitat, and recreational opportunities on public lands.
ACKNOWLEDGEMENTS
This project was made possible through funding provided by the Federal Aid in Restoration
Act under Pittman-Robertson project W–147_R and partially supported by salary support for
Scott R. Winterstein (Project No. MICL02588) and for Henry Campa III (Project No.
MICL001646) from the USDA National Institute of Food and Agriculture. I extend my gratitude
to these organizations for their support.
First and foremost, I would like to thank my wife, Katie, and children, Marissa, Morgan,
Conor, Raya, and Camreigh for their unending love, support, encouragement, and patience
throughout the duration of this project. This dissertation is dedicated to you. I thank my brother,
Adam Williamson, for his support and encouragement during the many hours spent in
conversation over the phone throughout the duration of this project. I also thank my parents, Lori
and Don, and friends, Jason Doll, Harlan Eagan, Kevin Barnes, Scott Bergeson, and many others
for their support and encouragement. The years and months spent working on this project were
certainly made brighter and richer by all of you.
I thank my advisor, Dr. Henry (Rique) Campa III, for his support, guidance, patience, and
advice throughout this project, from which I have certainly become a better researcher, writer,
and wildlife professional. I also thank my committee members, Dr. Dean Beyer, Dr. Scott
Winterstein, Dr. Shawn Riley, Dr. Chuck Nelson and project collaborators, Dr. Alexandra
Locher, and Jeff Doser for their support, guidance, expertise, and contributions during this
project. I also thank my former advisor, Dr. Tim Carter, for his support and mentorship to a
former student throughout the duration of this project.
iv
I thank the many MDNR personnel, including Mark Monroe, Brian Mastenbrook, Shelby
Adams, Scott Whitcomb, Brad Johnson, Brian Roell, Erin Largent, Cody Norton, Cody Stevens,
Dan O’Brien, Jennifer Kleitch, Kevin Jacobs, Paige Perry and many others, that assisted with elk
captures and various field work that greatly increased the productivity of my research efforts. I
also thank the personnel from Wildlife Helicopter Services for their cooperation and collegiality
during the elk captures. I thank my technicians Jarod Reibel, Waldemar Ortiz-Calo, and Jon
Schafer for their friendship and assistance in field data collection. I thank Carrie Hirsch for the
many hours spent examining trail camera images. I also thank Barb Curtis, Bonnie Cornelius,
Darlene Alexander, Kerry Mase, Jeffrey Whiting, Andrew Knapp and many other volunteers
who generously shared their time and valuable insights on equestrian use and mountain biking in
the elk range.
v
TABLE OF CONTENTS
LIST OF TABLES ......................................................................................................................... ix
LIST OF FIGURES ..................................................................................................................... xiii
LIST OF SCHEMES.................................................................................................................... xvi
PREFACE ................................................................................................................................... xvii
GENERAL INTRODUCTION .......................................................................................................1
OBJECTIVES ..................................................................................................................................2
STUDY AREA DESCRIPTION .....................................................................................................3
Elk Herd and Habitat Management............................................................................................7
Management of Recreational Use ..............................................................................................8
CHAPTER 1: APPLICATIONS OF INTEGRATING ELK HABITAT SUITABILITY AND
HABITAT POTENTIAL MODELS................................................................................................9
INTRODUCTION ...........................................................................................................................9
METHODS ....................................................................................................................................13
Elk Habitat Suitability – Public Lands.....................................................................................15
Elk Habitat Suitability – Private Lands ...................................................................................20
Elk Habitat Potential ................................................................................................................22
RESULTS ......................................................................................................................................24
Elk Habitat Suitability – Public Lands.....................................................................................24
Elk Habitat Suitability – Private Lands ...................................................................................25
Elk Habitat Potential ................................................................................................................29
DISCUSSION ................................................................................................................................34
CHAPTER 2: EXAMINING TRAIL-BASED RECREATIONAL USE PATTERNS IN TWO
CONTIGUOUS STATE FORESTS WITH DIFFERENT USE REGULATIONS ......................41
INTRODUCTION .........................................................................................................................41
METHODS ....................................................................................................................................45
Trail Camera Placement Protocols and Camera Settings ........................................................45
PRC Description and Trail Camera Placement ........................................................................46
ASF Description and Trail Camera Placement ........................................................................51
Trail Camera Data Collection and Analyses ............................................................................53
RESULTS ......................................................................................................................................57
Relative Intensities, Temporal Characteristics, and Group Sizes of Recreational Users in the
PRC ..........................................................................................................................................57
Evaluation of recreational intensity in the PRC .................................................................57
Evaluation of recreational intensity by year and month in the PRC ..................................57
Evaluation of recreational intensity by day of week in the PRC .......................................64
vi
Evaluation of recreational intensity by time of day in the PRC ........................................64
Evaluation of recreational intensity during peak days of use in the PRC ..........................65
Characteristics of group size for recreation types in the PRC ...........................................66
Relative Intensities, Temporal Characteristics, and Group Sizes of Recreational Users in the
ASF ..........................................................................................................................................77
Evaluation of recreational intensity in the ASF .................................................................77
Evaluation of recreational intensity by year and month in the ASF ..................................77
Evaluation of recreational intensity by day of week in the ASF .......................................78
Evaluation of recreational intensity by time of day in the ASF.........................................79
Evaluation of recreational intensity during peak days of use in the ASF ..........................80
Characteristics of group size for recreation types in the ASF ...........................................86
Comparisons between the PRC and ASF.................................................................................88
DISCUSSION ................................................................................................................................90
CHAPTER 3: ELK SPACE-USE AND RESOURCE SELECTION PATTERNS IN RESPONSE
TO SUMMER TRAIL-BASED RECREATION ..........................................................................97
INTRODUCTION .........................................................................................................................97
METHODS ..................................................................................................................................100
Elk Capture ............................................................................................................................100
Elk Locations .........................................................................................................................100
Elk Movement and Behavior .................................................................................................101
Elk Resource Use ...................................................................................................................102
Landscape-scale resource use ..........................................................................................103
Home range-scale resource use ........................................................................................104
Influence of Habitat Suitability and Recreational use of Roads and Trails ...........................105
Elk use of suitable areas ...................................................................................................105
Elk use of areas near roads and trails ..............................................................................106
RESULTS ....................................................................................................................................109
Elk Capture ............................................................................................................................109
Elk Locations .........................................................................................................................109
Elk Movements and Behavior ................................................................................................110
Elk Resource Use ...................................................................................................................116
Landscape-scale resource use ..........................................................................................116
Home range-scale resource use ........................................................................................119
Influence of Habitat Suitability and Recreational use of Roads and Trails ...........................125
Elk use of suitable areas ...................................................................................................125
Elk use of areas near roads and trails ..............................................................................125
DISCUSSION ..............................................................................................................................129
CHAPTER 4: ELK RESPONSES TO EXPERIMENTAL EQUESTRIAN USE AND
MOUNTAIN BIKING EVENTS ON PUBLIC LANDS IN MICHIGAN ................................137
INTRODUCTION .......................................................................................................................137
METHODS ..................................................................................................................................140
Experimental Recreation Events ............................................................................................140
Elk Movements and Behavior ................................................................................................140
RESULTS ....................................................................................................................................142
Experimental Recreation Events ............................................................................................142
vii
Elk Movements and Behavior ................................................................................................142
DISCUSSION ..............................................................................................................................152
CONCLUSIONS AND MANAGEMENT IMPLICATIONS .....................................................156
APPENDICES .............................................................................................................................161
APPENDIX A: Elk Collaring and Capture Data ..................................................................162
APPENDIX B: Elk Collar Events History ............................................................................165
APPENDIX C: Experimental Recreation Use Events ..........................................................168
APPENDIX D: Metadata ......................................................................................................172
APPENDIX E: Outreach and Presentation Experience ........................................................175
LITERATURE CITED ................................................................................................................178
viii
LIST OF TABLES
Table 1.1. Habitat suitability values (scale 0–1; 1 = optimum) for cover types supporting each
life requisite (i.e., winter thermal cover, winter food, and spring food) for elk in northeastern
lower Michigan on state-owned public lands. Cover types selected based on MDNR forest
inventory data. Adapted from Beyer (1987). .................................................................................16
Table 1.2. Elk habitat suitability value modifiers for elk life requisites in northeastern lower
Michigan on state-owned public lands. Adapted from Beyer (1987) ............................................16
Table 1.3. Mean maximum daily movement diameter (MDMD) of GPS-collared elk during
winter (17 Feb–20 Mar, 2016; 21 Dec–20 Mar, 2017–2018) and spring (21 Mar–20 Jun, 2016–
2018) in northeastern lower Michigan. ..........................................................................................21
Table 1.4. Habitat suitability values (scale 0–1; 1 = optimum) for cover types supporting each
life requisite (i.e., winter thermal cover, winter food, and spring food) for elk in northeastern
lower Michigan on private lands. Cover types selected based on satellite imagery classification.
Adapted from Beyer (1987). ..........................................................................................................21
Table 1.5. Accuracy assessment of classification methods used on National Agriculture
Inventory Program (NAIP) imagery (1 x 1 m, 2012) to classify cover types (i.e., aspen,
hardwoods, upland conifers, lowland conifers, openings) used in an elk habitat suitability model
for private lands in northeastern lower Michigan. Reported accuracy percentages were calculated
by validation of classified cover types (30 x 30 m) found within public lands (664.9 km2) using
Michigan Department of Natural Resources forest inventory cover type polygons.. ....................27
Table 2.1. Trail-based recreational use regulations for primary summer-fall recreation types for
the Pigeon River Country (PRC) State Forest and Atlanta State Forest (ASF) Management Units
in the northern lower peninsula of Michigan.. ...............................................................................51
Table 2.2. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., RRI, MN6HR) captured by remote digital trail cameras on public lands within the
Pigeon River Country (PRC) State Forest and Atlanta State Forest (ASF) Management Units
during peak summer–fall recreational periods (i.e., May–October), 2016–2018.. ........................59
Table 2.3. Independent variable F-values produced by generalized linear mixed models with
Poisson distributions to determine which variables had the most effect on detection of a
recreation event within a 6-hour period in the Michigan elk range (i.e., Atlanta [ASF] and Pigeon
River Country [PRC] State Forest management units) during 2016–2018.. .................................60
ix
Table 2.4. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5-minute period) and recreational
intensity (i.e., RRI, MN6HR) for recreation types (i.e., equestrian use, hiking/foot-traffic,
mountain biking, ORV use) captured by remote digital trail cameras on public lands within the
Pigeon River Country State Forest management unit during peak summer–fall recreational
periods (i.e., May–October), 2016–2018.. .....................................................................................61
Table 2.5. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5-minute period) and recreational
intensity (i.e., RRI, MN6HR) by days of the week captured by remote digital trail cameras on
public lands within the Pigeon River Country (PRC) and Atlanta (ASF) State Forest
Management Units during peak summer–fall recreational periods (i.e., May–October), 2016–
2018................................................................................................................................................68
Table 2.6. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., RRI, MN6HR) by days of the week for recreation types (i.e., equestrian use,
hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public
lands within the Pigeon River Country State Forest management unit during peak summer–fall
recreational periods (i.e., May–October), 2016–2018.. .................................................................71
Table 2.7. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., MN6HR) by time of day (i.e., 3, 6-hour intervals within 5:00–23:00) captured by
remote digital trail cameras on public lands within the Pigeon River Country (PRC) State Forest
and Atlanta State Forest (ASF) Management Units during peak summer–fall recreational periods
(i.e., May–October), 2016–2018.. ..................................................................................................73
Table 2.8. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5-minute period) and recreational
intensity (i.e., MN6HR) by time of day for recreation types (i.e., equestrian use, hiking/foot-
traffic, mountain biking, ORV use) captured by remote digital trail cameras on public lands
within the Pigeon River Country State Forest management unit during peak summer–fall
recreational periods (i.e., May–October), 2016–2018.. .................................................................75
Table 2.9. Mean group sizes of trail-based recreation events (i.e., any number of individuals of
the same recreation type passing by a camera in the same direction within a 5 minute period) for
recreation types (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) captured by
remote digital trail cameras on public lands in the Pigeon River Country State Forest
management unit during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Trail cameras operated from 20–May to 31–October, 2016; 16–May to 28–October, 2017; 24–
May to 30–September, 2018.. ........................................................................................................76
x
Table 2.10. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5-minute period) and recreational
intensity (i.e., RRI, MN6HR) for recreation types (i.e., equestrian use, hiking/foot-traffic,
mountain biking, ORV use) captured by remote digital trail cameras on public lands within the
Atlanta State Forest management unit during peak summer–fall recreational periods (i.e., May–
October), 2016–2018.. ...................................................................................................................81
Table 2.11. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., RRI, MN6HR) by days of the week for recreation types (i.e., equestrian use,
hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public
lands within the Atlanta State Forest management unit during peak summer–fall recreational
periods (i.e., May–October), 2016–2018.. .....................................................................................83
Table 2.12. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., MN6HR) by time of day (i.e., 3, 6-hour intervals within 5:00–23:00) for recreation
types (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) captured by remote
digital trail cameras on public lands within the Atlanta State Forest management unit during peak
summer–fall recreational periods (i.e., May–October), 2016–2018. .............................................85
Table 2.13. Mean group sizes of trail-based recreation events (i.e., any number of individuals of
the same recreation type passing by a camera in the same direction within a 5 minute period) for
recreation types (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) captured by
remote digital trail cameras on public lands in the Atlanta State Forest management unit during
peak summer–fall recreational periods (i.e., May–October), 2016–2018. Trail cameras operated
from 20–May to 31–October, 2016; 16–May to 28–October, 2017; 24–May to 30–September,
2018................................................................................................................................................87
Table 3.1. Locations of GPS-collared elk on public (i.e., Atlanta State Forest [ASF], Pigeon
River Country State Forest [PRC]) and private lands inside and outside of the Michigan
Department of Natural Resources designated elk range in northern lower Michigan from 1–May
to 30–September, 2016–2018. .....................................................................................................112
Table 3.2. Mean summer (i.e., 1–May to 30–September) elk home ranges (i.e., 95% utilization
distribution [UD]) and core areas (i.e., 50% UD) as estimated by dynamic Brownian bridge
movement models in Atlanta State Forest (ASF) and Pigeon River Country (PRC) State Forest in
the northern lower peninsula of Michigan, 2016–2018. ..............................................................112
Table 3.3. Mean monthly elk home ranges (i.e., 95% utilization distribution [UD]) and core areas
(i.e., 50% UD) as estimated by dynamic Brownian bridge movement models in Atlanta State
Forest (ASF) and Pigeon River Country (PRC) State Forest in the northern lower peninsula of
Michigan during summer 2016–2018. .........................................................................................113
Table 3.4. Average median elk daily linear movement distances (km) in Atlanta State Forest
(ASF) and Pigeon River Country (PRC) State Forest in the northern lower peninsula of
Michigan, 2016–2018. .................................................................................................................114
xi
Table 3.5. Mean proportions of cover types and elk locations found within GPS-collared elk
home ranges (i.e., 95% utilization distribution [UD]) and core areas (i.e., 50% UD) in the Atlanta
State Forest (ASF) and Pigeon River Country (PRC) State Forest in the northern lower peninsula
of Michigan from May–September, 2016–2018 ..........................................................................121
Table 3.6. Distances of key cover types (i.e., regenerating aspen, openings, northern
hardwoods/maple) and elk locations to the nearest road and trail (i.e., multi-use trail [trail],
biking trail [bike], equestrian trail [horse]) in the Atlanta State Forest (ASF) and Pigeon River
Country (PRC) State Forest in the northern lower peninsula of Michigan. Elk locations were
recorded from GPS-collared elk (n=53, 25 PRC [13 cows, 12 bulls], 28 ASF [14 cows, 14 bulls])
from 1–May to 30–September, 2016–2018 .................................................................................127
Table 3.7. Distances of primary cover types (i.e., regenerating aspen, openings, northern
hardwoods/maple) within elk home ranges (i.e., 95% UD) and elk locations within core areas
(i.e., 50% UD) to the nearest road and trail (i.e., multi-use trail [trail], biking trail [bike],
equestrian trail [horse]) in the Atlanta State Forest (ASF) and Pigeon River Country (PRC) State
Forest in the northern lower peninsula of Michigan. Elk locations were recorded from GPS-
collared elk (n=53, 25 PRC [13 F, 12 M], 28 ASF [14 F, 14 M]) from 1–May to 30–September,
2016–2018....................................................................................................................................128
Table 4.1. Summary of experimental equestrian use and mountain biking events within the
Pigeon River Country (PRC) and Atlanta State Forests (ASF) from 31–August to 30–September,
2018..............................................................................................................................................145
Table 4.2. Encounters (i.e., any 5-minute period where elk were within 120 m of an equestrian
user or mountain biker carrying a handheld GPS receiver) between equestrian users and elk
during experimental recreation events within the Pigeon River Country State Forest from 31–
August to 30–September, 2018 ....................................................................................................145
Table A1. Elk collaring and capture data from 3 capture events in the Pigeon River Country
(PRC) and Atlanta State Forests (ASF) from 15–16 February, 2016, 9–April, 2017, and 22–
February, 2018 .............................................................................................................................163
Table B1. Elk collar events history (i.e., collar deployments, elk mortalities, collar failures, collar
retrievals) in northern lower Michigan from 15–February, 2016 to 2–February, 2020 ..............166
Table C1. Experimental equestrian use and mountain biking events monitored with handheld
GPS receivers in the Pigeon River Country (PRC) and Atlanta State Forests (ASF) from 31–
August to 30–September, 2018 ....................................................................................................169
Table E1. Professional conferences, meetings, and events attended to promote awareness and
community engagement with my project focusing on elk responses to habitat potential and
human recreation use in the Michigan elk range, from 2016–2020 ............................................177
xii
LIST OF FIGURES
Figure 1.1. Location of the 1,220 km2 Michigan Department of Natural Resources (MDNR)
designated elk range and study area in the northern lower peninsula of Michigan. The Pigeon
River Country State Forest (458.4 km2) and a portion of the Atlanta State Forest (168 km2) are
considered the core of the Michigan elk range ................................................................................6
Figure 1.2. Distribution of elk winter thermal cover habitat suitability (A), winter food habitat
suitability (B), and spring food habitat suitability (C) on public lands (665 km2) within the
MDNR defined elk range (1,220 km2) in the northern lower peninsula of Michigan. Values are
based on a scale of 0–1; 1 = highest suitability .............................................................................26
Figure 1.3. Distribution of elk winter thermal cover habitat suitability (A), winter food habitat
suitability (B), and spring food habitat suitability (C) within the MDNR defined elk range (1,220
km2) in the northern lower peninsula of Michigan. Values are based on a scale of 0–1; 1 =
highest suitability ...........................................................................................................................28
Figure 1.4. Proportion (Prop.) and location of habitat types within the MDNR defined elk range
(1,220 km2) in the northern lower peninsula of Michigan. Vegetation codes for early-, middle-,
and late-successional stages are as follows: A = aspen, BA = Black Ash, Bas = basswood, Bee =
beech, BF = balsam fir, Bir = white birch, BP = balsam poplar, BC = black cherry, BS = black
spruce, C = cedar, G = grass species, H = hemlock, JP = jack pine, LBr = lowland brush, O = oak
species, RM = red maple, RO = red oak, RP = red pine, Shr = shrub, SM = sugar maple, T =
tamarack, WA = white ash, WP = white pine ................................................................................30
Figure 1.5. Distribution of elk winter thermal cover habitat potential (A), winter food habitat
potential (B), and spring food habitat potential (C) in the MDNR defined elk range (1,220 km2)
in the northern lower peninsula of Michigan. Distribution of areas with low habitat suitability (<
0.33) and high habitat potential (> 0.66) for elk winter thermal cover (D), winter food (E), and
spring food (F). Habitat potential/suitability values are based on a scale of 0–1; 1 = highest
potential/suitability ........................................................................................................................31
Figure 2.1. Location of designated recreational trails and campgrounds in the 1,220 km2
Michigan Department of Natural Resources designated elk range and study area in the northern
lower peninsula of Michigan. Trail camera locations shown (n=78) were during 24 May–30
September, 2018 ............................................................................................................................47
Figure 2.2. Mean relative recreational intensity (i.e. RRI = number of recreation events per
camera hour) of equestrian use, hiking/foot-traffic, mountain biking, and ORV use during peak
summer–fall recreational periods (i.e., May–October), 2016–2018. Data were captured by remote
digital trail cameras in the northern lower peninsula of Michigan: (A) Pigeon River Country
State Forest management unit, (B) Atlanta State Forest management unit. Variation depicted by
error bars is SD of mean ................................................................................................................63
xiii
Figure 2.3. Mean relative recreational intensity (i.e. RRI = number of recreation events per
camera hour) of equestrian use, hiking/foot-traffic, mountain biking, and ORV use by day of
week during peak summer–fall recreational periods (i.e., May–October), 2016–2018. Data were
captured by remote digital trail cameras in the northern lower peninsula of Michigan: (A) Pigeon
River Country state forest management unit, (B) Atlanta state forest management unit. Variation
depicted by error bars is SD of mean .............................................................................................70
Figure 2.4. Human recreation use (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV
use) events (i.e., any number of individuals of the same recreation type passing by a camera in
the same direction within a 5 minute period) by hourly time intervals during peak summer–fall
recreational periods (i.e., May–October), 2016–2018. Data were captured by remote digital trail
cameras in the northern lower peninsula of Michigan: (A) Pigeon River Country State Forest
management unit, (B) Atlanta State Forest management unit .......................................................74
Figure 3.1. Elk capture and GPS-collaring locations and recreational trails and campgrounds
within the Michigan Department of Natural Resources designated elk range in the northern lower
peninsula of Michigan. Elk capture and collaring was during 15–16 February, 2016 (n=40, 18
PRC, 22 ASF), 9 April, 2017 (n=1, ASF), and 22 February, 2018 (n=12, 7 PRC, 5 ASF).........108
2
Figure 3.2. Average median hourly Brownian motion variance (𝜎𝑚 ) for GPS-collared elk in the
Atlanta State Forest (A = cows [n = 13], B = bulls [n = 14]) and Pigeon River Country (C = cows
[n = 13], D = bulls [n = 11]) State Forest in the northern lower peninsula of Michigan from May–
September, 2016–2018. Gray-filled areas represent mean times of day between sunset and
sunrise during our monitoring periods .........................................................................................115
Figure 3.3. Relative probability of use of cover types by GPS-collared elk in the Atlanta State
Forest (A = cows [n = 13], B = bulls [n = 14]) and Pigeon River Country (C = cows [n = 13], D =
bulls [n = 11]) State Forest in the northern lower peninsula of Michigan from May–September,
2016–2018....................................................................................................................................118
Figure 3.4. Proportional monthly core area (i.e., 50% utilization distribution [UD]) use of cover
types within elk home ranges (i.e., 95% UD) in the Atlanta State Forest (A = cows [n = 13], B =
bulls [n = 14]) and Pigeon River Country (C = cows [n = 13], D = bulls [n = 11]) State Forest in
the northern lower peninsula of Michigan from May–September, 2016–2018. Dashed horizontal
lines represent proportional availability of each cover type within elk 95% UDs. Oak and
lowland conifer cover types were omitted due to low proportional use during the sampling period
......................................................................................................................................................122
Figure 3.5. Proportional daily core area (i.e., 50% utilization distribution [UD]) use of cover
types within elk home ranges (i.e., 95% UD) in the Atlanta State Forest (A = females, B = males)
and Pigeon River Country (C = females, D = males) State Forest in the northern lower peninsula
of Michigan from May–September, 2016–2018. Dashed horizontal lines represent proportional
availability of each cover type within elk 95% UDs ...................................................................123
xiv
Figure 3.6. Proportional hourly core area (i.e., 50% utilization distribution [UD]) use of cover
types within elk home ranges (i.e., 95% UD) in the Atlanta State Forest (A = cows [n = 13], B =
bulls [n = 14]) and Pigeon River Country (C = cows [n = 13], D = bulls [n = 11]) State Forest in
the northern lower peninsula of Michigan from May–September, 2016–2018. Dashed horizontal
lines represent proportional availability of each cover type within elk 95% UDs. Oak and
lowland conifer cover types were omitted due to low proportional use during the sampling
period. Gray-filled areas represent mean times of day between sunset and sunrise during our
monitoring periods .......................................................................................................................124
Figure 4.1. Location of experimental equestrian use (n = 69) and mountain biking (n = 3) events
within the Pigeon River Country and Atlanta State Forests from 31–August to 30–September,
2018. User locations were recorded in 1-minute intervals using handheld Global Positioning
System receivers carried during recreation events.......................................................................144
Figure 4.2. Encounter between equestrian users and a cow elk (collar ID = 29868) during an
experimental recreation event within the Pigeon River Country State Forest on 31–August, 2018.
Cow elk locations shown occurred in 30-minute intervals from 10:00 to 13:00 .........................146
Figure 4.3. Encounter between equestrian users and a cow elk (collar ID = 29868) during an
experimental recreation event within the Pigeon River Country State Forest on 2–September,
2018. Cow elk locations shown occurred in 30-minute intervals from 10:00 to 13:00 ...............147
Figure 4.4. Encounter between equestrian users and a cow elk (collar ID = 29868) during an
experimental recreation event within the Pigeon River Country State Forest on 27–September,
2018. Cow elk locations shown occurred in 30-minute intervals from 07:00 to 10:00 ...............148
Figure 4.5. Encounter between equestrian users and a cow elk (collar ID = 29868) during an
experimental recreation event within the Pigeon River Country State Forest on 27–September,
2018. Cow elk locations shown occurred in 5-minute intervals from 30 minutes before to 30
minutes after the closest linear distance (i.e., 92 m at 17:40) between the equestrian users and the
cow ...............................................................................................................................................149
Figure 4.6. Locations of GPS-collared elk and elk capture locations in relation to experimental
equestrian use (n = 69) and mountain biking (n = 3) events within the Pigeon River Country and
Atlanta State Forests from 31–August to 30–September, 2018 ...................................................150
Figure 4.7. Locations of GPS-collared elk in relation to habitat suitability and experimental
equestrian use (n = 69) and mountain biking (n = 3) events within the Pigeon River Country and
Atlanta State Forests from 31–August to 30–September, 2018 ...................................................151
xv
LIST OF SCHEMES
Scheme 1.1. Schematic of model components and processes used to develop habitat suitability
and habitat potential models for public and private lands within the elk range in northeastern
lower Michigan ..............................................................................................................................14
xvi
PREFACE
This dissertation has been organized into a General Introduction, Study Objectives, Study
Area Description, 4 chapters that focus on our 4 primary study objectives, and 5 appendices. The
formatting of this dissertation follows the formatting guidelines for manuscripts submitted to the
Journal of Wildlife Management. Chapter 1 was submitted to the Wildlife Society Bulletin on 20–
October, 2019, accepted on 9–June, 2020, and published on 18–February, 2021. The co-authors
for Chapter 1 were Henry Campa III, Alexandra B. Locher, Scott R. Winterstein, and Dean E.
Beyer Jr. Chapter 2 was submitted to the Journal of Outdoor Recreation and Tourism on 16–
September, 2020, and a revision that was submitted on 28–February, 2021 is in review. The co-
authors for Chapter 2 were Henry Campa III, Scott R. Winterstein, Charles M. Nelson, and Dean
E. Beyer Jr. Chapter 3 was submitted to the Journal of Wildlife Management on 6–July, 2021
and is in review. The co-authors for Chapter 3 were Henry Campa III, Jeffrey W. Doser, Scott R.
Winterstein, and Dean E. Beyer Jr. Chapter 4 is being modified for submission to the Journal of
Human-Wildlife Interactions in fall 2021. The co-authors for Chapter 4 were Henry Campa III
and Dean E. Beyer Jr. Appendix A provides elk collaring and capture data during our study.
Appendix B provides a history of elk collar deployments, elk mortalities, collar failures, and
collar retrievals. Appendix C provides a detailed summary of experimental recreation events
from Chapter 4. Appendix D contains metadata describing the organization and storage locations
of all data associated with this project. Appendix E contains a description of outreach experience
and a list of presentations given to numerous committees and at professional conferences.
xvii
GENERAL INTRODUCTION
Trail-based recreation (e.g., equestrian use, mountain biking, ORV use) has increased on
public lands within the Michigan elk range over the last 50 years with noticeable growth in
visitor numbers and interest in summer trail-based recreational opportunities (e.g., equestrian
use, mountain biking, ORV use; MDNR 2007, MDNR 2012). Although Michigan provides
32,375 km2 of public lands for recreation, the presence of a visible elk (Cervus elaphus nelsoni)
herd in the Pigeon River Country State Forest (PRC) and part of the adjoining Atlanta State
Forest (ASF), makes these areas an attractive destination (MDNR 2018, Hunt 2019). In the last
10 years, increased reports of elk causing agricultural depredation outside of their core range has
raised concerns among natural resource managers over the potential impacts of recreational
activities on elk movements and behaviors (B. Mastenbrook, MDNR, personal communications).
Consequently, natural resources managers have hypothesized that elk may be selecting areas
with less recreational activity. Long-term effects of repeated interactions between wildlife and
recreational users, such as avoidance of frequently used areas, create challenges for managers
attempting to provide wildlife habitat components. Wildlife avoidance of areas used by
recreational users can lead to indirect habitat degradation and has been observed in large
herbivores such as red deer (Cervus elaphus), mountain caribou (Rangifer tarandus caribou),
and elk (Sibbald et al. 2011, Lesmerises et al. 2018, Wisdom et al. 2018). However, other
research has demonstrated that providing suitable habitat may mitigate the negative effects of
human-wildlife interactions on wildlife populations (Coppes et al. 2018). To determine the
effects of trail-based recreation and habitat suitability and potential on the elk population in
northern Michigan, the MDNR and Michigan State University partnered to conduct research
examining the interactions among elk, their habitat, and summer trail-based recreational activity.
1
OBJECTIVES
The objectives of this study were focused on providing natural resources managers with an
understanding of the potential effects of interactions among the elk population, their habitat, and
summer trail-based recreational users in the Michigan elk range. The objectives were to:
1) Quantify and compare current elk habitat suitability and habitat potential for public and
private lands in the Michigan elk range.
2) Quantify and compare the number, relative intensity, frequency, and geographic scope of
summer trail-based recreational users (i.e., equestrian users, hikers, mountain bikers, off-road
vehicle users) between the Pigeon River Country State Forest (with restricted recreational
activities) and portions of the Atlanta State Forest (with limited recreation restrictions) within the
Michigan elk range.
3) Quantify and compare elk space-use and habitat selection patterns in response to habitat
suitability and the intensity of summer trail-based recreation types (i.e., equestrian use, hiking,
mountain biking, off-road vehicle use) at different temporal periods for the Pigeon River Country
State Forest and portions of the Atlanta State Forest within the Michigan elk range.
4) Quantify and compare the fine scale responses of GPS-collared elk to experimental equestrian
use and mountain biking events within the Pigeon River Country State Forest and portions of the
Atlanta State Forest within the Michigan elk range.
2
STUDY AREA DESCRIPTION
Our study area was the 122,000 ha (66,500 ha of public lands, 55,500 ha of private lands)
MDNR defined elk range and surrounding public and private lands within Cheboygan,
Montmorency, Otsego, and Presque Isle counties in the northern portion of the lower peninsula
of Michigan (Figure 1.1). The region has a humid continental climate with a mean annual
temperature of 5.6° C and a mean annual precipitation of 77.9 cm (Michigan Weather Service
1974, NOAA 2016). Temperature extremes can reach 34.4° C in summers and –28.8° C in
winters, with a mean snowfall of 198.6 cm (NOAA 2016).
The topography consists of moderately sloped ground moraines and outwash plains in the
south connecting to steep moraine ridges among outwash plains in the north (Albert 1995). The
southern ground moraines are dominated by drumlin fields that are typically <20 m in elevation,
0.2–0.4 km in width, and 1.6 km in length, separated by poorly drained outwash plains. Steep
moraine ridges in the north are surrounded by broad well-drained outwash plains and narrow
poorly drained outwash channels, and are characterized by elevation changes of >60 m in
distances <1.6 km and are among the steepest in the lower peninsula of Michigan (Albert 1995).
Elevation in the study area ranges from 181–335 m in the south to 274–396 m in the north. The
Pigeon and Black Rivers originate and drain within the elk range and are prominent within the
study area. Soil types range from dry sandy soils with low fertility in outwash plains to sandy
loam soils with medium–high fertility in till plains and moraines (USDA 2019).
Vegetation types within the study area vary depending on soil types, drainage, fertility,
exposure, and land management practices. Prior to the 1940’s, extensive logging, repeated
burning, and scattered attempts at farming further influenced the formation of numerous
vegetation types within the study area (Moran 1973, MDNR 2007). Vegetation types within the
3
elk range have been classified into 6 physiographic distributions, namely morainic uplands, steep
morainic slopes, outwash plains – morainic ecotones, sandy outwash plains, riverbanks and
bottomlands, and coniferous swamps (Spiegel et al. 1963, Moran 1973). Morainic uplands
primarily support northern hardwood forest types dominated by sugar maple (Acer saccharum),
basswood (Tilia americana), hemlock (Tsuga canadensis), red maple (A. rubrum), and beech
(Fagus grandifolia; Albert 1995). Steep morainic slopes are characterized by aspen (Populus
spp.) with red maple occurring at the base of slopes and pine (Pinus spp.)-hardwood mixtures
occurring at mid–high elevations (Albert 1995). Outwash plains – morainic ecotones are
considered transitional zones that are often dominated by red maple, white birch (Betula
papyrifera), and aspen (Moran 1973). Sandy outwash plains primarily support jack pine (Pinus
banksiana), grasses (Poaceae spp.), and forbs with interspersions of cherry (Prunus spp.),
willow (Salix spp.), and juneberry (Amelanchier Canadensis; Albert 1995). Riverbanks and
bottomland areas support lowland hardwoods and alluvial silt plain sites with common species
such as ash (Fraxinus spp.), grey alder (Alnus incana), dogwood (Cornus spp.), and willow
(Moran 1973). Coniferous swamps are dominated by northern white-cedar (Thuja occidentalis),
balsam fir (Abies balsamea), black spruce (Picea mariana), and balsam poplar (Populus
balsamifera; Albert 1995). Open areas occur throughout the study area as maintained wildlife
openings, natural grasslands (i.e., typically part of barren or savannah communities), and old
field grasslands (MDNR 2008).
Land use within our study area is influenced by accessibility to public lands for recreation,
production of timber products, and occurrence of agricultural crops and hunt clubs on private
lands. The study area is bordered by Federal and state highways with Interstate 75 to its west,
Michigan Route 68 to its north, Michigan Route 33 to its east, and Michigan Route 32 to its
4
south. Despite the absence of federal or state highways within the study area, county and
seasonal MDNR forest roads occur throughout the Pigeon River Country (PRC) and Atlanta
(ASF) State Forests. According to the MDNR Resource Assessment Unit, 88% of the PRC is
within 0.8 km of a road (S. Whitcomb, MDNR, personal communication). Approximately 16.2%
of private lands within the elk range are club lands (i.e., Black River Ranch [35.5 km2] and
Canada Creek Ranch [54.6 km2]) that offer hunting, fishing, and outdoor reaction opportunities
(Black River Ranch 2019, Canada Creek Ranch 2019). Primary types of public and private land
recreation include camping, hunting, fishing, mushroom hunting, berry picking, equestrian use,
and mountain biking (MDNR 2007, MDNR 2012).
Public lands within the Michigan elk range included the PRC (45,840 ha) and portions of the
ASF (16,800 ha) and Gaylord State Forest Management Units of the Mackinaw State Forest
(Figure 1.1). The PRC and portions of the ASF inside of the elk range are considered the core of
the elk range. Although both are state forests balancing multiple management objectives for
sustainability of forest and wildlife resources and recreational opportunities, the PRC has been
designated as a Special Management Unit since 1972 to safeguard its unique variety of
undeveloped land cover types and natural resources from overuse and development (MDNR
2007). Notably, the PRC’s Concept of Management (COM) describes its specific objectives and
guidelines for managing elk and other fish and wildlife species and their habitat, forest and
mineral resources, and for providing recreational opportunities (MDNR 2007).
5
Figure 1.1. Location of the 1,220 km2 Michigan Department of Natural Resources (MDNR) designated elk range and study area in the
northern lower peninsula of Michigan. The Pigeon River Country State Forest (458.4 km2) and a portion of the Atlanta State Forest
(168 km2) are considered the core of the Michigan elk range.
6
Elk Herd and Habitat Management
The first objective of the COM is to “manage the elk population and elk habitat so the Pigeon
River Country State Forest remains the nucleus of Michigan’s elk herd” (MDNR 2007:14). The
elk herd has persisted in the region since the introduction of 7 Rocky Mountain elk in 1918
(MDNR 2007). The creation of the Michigan Elk Management Plan by the MDNR (1975)
established elk as a priority species and outlined management objectives. The 1984 Elk
Management Plan designated the elk range, set a population objective of 600–800 elk,
established recreational hunting as the primary method of population management, and
recognized the importance of elk viewing opportunities (MDNR 1984). The 2012 Elk
Management Plan revised the population objective to 500–900 elk (MDNR 2012). Elk hunts
have occurred annually since 1984, and since 2006 the MDNR evaluates herd size using a
sightability model (MDNR 2012, Walsh 2007, Walsh et al. 2009). The estimated elk population
has remained relatively stable (800–1,400) since 1984, and from 2006–2019 the mean population
estimate was 1,065 (SD = 217.4; MDNR 2012, S. Adams, MDNR, personal communication).
Elk habitat is managed on public lands and based on goals defined by the Michigan Elk
Management Plan and the PRC’s COM (MDNR 2007, MDNR 2012). Beyer (1987) identified
the primary life requisites for elk in Michigan as spring food, winter food, and winter thermal
cover and thus the management of cover types supporting these requisites is essential for
sustaining the population. The goals for forest management include: 1) maintaining 6–7% of the
range as grass and upland brush; 2) managing aspen for no net loss, with a goal of aspen
representing 27% of the range; 3) maintaining mast production by oak and beech (Fagus
grandifolia), and increasing production if possible; and 4) sustaining mixed pine (Pinus spp.)
stands by promoting natural regeneration of coniferous and deciduous species. The goals for
7
openings include maintaining an even distribution of managed (i.e., planted, mowed, burned)
openings of at least 400 ha throughout the elk range. The PRC’s COM has guidelines for even-
aged management that retains <8% of stems or <0.93 m2 of basal area to be no greater than
approximately 16 ha, while the ASF has no size restrictions on even-aged management (MDNR
2007). For private lands, the MDNR communicates options and assists with improving elk
habitat if desirable by landowners (MDNR 2012).
Management of Recreational Use
The PRC has regulations for recreational use that restricts specific types of trail-based
recreation to a greater extent than the ASF and other state forests in Michigan (MDNR 2016,
MDNR 2018a). For example, equestrian users and mountain bikers may only use designated
trails and forest roads in the PRC, while being permitted anywhere within the ASF (MDNR
2016, M. Fry, M. Monroe, MDNR, personal communication). Off-road vehicle use is prohibited
in the PRC, however, permitted on designated trails and forest roads in the ASF (MDNR 2016,
MDNR 2018b). Despite the PRC’s unique recreational regulations, the forest offers a greater
quantity and variety of recreation-based amenities (e.g., designated campgrounds and trails) than
the ASF (Williamson et al. in review). Furthermore, the PRC is arguably more publicized and
accessible due to its history as the center of the elk range and close proximity to nearby
Interstate-75 and population centers (Williamson et al. in review).
8
CHAPTER 1: APPLICATIONS OF INTEGRATING ELK HABITAT SUITABILITY
AND HABITAT POTENTIAL MODELS
INTRODUCTION
Understanding wildlife-habitat relationships is fundamental for wildlife managers attempting
to develop habitat management strategies and predict population responses. Survival and
reproductive success of species are, in part, dependent on the amount, condition, and spatial
arrangement of habitat components (Van Horne and Wiens 2015). Therefore, it is vital for
managers to understand the ecosystem processes that influence the presence and distribution of a
species and identify areas that may or may not be suitable for a species’ life requisites. Using a
habitat-based perspective to examine the suitability of a landscape to support a population is a
common approach for predicting the spatial distribution of species (Hirzel and Le Lay 2008).
Habitat Suitability Index (HSI) models allow wildlife managers to assess the availability and
quality of habitat for a species and continue to be one of the most widespread management tools
used by government agencies (Brooks 1997, Latif et al. 2015). Habitat Suitability Index models
often rely on expert knowledge to define relevant habitat attributes to describe species’ life
requisites, and managers often use data from existing vegetation conditions or land-cover
databases to identify those attributes (Leblond et al. 2014). Van Horne and Wiens (1991:3)
suggested HSIs be viewed as “quantitative expressions of our best working understanding of the
relations between easily measured environmental variables and habitat quality for a species.”
While HSI models can provide spatial information about the quality and distribution of habitat
that is available, they are typically limited to current conditions and do not provide predictions of
future variations in quality, distribution, or availability of habitat (Thuiller and Münkemüller
2010).
9
For wildlife managers to make effective habitat management decisions, it is useful to have
information about the potential of areas to remain or become wildlife habitat for given species.
Wildlife habitat is not static, and the dynamic relationship between wildlife and their habitat is a
direct result of the processes by which landscapes and their associated vegetation types change
over time (Cushman and McGarigal 2007, Felix et al. 2007b). Therefore, natural resources
agencies attempting to model and manage populations should consider the arrangement of
species-specific habitat attributes and how they affect wildlife populations across space and time.
Previous research has focused on using “habitat types” (Daubenmire 1966) for predicting
changes in vegetation types through time and quantifying habitat potential (e.g., Felix et al. 2004,
Felix et al. 2007a, Windmuller-Campione et al. 2015). Felix (2004:796) defined “habitat
potential” as “the capability of an area being or becoming habitat based on biological and
geological characteristics.” Areas with the same ecological characteristics (e.g., soil
characteristics, landforms, climate) and successional trajectories are defined as habitat types
(Daubenmire 1966). Delineating habitat types and their boundaries allows managers to identify
ecologically similar land units where key information (i.e., measurements) can be extrapolated to
all areas of the same type (Kotar 1986). Additionally, identifying habitat types allows managers
to quantify habitat potential through the assignment of suitability values (SV) to successional
stages based on their ability to provide habitat components for wildlife. Understanding temporal
variations in habitat suitability through forest succession can provide insights on effects of
landscape changes on wildlife and their habitat, and how wildlife habitat may respond to land-
use decisions.
Consideration of current and future habitat availability and quality offers wildlife managers
insights on management strategies to maintain or provide additional habitat for species of
10
interest. For example, since the elimination of the eastern subspecies of elk (Cervus elaphus
canadensis) from Midwestern and eastern North America in the late 1800’s, wildlife managers
have attempted to manage populations of introduced Rocky Mountain elk (Cervus elaphus
nelsoni) with varying results (Witmer 1990, O’Gara and Dundas 2002, Keller et al. 2015).
Approximately 40% of elk restoration efforts have failed in eastern North America within 5–94
years (Popp et al. 2014). Notably, the most common explanation for elk restoration failure was
lack of appropriate habitat quality and/or quantity (Witmer 1990, Popp et al. 2014). Witmer
(1990) suggested the use of habitat suitability models or other methods to evaluate elk habitat in
areas where restorations are being proposed to increase likelihood of success. While some
regions examined potential restoration sites or herd expansion of current elk ranges (Van Deelen
1997, Telesco et al. 2007, Gilbert et al 2010), only current habitat availability and quality was
considered. Investigation of current habitat suitability and habitat potential in areas where
managers are attempting to establish new or maintain existing populations could provide insights
on the spatiotemporal dynamics of habitat availability and quality and likely increase probability
of successful restoration efforts and management plans. For example, a habitat potential model
can be used to identify habitat types that support aspen and hardwood-dominated vegetation
types (i.e., areas of high habitat suitability for elk winter and spring food) as potential restoration
sites or focus areas for maintaining or providing additional habitat for elk. Subsequently, an elk
habitat suitability model can be used to identify sites that are currently in low suitability within
those habitat types as focus areas for elk habitat management efforts.
We integrated HSI and habitat potential models for elk in Michigan to identify the spatial and
temporal dynamics of their habitat so that results could guide current and future habitat
management. In Michigan, one of the primary goals of the Michigan Department of Natural
11
Resources (MDNR) Elk Management Plan is to “manage for a sustainable elk population in
balance with the habitat” (MDNR 2012: 21). Since its conception, Michigan’s elk management
plan has outlined strategies to manage forests, openings, and private lands to maintain and
improve habitat for elk (MDNR 1975, MDNR 1984, MDNR 2012).
Beyer (1987) identified 3 potential habitat limiting factors for elk in Michigan, namely winter
thermal cover (WTC), winter food (WF), and spring food (SF). Using these limiting factors,
Beyer (1987) developed an HSI model to evaluate the quality of elk habitat throughout the year.
Natural resources managers can assess habitat quality to inform management decisions within
the elk range. However, determining existing vegetation conditions through land-cover databases
or field examination only provides an index of current conditions, requiring managers to make
assumptions about potential vegetation and successional dynamics of different habitat types
(Felix et al. 2004, Thuiller and Münkemüller 2010). Using models to examine habitat suitability
and potential can provide managers with insights on the spatiotemporal dynamics of wildlife
habitat for any species of interest with known habitat requirements in a region. Our objective was
to demonstrate how the development and integration of elk habitat suitability and habitat
potential models can identify desirable landscapes for elk conservation and habitat management
and planning.
12
METHODS
To demonstrate the application of integrating habitat models for wildlife habitat
management, we created a set of models (Elk Habitat Suitability – Public Lands, Elk Habitat
Suitability – Private Lands, Elk Habitat Potential) to quantify elk habitat suitability and habitat
potential for state-owned (hereafter public) and private lands within the MDNR designated elk
range (Scheme 1.1). Each model used a framework for quantifying habitat suitability or habitat
potential values based on elk habitat requirements (i.e., WTC, WF, SF) and supporting cover
types (e.g., aspen, cedar) determined by Beyer (1987) for Rocky Mountain elk in Michigan.
Winter thermal cover is provided by cedar and other lowland conifer swamps in Michigan and
allows elk to maintain homeothermy during severe winter conditions (Moran 1973, Beyer 1985).
Winter food is vital during harsh weather conditions, and is considered as browse (e.g., aspen)
that is available to elk above snow cover (Beyer 1987). Spring food is critical for elk to recover
any loss of physical condition during winter and generally considered to be high nutritional
quality forage made available following or during snow melt (Beyer 1987). While all 3 of these
habitat requirements for elk in Michigan occur in winter and spring, availability of food and
cover are most critical during these seasons and habitat suitability ratings are assumed to relate
directly to quality of habitat throughout the year.
13
Scheme 1.1. Schematic of model components and processes used to develop habitat suitability and habitat potential models for public
and private lands within the elk range in northeastern lower Michigan.
14
Elk Habitat Suitability – Public Lands
We created an HSI model using 8 dominant cover types (i.e., aspen, northern
hardwoods/maple, oak, other hardwoods, upland conifers, cedar, other lowland conifers,
openings) in our study area as identified using MDNR forest inventory data to quantify elk
habitat suitability on public lands in Michigan. Michigan DNR forest inventory data are
maintained through field inventories of designated forest compartments on a 10-year rotation,
and contain information describing key forest stand attributes (e.g. species, age, basal area,
percent canopy closure, management strategy) that allowed us to quantify elk habitat suitability
on public lands. All MDNR forest stands were categorized into cover types (Table 1.1) using
ArcGIS version 10.6.1 (Environmental Systems Research Institute, Redlands, CA, USA). We
assigned suitability values (SV) ranging from 0 (unsuitable habitat) to 1 (suitable habitat) based
on the ability of each cover type to provide seasonal life requisites based on Michigan elk winter
and spring habitat use patterns (Beyer 1987). Individual SVs were modified for each cover type
to more accurately describe suitability using key forest stand attributes identified by Beyer
(1987; Table 1.2).
15
Table 1.1. Habitat suitability values (scale 0–1; 1 = optimum) for cover types supporting each
life requisite (i.e., winter thermal cover, winter food, and spring food) for elk in northeastern
lower Michigan on state-owned public lands. Cover types selected based on MDNR forest
inventory data. Adapted from Beyer (1987).
Cover type Winter thermal cover Winter food Spring food
Aspen 0.3 1.0 0.7
Northern hardwoods/maple 0.3 1.0 0.5
Oak 0.3 0.7 0.4
Other hardwoods 0.3 0.5 0.3
Cedar 1.0 1.0 0.7
Upland conifers 0.5 0.7 0.5
Other lowland conifers 1.0 0.2 0.5
Openings 0.0 0.0 1.0
Table 1.2. Elk habitat suitability value modifiers for elk life requisites in northeastern lower
Michigan on state-owned public lands. Adapted from Beyer (1987).
Winter thermal cover Winter food/Spring food
Width of conifer stands Age of aspen stands
% canopy closure of conifer stands Density of cedar stands
Even/unevenaged management of conifer stands Density of hardwood stands
Stand age of conifers
Hardwood stand basal area
Hardwood dbh
Presence of conifer understory in hardwood stands
16
Similar to Beyer (1987), we determined the quality of WTC that conifer stands provide for
elk to be a function of stand width (MODStWid), percent canopy closure (MOD%CC), whether a
stand has been managed to be even or uneven-aged (MODEven/Uneven), and the age of a stand if
even-aged (MODConAge).
SVWTC Conifer = SVcover type x MODStWid x MOD%CC x MODEven/Uneven x MODConAge
Stand width was used based on the environmental differences (i.e., temperature, wind
currents, snow depths) between stand edge and interior, and the necessity of adequate stand size
to accommodate elk herding behavior (Beyer 1987). Beyer (1987) determined the optimal stand
width for elk to be 150 m or greater, with stands providing proportionally lower value as they
decrease in size. We measured individual stand widths using ArcGIS, and considered the largest
diameter within each stand to be its maximum width (i.e., to avoid assigning higher values to
stands that were wider in one direction than another). Stand canopy closure of conifer stands is
essential to elk for thermoregulation and energy conservation provided by a 75–100% complete
canopy closure (Verme 1965, Thomas et al. 1979). We identified stands managed under even-
aged management strategies due to their ability to modify wind currents and reduce heat loss to
elk (Verme 1965). Additionally, stand age was used for even-aged stands due to the relationship
between taller tree height and greater canopy depth, thus increasing the ability of a stand to
reduce snow depths and modify stand conditions at ground level (Beyer 1987). According to
Johnston (1977), conifers ≥12 m in height provide optimal thermal cover. Therefore, we used
site index curves for lowland conifer species (e.g., black spruce, northern white-cedar, balsam
fir) in the eastern United States to determine that a stand age of ≥33 years provides optimal WTC
for elk in Michigan (Carmean et al. 1989). Percent canopy closure, stand management strategy,
and stand age were determined using MDNR forest inventory data.
17
We determined the quality of WTC that hardwood stands provide using a function of the
average basal area of a stand (MODBA), tree size (dbh) (MODDBH), and presence of a conifer
understory in a stand (MODConUnd, Beyer 1987).
SVWTC Hardwood = SVcover type x MODBA x MODDBH x MODConUnd
We used basal area (MODBA) and maximum tree size (MODDBH) of hardwood stands to
modify WTC suitability values due to the ability of tree trunks to reduce air movements within
forest stands (Beyer 1987). Similar to Beyer (1987), we used the MDNR’s forest inventory
defined equivalent value of a well-stocked stand (i.e., basal area of 16m2/ha or greater) as the
optimal basal area to provide WTC for elk. Additionally, elk have been shown to select bed sites
next to the largest diameter trees in forest stands during winter (Beall 1974). Therefore, we
identified hardwood stands with tree sizes of ≥35 cm dbh to be optimal for providing WTC
(Beyer 1987). However, basal area and dbh modifiers were only used if a conifer understory was
present to provide horizontal cover and reduce air movements (Beyer 1987). Hardwood stands
without conifers in the understory were considered to have no WTC value for elk. We used
MDNR forest inventory data to identify basal area, tree dbh, and presence of conifer understory.
According to Beyer (1987), the quality of aspen as a WF or SF source for elk is a function of
aspen stand age.
SVWF Aspen and SVSF Aspen = SVAspen x MODAspenAge
Aspen is a valuable winter and spring food source for elk in Michigan, but previous research
found declines in browse use as age of aspen increased (Campa 1989, Campa et al. 1993,
Raymer 2000). Aspen stands 1–2.5 m tall (i.e., age 1–3 years) are ideal for normal browsing with
18
only larger animals being able to access browse from trees >3 m in height (Beyer 1987). We
determined aspen stands <7 years of age to have winter and spring food value for elk.
The quality of cedar as a WF or SF source for elk also is a function of stand age (Verme
1965).
SVWF Cedar and SVSF Cedar = SVCedar x MODCedarAge
Verme (1965) found cedar stands between 5–9 years-old provide the best quality and
quantity of food for deer in the Upper Peninsula of Michigan. While elk can reach higher for
browse, Beyer (1987) suggested greater snow depths in the Upper Peninsula than in the Northern
Lower Peninsula would negate the height advantage for reaching browse. We used MDNR forest
inventory data to identify aspen and cedar stand ages on public land. Additionally, the amount of
WF and SF that upland conifer stands can provide is a function of the presence of hardwood
species used as browse by elk in the understory (Beyer 1987). We considered upland conifer
stands without hardwood species in the understory to have no WF or SF value for elk.
Using ArcGIS, we created elk habitat suitability maps for public lands for each life requisite at a
resolution of 30 x 30 m (i.e., to remain consistent with the resolution of subsequent models). We
used a roving window with a focal mean function to recalculate suitability for each life requisite
to consider the spatial influence of elk movement patterns. Our roving window sizes (i.e., 1,053 m
for WTC and WF, 1,690 m for SF) were based on the mean maximum daily movement distance
(i.e., diameter) of radio-collared elk during winter 2016–2018 for WTC and WF, and spring 2016–
2018 for SF (Table 1.3). Additionally, we calculated the amount of area (km2) represented by each
suitability value (i.e., within our range of 0–1) to assess the distribution of habitat suitability for
each life requisite.
19
Elk Habitat Suitability – Private Lands
To quantify elk habitat suitability on private lands in Michigan, we classified cover types for
our entire study area (i.e., private and public lands) using satellite imagery to identify forest
stands since forest inventory data equivalent to those for state lands do not exist for private lands.
We used National Agriculture Inventory Program (NAIP) imagery (1 x 1 m, 2012) purchased
through the U.S. Department of Agriculture in ArcGIS to identify 5 cover types (i.e., aspen,
hardwoods, upland conifers, lowland conifers, openings) based on visual characteristics (e.g.,
shape, texture, color). We used the Image Classification toolbar in ArcGIS and delineated 10–15
training samples (i.e., polygons) representing a minimum of at least 5% of each cover type
through MDNR forest inventory data in public land areas. We used the Maximum Likelihood
Classification Tool to classify the NAIP imagery into our 5 cover types. To remove irrelevant
detail and improve classification, we resampled the imagery at a resolution of 30 x 30 m. To
determine the accuracy of satellite imagery classification methods, we validated each class using
cover type descriptions from MDNR forest inventory data for all public lands (664.9 km2) found
within our study area.
Cover types were assigned SVs (Table 1.4) based on Michigan elk winter and spring habitat
use patterns determined by Beyer (1987). We modified upland and lowland conifer cover types
for percent canopy closure using the Landscape Fire and Resource Management Planning Tools
(LANDFIRE) Forest Canopy Cover (CC) layer from 2012 (available at
https://www.landfire.gov/cc.php). We applied roving window sizes described in our previous
model to recalculate suitability for each life requisite, and produced elk habitat suitability maps
and plotted the distribution of HSI values by area (km2) for public and private lands for each
suitability map.
20
Table 1.3. Mean maximum daily movement diameter (MDMD) of GPS-collared elk during
winter (17 Feb–20 Mar, 2016; 21 Dec–20 Mar, 2017–2018) and spring (21 Mar–20 Jun, 2016–
2018) in northeastern lower Michigan.
Days
Season N monitored MDMD (m) SE
Winter 46 6,328 1,053.1 8.6
Spring 47 8,798 1,689.6 10.0
Table 1.4. Habitat suitability values (scale 0–1; 1 = optimum) for cover types supporting each
life requisite (i.e., winter thermal cover, winter food, and spring food) for elk in northeastern
lower Michigan on private lands. Cover types selected based on satellite imagery classification.
Adapted from Beyer (1987).
Cover type Winter thermal cover Winter food Spring food
Aspen 0.3 1.0 0.7
Hardwoods 0.3 0.8 0.5
Upland conifers 0.5 0.7 0.5
Lowland conifers 1.0 0.7 0.6
Openings 0.0 0.0 1.0
21
Elk Habitat Potential
We quantified elk habitat potential by delineating habitat types for northern lower Michigan
using a procedure similar to the one described by Felix et al. (2004). We overlaid 3 digital spatial
datasets in ArcGIS software (Scheme 1.1). Essentially, habitat types were the intersection of
landtype associations (LTAs), soils, and vegetation (Felix et al. 2004). The landtype associations
layer (Corner et al. 1999) helped describe plant species recruitment patterns and direction of
compositional and structural change across a landscape (Cleland et al. 1993) based on landform
and topographic characteristics. We used SSURGO soils data (Natural Resources Conservation
Service 2010) to describe soil moisture and texture, which limit potential vegetation types and
are important characteristics for classifying habitat types (Kotar and Burger 2000). We identified
specific land cover classes in various seral stages within habitat types using land cover
classifications from the Integrated Forest Monitoring, Assessment, and Prescription (IFMAP)
data (MDNR 2003) at a resolution of 30 x 30 m.
We converted the soils and LTAs from vector data models to raster and assigned grid codes
based on texture and moisture (soils) and landform (LTAs). All datasets were combined using
the Raster Calculator in ArcMap. Habitat types were identified using habitat type classification
guides (i.e., Coffman et al. 1980, Burger and Kotar 1999) and a hierarchical decision protocol
whereby soils and LTAs determined possible successional pathways and the vegetation dataset
(MDNR 2003) validated the habitat type assignment. In cases where discrepancies were
identified based on inaccuracies and inconsistencies in the datasets, we designated habitat types
by evaluating the landscape patterns of vegetation composition and structure using high
resolution (1 x 1 m) imagery purchased from the USDA (National Agriculture Imagery Program
2012).
22
We assigned SVs to each successional stage for each habitat type based on the maximum
value of suitability provided by cover types for each elk life requisite (i.e., based on prior
research by Beyer [1987]). Habitat potential was determined by selecting the highest SV of any
successional stage for each habitat type, and habitat potential maps were created for each life
requisite at a resolution of 30 x 30 m to reflect elk movements and feeding behavior.
Additionally, we identified key public land areas for elk habitat management focus where current
habitat suitability is low (i.e., ≤0.33) and habitat potential is high (i.e., >0.66).
23
RESULTS
Elk Habitat Suitability – Public Lands
We identified 8,625 cover type polygons from MDNR forest inventory data for public lands
within the Michigan elk range. The most abundant cover types on public lands (664.9 km2) were
aspen (25.43%, 163.97 km2), upland conifers (23.09%, 148.86 km2), and northern
hardwoods/maple (15.36%, 99.05 km2), with openings (i.e., maintained wildlife openings,
natural grasslands, old field grasslands), other lowland conifers, cedar, other hardwoods, oak,
water, and other making up ≤10% of public lands, respectively.
For WTC, approximately 79.06% (509.74 km2) of public lands were identified as areas with
low suitability (i.e., 0–0.33), 16.88% (108.87 km2) were medium suitability (i.e., 0.34–0.66), and
4.06% (26.16 km2) were high suitability (i.e., 0.67–1; Figure 1.2). The primary areas of high
suitability were cedar (i.e., 52.33%, 13.69 km2) and other lowland conifer (i.e., 41.74%, 10.92
km2) stands located in large, isolated clusters (i.e., 5–13 km2) in the northern and southern
portions of public lands (Figure 1.2). For WF, approximately 40.55% (261.46 km2) of public
lands were identified as areas with low suitability, 44.61% (287.66 km2) were medium
suitability, and 14.83% (95.64 km2) were high suitability (Figure 1.2). The primary areas of high
suitability were northern hardwoods/maple (i.e., 60.62%, 57.98 km2), upland conifer (i.e.
20.19%, 19.31 km2), and aspen (i.e. 6.07%, 5.81 km2) stands located throughout public lands
(Figure 1.2). For SF, approximately 32.53% (209.74 km2) of public lands were identified as
areas with low suitability, 66.4% (428.15 km2) were medium suitability, and 1.07% (6.88 km2)
were high suitability (Figure 1.2). The primary areas of high suitability were openings (i.e.,
65.35%, 4.5 km2) located in the central west portion of public lands, and northern
24
hardwoods/maple (i.e., 33.09%, 2.28 km2) stands that were interspersed among those openings
(Figure 1.2).
Elk Habitat Suitability – Private Lands
Cover type classification distributions for private lands (555.1 km2) were hardwoods =
40.6% (225.2 km2), openings = 20% (111.3 km2), upland conifers = 16.6% (92.3 km2), aspen =
14.3% (79.6 km2), and lowland conifers = 6.5% (36.4 km2). Our satellite imagery classification
method was moderately successful for upland conifers (66.1% accuracy), hardwoods (57.6%),
openings (56.8%), and lowland conifers (54.8%) on public lands where MDNR cover type data
was available for determination of accuracy (Table 1.5). Classification accuracy of aspen stands
on public lands was lower (36.4%) primarily due to mature aspen stands being visually
indistinguishable from mature hardwood stands (i.e., 45.4% were classified as hardwoods).
For WTC, approximately 66.1% (367 km2) of private lands were identified as areas with low
suitability, 33.8% (188 km2) had medium suitability, and 0.1% (<1 km2) had high suitability
(Figure 1.3). The paucity of private land areas with high suitability for WTC can be attributed to
the absence of mature cedar and lowland conifer stands that were large enough to provide high
suitability after roving window averaging. For WF, approximately 2% (11 km2) of the public
lands were identified as areas with low suitability, 53.5% (297 km2) had medium suitability, and
44.5% (247 km2) had high suitability (Figure 1.3). Most of the high suitability areas were
hardwood (i.e., 55.8%, 139 km2) and aspen (i.e., 19.2%, 48 km2) stands found throughout private
lands, and when combined account for over half (i.e., 52.6%) of all private land cover types. For
SF, we found no areas with low suitability, 82.6% (459 km2) had medium suitability, and 17.4%
(96.5 km2) had high suitability (Figure 1.3). The majority of areas with high suitability were
openings (i.e., 44.6%, 43 km2) found in clusters throughout private lands.
25
Figure 1.2. Distribution of elk winter thermal cover habitat suitability (A), winter food habitat
suitability (B), and spring food habitat suitability (C) on public lands (665 km2) within the
MDNR defined elk range (1,220 km2) in the northern lower peninsula of Michigan. Values are
based on a scale of 0–1; 1 = highest suitability.
26
Table 1.5. Accuracy assessment of classification methods used on National Agriculture Inventory Program (NAIP) imagery (1 x 1 m,
2012) to classify cover types (i.e., aspen, hardwoods, upland conifers, lowland conifers, openings) used in an elk habitat suitability
model for private lands in northeastern lower Michigan. Reported accuracy percentages were calculated by validation of classified
cover types (30 x 30 m) found within public lands (664.9 km2) using Michigan Department of Natural Resources forest inventory
cover type polygons.
Classified Aspen1 Hardwoods1 Upland conifers1 Lowland conifers1 Openings1
cover types Pixels2 % Pixels2 % Pixels2 % Pixels2 % Pixels2 %
a
Aspen 66,366 36.38 32,963 18.37 5,306 3.20 7,584 6.46 5,261 7.17
a
Hardwoods 82,751 45.36 103,362 57.60 13,405 8.10 22,105 18.82 14,844 20.24
a
Upland conifers 5,014 2.75 19,527 10.88 109,492 66.11 3,973 3.38 4,032 5.50
a
Lowland conifers 3,816 2.09 3,494 1.95 3,109 1.88 64,428 54.84 7,348 10.02
a
Openings 24,394 13.37 19,966 11.13 34,050 20.56 19,191 16.33 41,653 56.80
Water 96 0.05 121 0.07 251 0.15 203 0.17 200 0.27
1
Cover types identified within MDNR forestry inventory data.
2
The number of 30 x 30 m pixels contained within the classification raster.
a
Correct classification according to cover type descriptions within MDNR forest inventory data.
27
Figure 1.3. Distribution of elk winter thermal cover habitat suitability (A), winter food habitat
suitability (B), and spring food habitat suitability (C) within the MDNR defined elk range (1,220
km2) in the northern lower peninsula of Michigan. Values are based on a scale of 0–1; 1 =
highest suitability.
28
Elk Habitat Potential
We classified and delineated 13 habitat types in the Michigan elk range and determined
successional pathways for each based on biotic and abiotic characteristics (i.e., LTAs, soils, and
vegetation; Figure 1.4). Approximately 19% of the elk range is composed of a sandy upland dry
habitat type (i.e., RO/RM/WP) in which jack pine and aspen dominate early-successional (i.e.,
<30 years) stages; jack pine and red pine dominate mid-successional (i.e., 30–100 years) stages;
and red oak, red maple, and white pine dominate late-successional (i.e., >100 years) stages
(Figure 1.4). Approximately 18.7% of the elk range is composed of a sandy upland dry-mesic
habitat type (i.e., SM/Bee/H) in which aspen and white birch dominate early-successional stages;
red maple, beech, and white ash dominate mid-successional stages; and sugar maple, beech, and
hemlock dominate late-successional stages (Figure 1.4). Approximately 10.7% of the elk range is
composed of a sandy dry habitat type (i.e., WP/RP/O/JP) in which shrubs and grasses dominate
early-successional stages due to poor soil fertility; jack pine and red pine dominate mid-
successional stages; and white pine, red pine, and oak dominate late-successional stages (Figure
1.4). All other habitat types each represent <10% of the elk range (Figure 1.4).
Habitat types vary in their potential to provide elk life requisites. For WTC, 5 habitat types
provide maximum habitat potential (1.0), 5 provide medium potential (0.5), and 3 provide low
potential (0.3). However, habitat potential was high (0.7–1.0) across all habitat types for winter
and SF (Figure 1.5). Notably, all but one habitat type (i.e., WP/RP/O/JP) provide maximum
habitat potential during at least one successional stage for WF. While each habitat type provides
different potential across successional trajectories for each elk life requisite, general trends were
evident.
29
Figure 1.4. Proportion (Prop.) and location of habitat types within the MDNR defined elk range (1,220 km2) in the northern lower
peninsula of Michigan. Vegetation codes for early-, middle-, and late-successional stages are as follows: A = aspen, BA = Black Ash,
Bas = basswood, Bee = beech, BF = balsam fir, Bir = white birch, BP = balsam poplar, BC = black cherry, BS = black spruce, C =
cedar, G = grass species, H = hemlock, JP = jack pine, LBr = lowland brush, O = oak species, RM = red maple, RO = red oak, RP =
red pine, Shr = shrub, SM = sugar maple, T = tamarack, WA = white ash, WP = white pine.
30
Figure 1.5. Distribution of elk winter thermal cover habitat potential (A), winter food habitat
potential (B), and spring food habitat potential (C) in the MDNR defined elk range (1,220 km2)
in the northern lower peninsula of Michigan. Distribution of areas with low habitat suitability (<
0.33) and high habitat potential (> 0.66) for elk winter thermal cover (D), winter food (E), and
spring food (F). Habitat potential/suitability values are based on a scale of 0–1; 1 = highest
potential/suitability.
31
For WTC, all habitat types provide highest potential during mid to late-successional stages
when lowland conifer species are present and stand canopy closure is at least 75%.
Approximately 20% of the elk range has high potential (1.0) for WTC during mid to late-
successional stages. For WF, approximately 71% of the elk range has high potential (1.0) in
early-successional stages due to habitat types supporting young, regenerating aspen stands (i.e.,
<7 years-old) providing browse. Additionally, approximately 45% of the elk range has high
potential (1.0) in mid to late-successional stages when northern hardwood (e.g., sugar maple, red
maple, hemlock, basswood) stands provide browse, and approximately 13.8% of the elk range
has high potential (1.0) during mid to late-successional stages when young cedar stands (i.e., <25
years-old) provide browse. For SF, 10.7% of the elk range has high potential (1.0) during early-
successional stages when openings provide herbaceous forage. The remaining 89.3% of the elk
range has high (0.7) potential during early-successional stages when regenerating aspen stands
provide browse or mid to late-successional stages when young cedar stands provide browse.
To identify key public land areas for elk habitat management, we identified areas that had
low habitat suitability and high potential for each life requisite (Figure 1.5). For WTC,
approximately 15.2% (101 km2) of public lands in the elk range had low habitat suitability
(≤0.33), but high potential. The majority of areas in low suitability were aspen (33 km2), upland
conifers (14 km2), cedar (12 km2), openings (12 km2), other lowland conifers (11 km2), and other
hardwoods (11 km2) cover types. Aspen, upland conifers, openings, and other hardwoods cover
types that had low suitability were in early-successional stages of habitat types that have high
potential for WTC in late-successional stages. Cedar and lowland conifers cover types that had
low suitability were degenerating due to age or were too small (<50 m in diameter) in size to
provide WTC for elk (Beyer 1987).
32
For WF, approximately 39.4% (262 km2) of public lands within the elk range had low habitat
suitability, but high potential. The majority of areas in low suitability were aspen (94 km2),
lowland conifers (39 km2), openings (38 km2), and cedar (38 km2) cover types. Approximately
36% of areas that are currently in low suitability for WF were aspen stands ≥7 years-old that
were too mature to provide available browse for elk (Campa 1989, Campa et al. 1993, Raymer
2000). Similarly, cedar stands with low suitability were too mature (i.e., >25 years-old) to
provide WF (Verme 1965, Beyer 1987). Openings with low suitability were in habitat types that
will not provide WF unless they are allowed to reach mid to late-successional stages.
For SF, approximately 32.6% (217 km2) of public lands within the elk range had low habitat
suitability, but high potential. The majority of areas in low suitability were aspen (96 km2),
upland conifers (30 km2), openings (20 km2), and other hardwoods (20 km2) cover types. Similar
to WF, aspen stands that had low suitability for SF were too mature (i.e., ≥7 years-old) to
provide elk with ample browse. The majority of upland conifer and other hardwood stands with
low suitability were habitat types that provide aspen in early-successional stages. Openings that
had low suitability for SF were reduced in value due to their juxtaposition to comparatively
larger areas of lower value cover types from roving window averaging.
33
DISCUSSION
While our analyses demonstrated the utility of integrating habitat suitability and potential
models for elk habitat management planning in Michigan, other wildlife species with known
habitat requirements can benefit from habitat management using this approach. For example,
species of concern such as American marten (Martes americana) require specific structural
habitat characteristics (e.g., canopy cover, tree size, coarse woody debris) within specific habitat
types supporting pine-dominated vegetation types with few mixed hardwoods (Hargis and
McCullough 1984, Thompson and Colgan 1994, Godbout and Ouellet 2010). A habitat potential
model can be used to identify habitat types that support pine-dominated vegetation types (i.e.,
areas of high habitat potential), and a habitat suitability model can then be used to identify sites
that are currently in low suitability within those habitat types as focus areas for marten habitat
management efforts.
The spatial and temporal components of our approach may be especially useful for
consideration of where and when to implement management strategies for threatened or
endangered species with known habitat requirements. Habitat degradation and fragmentation are
commonly cited as primary sources of species endangerment, necessitating the importance of
identifying current and potential sources of habitat for threatened and endangered species
(Wilcove et al. 1998, Kerr and Cihlar 2004, Schaffer-Smith et al. 2016). Linden (2011) used a
habitat potential model for Canada lynx to determine potential hare and lynx densities in the
Upper Peninsula of Michigan. While his model defined habitat types to identify vegetation
attributes for the prediction of hare and lynx densities, it did not assign habitat suitability values
to successional stages for consideration of habitat potential for life requisites of lynx. Managers
34
could apply our approach of integrating habitat suitability and potential models to target areas of
highest potential to maintain or improve lynx habitat quality for conservation efforts.
Additionally, reintroduction or restoration efforts may benefit by integrating habitat
suitability and potential models. The persistence of a reintroduced population is dependent on the
ability of managers to maintain the necessary habitat requirements for a given species
(Armstrong and Reynolds 2012); thus, requiring collection of data to determine habitat
suitability before introduction and monitoring of key habitat variables following a release. For
instance, many mangers use population viability analysis (PVA) to identify key parameters (e.g.,
habitat restoration) that will increase the probability of success for a reintroduced species
(Haines et al. 2006, Kindall et al. 2011). Some PVAs have linked estimates of habitat availability
to population viability by simulating decreases in habitat quality or quantity over time (Akçakaya
et al. 1995, Nickelson and Lawson 1998, Larson et al. 2004). While these examples incorporated
temporal changes to habitat suitability and availability, each only addressed specific scenarios or
simulations for changes to habitat. We believe using habitat types to determine potential habitat
suitability for any successional stage provides the flexibility to understand the spatial and
temporal dynamics of wildlife habitat for any species with known habitat requirements in any
landscape.
In our case study, we determined less than half (i.e., 38%) of the Michigan elk range had
high habitat suitability for at least one life requisite of elk. Conversely, nearly all (i.e., 96%) of
the elk range had high habitat potential for at least one life requisite for elk. While the proportion
of areas with low habitat suitability varied for each life requisite, the potential for those areas to
provide habitat for elk was high (Figure 1.6). For example, habitat suitability was low for WTC
on private and public lands, with only a few primary areas (26.16 km2) of high suitability on
35
public lands in the north and southern areas of the elk range (Figure 1.3). However, habitat types
that provided high potential for WTC composed 21% (256 km2) of the elk range; hence, only
10.2% of areas capable of providing WTC for elk were currently in high suitability. Beyer
(1987) suggested that ideally elk habitat should be comprised of 10% WTC with the highest
suitability (i.e., 1.0). While only 4.06% of public lands and 0.1% of private lands were
determined to provide high suitability (i.e., 0.67–1.0) in our models, 16.88% of public lands and
33.8% of private lands were in medium suitability (i.e., 0.34–0.66) and should be considered
valuable for providing WTC for elk. Additionally, we believe the low proportion of high
suitability WTC areas in the Michigan elk range is sufficient based on observations made by
Moran (1973) where elk were not observed using thermal cover until snow depths exceeded 46
cm. In Pennsylvania, elk have been found using conifer stands and lowland drainages when snow
depths were ≥ 60 cm (DeBerti 2006). Additionally, Moran’s (1973) findings were similar to
western states where elk movements were restricted at depths of 41 cm (Sweeney and Steinhoff
1976) and 46 cm (Beall 1974, Leege and Hickey 1977). Notably, elk nearly exclusively used
conifer stands when snow depths exceeded 60 cm in Glacier National Park, MT (Martinka 1976).
According to NOAA (2018), snow depth records within 8 km of our study area exceeded 46 cm
during 20% of days during the winters from 2008–2018. While it is vital to maintain areas that
provide WTC within the elk range, the low proportion of winter days necessitating the use of
WTC may allow managers to focus management efforts on maintaining and improving habitat
for winter and SF for elk.
In contrast to WTC, results indicated habitat suitability for winter and SF was higher and
more evenly distributed across public and private lands within the elk range. We attribute an
abundance of high suitability areas to habitat management by the MDNR on public lands since
36
the mid 1970’s (MDNR 2012), private land management on club lands, and high habitat
potential for winter and SF across nearly all of the elk range. The only areas without high habitat
potential are developed, urban, or agricultural areas; <4%. The potential for all habitat types
within the elk range to provide high suitability for winter and SF is primarily due to: 1)
approximately 73.5% of the elk range composed of habitat types that support aspen; 2)
approximately 24.5% of the elk range composed of a habitat type that supports openings
throughout early-successional stages.
The differences between habitat suitability and potential in the Michigan elk range were due
to habitat types not being in successional stages that currently provide high suitability for elk life
requisites (i.e., food or thermal cover). For example, 75.3% of the elk range is composed of
habitat types that support aspen during early or mid-successional stages. Aspen was the most
abundant cover type on public lands within the elk range, which was consistent with the
statewide distribution of forest types for the Lower Peninsula of Michigan (MDNR 2008).
However, only 4.6% of aspen stands on public lands were <7 years-old and provided value for
elk winter and SF in our models (i.e., the majority [70.9%] of aspen stands were 20–50 years-
old). Despite the low proportion of young stands, aspen still composed 6.07% of the high
suitability areas for WF. The MDNR (2012) currently manages for no net loss of aspen stands
(i.e., 27% of public land forest cover types), and for numerous age classes. Harvesting an
extensive number of mature (i.e., > 40 years-old) aspen stands in any year may promote an
abundance of regeneration and available browse for elk and other wildlife species (e.g., ruffed
grouse [Bonasa umbellus], white-tailed deer [Odocoileus virginianus]), but would reduce the
amount of aspen stands available for cutting in subsequent years until harvest age is reached
creating a “boom and bust” scenario for wildlife populations depending on various stages of
37
aspen development (MDNR 2008:56, Felix-Locher and Campa 2010). While the majority of the
elk range has high habitat potential due to its ability to support aspen in early successional
stages, we believe the current forest management strategy to maintain no net loss of aspen of
numerous age classes provides an adequate winter and SF source for elk in Michigan. Beyer
(1987) described optimal elk habitat in Michigan as providing ≥15% of available cover types in
highest suitability (i.e., 1.0) for WF and ≥10% of available cover types in highest suitability (i.e.,
1.0) for SF. We found 14.83% of public lands (i.e., 95.64 km2) in high suitability (i.e., 0.67–1.0)
for WF with aspen contributing 6.07% (i.e., 5.81 km2). For SF, we found only 1.07% of public
lands in high suitability (0.67–1.0). However, 44.61% of public lands were found to be in
medium suitability for WF with aspen contributing 21.96% (63.16 km2), and 66.4% of public
lands were found to be in medium suitability for SF with aspen contributing 15.52% (66.4 km2).
We believe the abundance of medium suitability areas for winter and SF should be considered
valuable when considering the availability of food for elk. Notably, the Michigan elk population
is considered healthy with a stable population size and successful hunting seasons since 1984
(MDNR 2012, MDNR 2019).
Identifying aspen stands on private lands will allow managers to realize potential areas of
high suitability that may attract elk or other wildlife species from public land areas. For private
lands, we found aspen only composed 14.3% of the cover types. However, our satellite imagery
classification of aspen stands on public lands only classified 36.38% of aspen stands correctly,
with 45.36% of aspen being misclassified as hardwoods. We attributed the misclassification of
aspen to the similar appearance of hardwoods in our satellite imagery. While we recognize this
as a limitation in our model, the misclassification of aspen was conservative and did not
overestimate habitat suitability in our model. The misclassification of aspen to hardwoods only
38
reduces our HSI values from 1.0 (i.e., 100% aspen) to 0.8 (i.e., 100% hardwoods) for WF and 0.7
(i.e., 100% aspen) to 0.5 (i.e., 100% hardwoods) for SF. Additionally, our inability to determine
the age of aspen stands on private lands likely overestimates the suitability value of aspen stands
≥7 years of age. This is especially true for SF where all aspen stands were valued at 0.7, which
likely inflated the proportion of areas in high suitability. While we are aware of these limitations
in our private lands model, we believe the reduction in suitability value for 63.6% of aspen
stands on private lands may more accurately represent winter and SF values for aspen on private
lands when considering the aforementioned 70.9% of aspen stands within the elk range are 20–
50 years-old. While we believe the misclassification error in our private lands model did not
strongly affect our habitat suitability values, we advise wildlife managers to be aware of model
limitations and additional information such as stand structural conditions (e.g., age) to inform
habitat management decisions. However, use of data with relatively low accuracy may still be
important for identifying areas of interest to improve accuracy with additional analyses, or to
validate models through field sampling.
Openings have been recognized as a vital habitat component for elk in the eastern US
(Devlin and Tzilkowski 1986 [Pennsylvania], Dahl 2008 [Kentucky]), and are used by elk in
Michigan more often than all but regenerating deciduous and northern hardwood stands (Beyer
and Haufler 1994). The MDNR (2012) actively manages public lands within the elk range to
maintain 6–7% of cover types as openings (i.e., grass/upland brush). While only 1.07% of public
lands and 17.4% of private lands were found to be high suitability for SF, openings accounted for
most of those areas at 65.35% and 44.6%, respectively. More notably, the use of a roving
window (i.e., to average habitat suitability values in consideration of elk movement patterns)
reduced the amount of high suitability in areas where openings were small enough to be reduced
39
in value by surrounding lower value cover types for SF (e.g., upland conifers). The MDNR Elk
Management plan suggests that openings should be distributed across the elk range “as even as
possible considering ecological conditions” (MDNR 2012:23). Openings within the elk range
can vary in size (<1–57 ha) depending on management goals and site conditions (S. Heistand,
MDNR, personal communication). While openings that are smaller in size than nearby larger
areas of lower suitability may be reduced in value by roving window averaging and not reflect
high suitability areas in our model, they accounted for 9.29% (39.77 km2) of medium suitability
areas for SF which reflects the overall value and contribution to elk habitat suitability at a
landscape level.
40
CHAPTER 2: EXAMINING TRAIL-BASED RECREATIONAL USE PATTERNS IN
TWO CONTIGUOUS STATE FORESTS WITH DIFFERENT USE REGULATIONS IN
MICHIGAN
INTRODUCTION
The growing use of wild areas for outdoor, nature-based recreation has necessitated the
monitoring of recreational activities and associated ecological impacts around the world
(Balmford et al. 2009, Balmford et al. 2015, Cordell 2012). During 2000-2009, the number of
US citizens participating in outdoor activities grew 7.5% and the number of participation days
increased by 32.5% (Cordell 2012). The most recent findings by the Outdoor Foundation (2020)
reported 50.7% (i.e., 153.6 million Americans) of the US population ≥ 6 years of age participated
in outdoor activities in 2019, which was an increase of 1.2% from 2018. Among these trends is
growth in nature-based activities with wildlife viewing, wildlife photography, off-highway
vehicle driving, and physically challenging activities (e.g., kayaking, surfing, snowboarding)
having the largest increases in participants and number of days per year during the first decade of
the 21st century (Cordell 2012).
Nature-based recreation has been recognized as one of the fastest growing sectors of tourism
(Winter et al., 2020). Consequently, numerous state and federal public lands have seen an
increase in recreational users and are an increasingly important destination for tourists (Cordell
2012, Winter et al. 2020). In the US, approximately 75% of backcountry activities (e.g.,
backpacking, day hiking, equestrian use, mountain biking), 58% of wildlife viewing and
photography, 53% of motorized activities (i.e., off-road vehicle use, snowmobiling), and 50% of
hunting occurred on public lands during 2005-2009 (Cordell 2012). Notably, nature-viewing and
photography were the most popular activities (i.e., approximately 10x higher than backcountry
41
activities) on public lands with 15.1 billion days of activity reported in eastern states, and 5.2
billion days in western states (Cordell 2012).
The trend in the US of increasing outdoor recreation is evident in Michigan with 63% of its
residents participating annually, which is approximately 12% greater than the national average
(Outdoor Foundation 2020, Outdoor Industry Association 2017). In 2017, outdoor recreational
activities in Michigan generated $26.6 billion (USD) in consumer spending, 232,000 direct jobs,
$7.5 billion (USD) in wages and salaries, and $2.1 billion in state and local tax revenue (Outdoor
Industry Association 2017). Outdoor recreation is a mainstay of Michigan culture and the state
offers numerous opportunities for camping, hiking, hunting, fishing, wildlife viewing, cycling,
equestrian use, snowmobiling, and off-road vehicle (ORV) use (MDNR 2018a). One
contributing factor for these patterns in recreation is every Michigan community is located
within 80.4 km of a state park or recreation area (MDNR 2018a). Michigan has approximately
32,375 km2 of public land with the Michigan Department of Natural Resources (MDNR)
managing 18,615 km2 in its state forests, parks, game areas, and recreation areas (MDNR
2018a). Notably, Michigan’s Division of Parks and Recreation provides 20,117 km of trails
statewide, including 2,100 km of equestrian trails, 6,500 km of hiking trails, 2,250 km of
mountain biking trails, 5,800 km of ORV trails, and 10,000 km of snowmobiling trails (MDNR
2018a). Among public lands managed by the MDNR, approximately 87% (i.e., 15,783 km2) are
composed of state forests that are managed to conserve natural resources and provide natural
resource-based economic activity and recreation. Whereas Michigan’s state parks and recreation
areas focus on providing recreational opportunities, its state forests must balance multiple
management objectives of forest “use and restoration within a framework of long-term
sustainability, while also enabling an expanding diversity of uses” (MDNR 2008:xii).
42
Although Michigan provides numerous public lands for recreation, the presence of a visible
elk (Cervus elaphus nelsoni) herd (i.e., approximately 1,196 elk in 2019 [unpublished data
provided by MDNR]) in the Pigeon River Country State Forest (PRC) and part of the adjoining
Atlanta State Forest (ASF), makes these areas an attractive destination (Hunt 2019). The use of
the PRC for nature-based recreation has increased over the last 50 years with noticeable growth
in visitor numbers and interest in trail-based recreational opportunities (e.g., equestrian use,
mountain biking, ORV use; MDNR 2007, MDNR 2012). Increasing use of the PRC by
equestrian users and ORV users was first noted in 1970, and the PRC Advisory Council helped
implement plans to prohibit ORVs in 1988 and control the “dramatic increase in horseback
riding” in 1994 (MDNR 2007:12).The MDNR documented increasing trends in mountain biking
and wildlife viewing via recreation surveys conducted in the PRC from 1981-1982, 1986-1987,
and 1997-1998 (MDNR 2007).
The PRC has different regulations for trail-based recreation than other state forests (e.g., ASF)
in Michigan, due to its designation as a Special Management Unit with a “Concept of
Management” (COM) created to safeguard the lower peninsula’s last “big wild” from overuse
and development (MDNR 2007:14). Notably, the first objective of the COM is the management
of elk and their habitat to conserve the core elk range within the PRC, while its third objective is
to provide recreational opportunities that maintain the PRCs wild character through control of
disruptive activities (MDNR 2007). Although these objectives are not mutually exclusive, the
growth in nature-based recreation in the elk range has created challenges for managers charged
with balancing the PRC’s COM objectives (MDNR 2007).
We examined recreation patterns in the PRC and a portion of the adjoining ASF (i.e., public
lands within the Michigan elk range) to inform natural resource managers of the extent and
43
characteristics of common summer-fall trail-based recreational activities under differing use
regulations. Our objectives were to: 1) quantify the intensity, temporal characteristics (i.e., year,
month, day of week, time of day), and group sizes of common types of summer-fall trail-based
recreation (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) occurring in the
PRC and ASF; 2) characterize differences in visitor-use patterns between the PRC and ASF; and
3) provide recommendations for sustaining diverse recreational opportunities in the PRC and
ASF.
44
METHODS
Trail Camera Placement Protocols and Camera Settings
We evaluated 4 types of trail-based recreational use (i.e., equestrian, hiking/foot-traffic,
mountain biking, ORV) within the PRC and ASF for summer and fall (i.e., May–October) during
2016–2018. We quantified these common types of recreation based on observed seasonal user
trends in our study regions and due to concerns of natural resources professionals regarding their
growing numbers and potential impacts to elk and other wildlife (B. Mastenbrook, MDNR,
personal communication). We used 3 remote digital trail camera brands and models (i.e.,
RECONYX PC900 HyperFireTM Professional High Output Covert IR, RECONYX, Inc.,
Holmen, WI; Browning Dark Ops Elite HD, Browning, Inc., Morgan, UT; Stealth Cam model #,
Stealth Cam, Inc., Grand Prairie, TX) to capture the diversity of recreation types likely to be seen
in our study regions. To capture fast moving recreational users (i.e., mountain biking, ORVs) we
used RECONYX trail cameras due to their fast shutter speeds, which have been found to have
high detection rates for humans and large mammals (Gompper et al. 2006) and outperform other
brands in multiple studies (Duke and Quinn 2008, Hughson et al. 2010, Kelly and Holub 2008).
Browning and Stealth Cam trail cameras were primarily used in areas where mountain biking or
ORV use was prohibited.
Each trail camera model used a passive infrared motion detector to capture motion-induced
changes in ambient infrared, and a no-glow covert infrared flash to remain unobtrusive. Cameras
were mounted on trees using locking cables approximately 3–5 m above the ground at distances
of 5–10 m depending on availability of trees, line of sight, and ability to obscure cameras from
recreational users (i.e., placement behind trees or branches, using brush to conceal cameras and
locks). We verified accuracy of camera mounting positions using the aiming function to ensure
45
high likelihood of capturing recreational users. All trail cameras were programmed to capture
images 3–5 times per motion-triggered event at 1-second intervals with no delay between image
captures at high sensitivity. Each camera operated 24 hours per day, and images were
programmed to include the camera ID and date and time of acquisition. We used 2 different
camera placement designs for monitoring recreation use to accommodate different recreation
regulations for each study region.
PRC Description and Trail Camera Placement
The PRC State Forest is 458.4 km2 of nearly contiguous land in the northeastern lower peninsula
of Michigan and designated as a Special Management Unit by the MDNR (MDNR 2007) (Figure
2.1). The area that became the PRC in the early 20th century was first referred to as “The Big
Wild” due to its lack of development and mosaic of numerous forest types, rolling hills, lakes
and streams, and swamps that create a “variety found nowhere else” in the lower peninsula of
Michigan (MDNR 2007:3). The PRC has been the core of the Michigan elk range since 1917 and
provides habitat for black bear (Ursus americanus), bobcat (Lynx rufus), white-tailed deer
(Odocoileus virginianus), ruffed grouse (Bonasa umbellus), American woodcock (Scolopax
minor), brook trout (Salvelinus fontinalis), and many other fish and wildlife species (MDNR
2007).
In 1970, recent drillings for newly discovered oil and gas led to increased concerns of
changes to the formerly quiet and undisturbed wild nature of the PRC (MDNR 2007). This
industrial activity and additional concerns over expansion of campgrounds, pathways, timber
harvests, wildlife cuttings, and increasing occurrences of equestrian users, ORVs, and
snowmobiles culminated in the creation of the Pigeon River Country Association (PRCA). The
PRCA submitted a request to designate the PRC as a special management area in 1972, which
46
Figure 2.1. Location of designated recreational trails and campgrounds in the 1,220 km2 Michigan Department of Natural Resources
designated elk range and study area in the northern lower peninsula of Michigan. Trail camera locations shown (n=78) were during 24
May–30 September, 2018.
47
provided restrictions to natural resources manipulations, restricted vehicular traffic, designated
its streams as “Wild Rivers”, and established a primary focus of sustainable management of
resources (MDNR 2007). However, continued pressure to develop the PRC’s oil and gas wells
resulted in a multi-year struggle that ended in the Michigan supreme courts. As a result, oil and
gas exploration and development would be limited to the southern third of the forest and a trust
fund was established in 1976 to secure lease revenues for development of new public lands for
recreation. Frequent misappropriations of the funds led to a 1984 constitutional amendment that
established the new Michigan Natural Resources Trust Fund (MNRTF), which was safeguarded
against further diversions of funds (MDNR 2007). Since its original conception in 1976, the
MNRTF has awarded (i.e., as of 20–September, 2019) approximately 1.2 billion US dollars to
2,366 projects occurring throughout all 83 counties in Michigan (MDNR 2020a), including
purchases of an additional 50 km2 for the PRC (MDNR 2007).
The PRC is a model of multiple-use land management, outlined in the management
objectives and guidelines in its COM (MDNR 2007). The first objective of the COM addresses
the management of elk and their habitat: “Manage the elk population and elk habitat so the
Pigeon River Country State Forest remains the nucleus of Michigan’s elk herd” (MDNR
2007:14). The second objective focuses on management of habitat for other fish and wildlife
species. The third objective addresses recreational use: “Provide recreational opportunities for
people in keeping with the wild character of the area and to provide peace and quiet through
control of disruptive activities” (MDNR 2007:14). The remaining 5 objectives address the
management of game and fish species, forest and mineral resources, and protection from overuse
and overdevelopment.
48
While the primary focus of the PRC’s policies and objectives concern the management of elk
and other wildlife species and their habitats, opportunities for recreation remains a concern
within the COM’s detailed recreational use criteria to maintain the forest’s wild character. Since
its conception, the PRC has been managed with more restrictions limiting development and
human activity than most other state forests in Michigan to conserve its wild character from
heavy use (MDNR 2007). The PRC’s COM outlines its recreational use criteria for each
recreational type that is common in the lower peninsula of Michigan. Notably, many of the rules,
regulations, and guidelines are different from what are outlined in Michigan’s State Forest
Management Plan (2008), and the COM states that the PRC “cannot be all things to all recreation
users” (MDNR 2007:23). For example, ORV use is prohibited in the PRC, while being restricted
to forest roads and designated trails in most other state forests (e.g., ASF; MDNR 2007, MDNR
2018b; Table 2.1). Equestrian use and mountain biking are permitted on county and forest roads
and designated trails, while off-road and off-trail riding is permitted in most other state forests
(e.g., ASF; MDNR 2007; M. Fry, M. Monroe, MDNR, personal communication; Table 2.1).
The PRC has a well-developed and maintained network of forest roads and designated trails
(Figure 2.1). According to the MDNR Resource Assessment Unit, 88% of the PRC is within 0.8
km of a road (S. Whitcomb, MDNR, personal communication). Additionally, as of 2019 there
were 208.9 km of recreational trails within the PRC with 79.8 km designated for bicycling, 69.3
km designated for equestrian use, and hiking allowed on all trails. The PRC also provides 12
rustic campgrounds including 6 designated for equestrian users, of which all but 1 provides
direct connections to designated equestrian trails (Figure 2.1). Notably, 2 (i.e., Elk Hill
Trail/Group Campground, Johnson’s Crossing Trail/Group Campground) of the 6 equestrian
campgrounds are designed for large groups (i.e., a minimum of 10 people per group with a
49
maximum campground capacity of 100 individuals) and provide amenities including fire rings,
tables, toilets, potable water, and manure bunkers. Additionally, a pavilion is available at the Elk
Hill Trail/Group Campground for recreational users.
We selected sites for camera placement to maximize capturing recreational user activities on
trails in the PRC by focusing on intersections of trails, intersections of trails and forest roads, and
trailheads (i.e., with an emphasis on trailheads and proximity to campgrounds). In 2016, we used
21 cameras (i.e., 4 Browning, 12 RECONYX, 5 Stealth Cam) in the PRC that operated 20 May–
31 October. In 2017, we used 28 cameras (i.e., 18 Browning, 5 RECONYX, 5 Stealth Cam) in
the PRC that operated 16 May–28 October. In 2018, we used 32 cameras (i.e., 14 Browning, 9
RECONYX, 9 Stealth Cam) in the PRC that operated 24 May–30 September (Figure 2.1). In
2018, we removed cameras at the end of September to avoid camera thefts and damage during a
period of increased wildlife viewing, hunting, and logging.
50
Table 2.1. Trail-based recreational use regulations for primary summer-fall recreation types for
the Pigeon River Country (PRC) State Forest and Atlanta State Forest (ASF) Management Units
in the northern lower peninsula of Michigan.
Recreation type PRC ASF
Equestrian use Limited to designated trails Permitted in all areas unless
and all forest/county roads1 specified2
Hiking/foot-traffic Permitted in all areas unless Permitted in all areas unless
specified1 specified2
Mountain biking Limited to designated trails Permitted in all areas unless
and all forest/county roads1 specified2
ORV use Prohibited within the PRC1 Limited to designated trails
and all forest/county roads,
except for hunters attempting
to retrieve deer, elk, or bear
at speeds of 8 kph or less3
1
(MDNR 2007)
2
(M. Fry, M. Monroe, MDNR, personal communication)
3
(MDNR 2018b)
ASF Description and Trail Camera Placement
Approximately 168 km2 of the Michigan elk range is in the ASF (Figure 2.1). Located
adjacent to the southwestern edge of the PRC, the ASF shares many of the same geographic
features and communities of plant, fish, and wildlife species. However, the Michigan State
Forest Management Plan (MDNR 2008) directs the management of fish and wildlife populations
and habitats, and the Michigan Statewide Outdoor Recreation Plan (MDNR 2018a) directs
recreational use in the ASF. Hence, there are differences in opportunities and policies for
recreation use between the ASF and PRC. For example, ORV use is permitted on all forest roads
and trails within the ASF (MDNR 2018b), and equestrian users and mountain bikers are
51
permitted to ride on all forest roads, recreational trails, and off-trail anywhere within the forest
(M. Fry, M. Monroe, MDNR, personal communication; Table 2.1). The ASF has very few
designated trails (i.e., 29.9 km), compared to the PRC, for recreation with only one primary trail
running west from the PRC and ending in an array of short trails near the eastern edge of the elk
range (Figure 2.1). Additionally, there are no designated camping areas or amenities within the
forest.
Due to the limited number of designated trails, we used field observations of recreational
activities and camping sites and discussions with users to determine areas (e.g., county roads,
forest roads, forest edges) where we likely would observe the greatest intensity of trail-based
recreational use. In 2016, we used 16 cameras (i.e., 5 Browning, 3 RECONYX, 8 Stealth Cam)
that operated 20 May–31 October. In 2017 and 2018, we used a camera movement sampling
design to evaluate additional locations throughout the field season to increase the probability of
capturing a greater number and diversity of recreational events. Cameras that failed to capture
recreation images after 2 weeks or captured < 10 images over a 3-week period were moved to a
new location. Additionally, we relocated cameras near (i.e., 50–200 m) cameras that consistently
captured images (i.e., reflecting high-use areas), but in areas that might capture riding behavior
that was not commonly observed (e.g., riding along forest edges instead of forest roads, edges of
forest clear-cuts or wildlife openings). In 2017, we used 32 cameras (i.e., 15 Browning, 9
RECONYX, 8 Stealth Cam) during 16 May–28 October, and in 2018 we used 43 cameras (i.e.,
36 Browning, 4 RECONYX, 3 Stealth Cam) during 24 May–30 September (Figure 2.1). In 2018,
we removed cameras at the end of September to avoid camera thefts and damage during a period
of increased wildlife viewing, hunting, and logging.
52
Trail Camera Data Collection and Analyses
We checked cameras weekly during May–August and bi-weekly during September–October
throughout 2016–2018. We recorded camera ID, date, time, recreation type, and group size for
each image. We removed images that contained no recreational users or when recreation type
could not be determined from analysis (e.g., blurred image, majority of user was out of image
frame). A recreation event was defined as any number of individuals of the same recreation type
passing a camera in the same direction within a 5-minute period. Individuals or groups that
passed a camera in any direction > 5 minutes after their initial pass were counted as separate
events. However, individuals or groups that turned around in front of or passed the same camera
in the opposite direction within a 5-minute period were not counted as separate events (Coltrane
and Sinnott 2015). Recreation events were sorted by study region (PRC, ASF), type (equestrian
use, hiking/foot-traffic, mountain biking, ORV use), and categorized by year (2016–2018),
month (May–Oct), day of week, and time of day (hour: 0–23).
To compare recreational intensity between regions for each temporal category, we divided
the number of recreation events by the total number of operational camera hours to correct for
differences in camera operating hours between regions and among operating days, months, and
years. We referred to the quotients as relative recreational intensity (RRI; i.e. the number of
recreation events per camera hour during each temporal category, respectively). We did not
calculate RRI for time of day since we monitored 24 hours per day and the number of camera
hours per time of day was equivalent for each hour interval (0–23). To evaluate recreational user
trends for peak individual days of use during 2016–2018, we extracted the outliers (i.e., > Q3 +
1.5 × interquartile range [IQR]) of median daily RRI values for each study region and evaluated
the distribution of days occurring during each study year, month, day of the week, holiday
53
weekends, and for each recreation type. Analyses were conducted in R (R Core Team 2018) and
figures were produced using the ggplot2 package (Wickham 2016).
To determine which variables were most predictive of recreational intensity, we used a
generalized linear mixed model (GLMM) with a Poisson distribution for each region. To
standardize evaluation periods among years, we removed data prior to 24–May (i.e., our latest
start date among 2016–2018) and after 30–September (i.e., we did not monitor recreation during
October 2018) from 2016 and 2017. Additionally, we modified the category for time of day to 3
6-hour periods (i.e., 5:00–10:59, 11:00–16:59, 17:00–23:00) to avoid model errors for consistent
periods of inactivity (i.e., 2% of recreation events during 2016–2018 occurred from 23:00–4:59).
Thus, we defined the response variable as the mean number of events detected during each 6-
hour period within a given time period (e.g., month, day of week). Our models’ fixed effects
included variables for recreation type (i.e., equestrian use, hiking/foot-traffic, mountain biking,
ORV use), year (i.e., 2016–2018), month (i.e., May–September), day of week, time of day (i.e.,
5:00–10:59, 11:00–16:59, 17:00–23:00), the two-way interactions between each variable, and a
random effect of the number of operational camera hours for a given day to account for
variability. We did not include ORV use in the PRC model since ORV use is prohibited and only
accounted for 0.6% of recreation events in the PRC during our study.
Our GLMM outputs (i.e., emmeans) were the logmean number of events during all 6-hour
periods for a given selection of the aforementioned variables (e.g., logmean number of mountain
biking events occurring during May of 2017 in the PRC). Emmeans were obtained by averaging
each level of aforementioned variables that was not currently being considered (e.g., the emmean
for mountain biking events occurring during May of 2017 was obtained by averaging over the
levels of day of week and 6-hour periods). We used post-hoc Tukey’s pairwise comparison tests
54
to interpret GLMM results for each variable and the interactions between variables. To determine
which variables had the most effect on detection of a recreation event, we compared the
magnitudes of F values for each independent variable. To make model outputs more informative,
we back-transformed emmeans to represent the actual estimated mean number of events
occurring during a 6-hour period for a given combination of variables (i.e., hereafter referred to
as MN6HR). Analyses were conducted in R (R Core Team 2018) using the lme4 (Bates et al.
2015) and emmeans (Lenth 2020) packages.
To examine the number of events occurring throughout the day at a finer scale than we used
in our GLMMs (i.e., 6-hour intervals), we evaluated for differences among the hourly
distributions (i.e., 1-hour intervals) of recreation events for each recreation type using non-
parametric Kruskal-Wallis (KW) tests. We used the number of events during 1-hour intervals as
the response variable and temporal categories (i.e., year, month, day of week, time of day) as
independent variables, and used post-hoc pairwise comparison tests to interpret model results
among hours and recreation types (Siegel and Castellan 1988).
We examined group sizes (i.e., number of users during a recreation event) of recreational
users based on previous findings of increased flight distances and decreased observations of
wildlife with increased visitor group sizes (Frid and Dill 2002, Hamr 1988, Remacha et al.
2011). To evaluate differences in group size among temporal categories (i.e., year, month, day of
week, time of day) for each recreation type, we used KW tests with group size as the response
variable and temporal categories as independent variables. We used post-hoc pairwise
comparison tests to interpret model results among temporal categories and recreation types. To
evaluate the temporal characteristics for the largest group sizes of each recreation type, we
extracted the outliers (i.e., > Q3 + 1.5 × IQR) of median group sizes for each recreation type in
55
each study region. We used KW tests for time of day and group size analyses to avoid violating
assumptions of normality for parametric tests. We used an alpha level of 0.05 for all tests for
significance. Analyses were conducted in R (R Core Team 2018) using the pgirmess (Giraudoux
2018) package.
56
RESULTS
Relative Intensities, Temporal Characteristics, and Group Sizes of Recreational Users in
the PRC
Evaluation of recreational intensity in the PRC
We captured 11,412 recreation events during 263,664 hours of monitoring in the PRC from
2016–2018 (Table 2.2). We censored 17 events due to inability to determine recreation type. The
overall RRI (i.e., relative recreational intensity) was 0.043 (i.e., events per camera hour) and the
overall MN6HR (i.e., mean number of events during a 6-hour period) was 0.829. Time of day
had the most effect on detection of a recreation event within a 6-hour period followed by
recreation type, month, day of week, and year (Table 2.3). Equestrian use was the most
frequently detected type of trail-based recreation (i.e., RRI = 0.025, MN6HR = 2.088), followed
by hiking/foot-traffic (i.e., RRI = 0.015, MN6HR = 1.208), mountain biking (i.e., RRI = 0.003,
MN6HR = 0.226), and ORV use (i.e., RRI < 0.001) (Table 2.4).
Evaluation of recreational intensity by year and month in the PRC
Total (i.e., all trail-based recreation types) annual RRI was less in 2018 (0.037) than 2016
(0.047) and 2017 (0.046), and the total MN6HR was less (p < 0.05) in 2018 (0.556) than 2016
(0.944) and 2017 (1.084) (Table 2.2). Additionally, the annual RRI and MN6HR were least
during 2018 for all recreation types (Table 2.4). Evaluation of recreational intensity by month
indicated September had the greatest total RRI each year, and the overall MN6HR was greater (p
< 0.05) for September (1.597) than all other study months (May = 0.765, June = 0.647, July =
0.596, August = 0.828) during 2016–2018 (Table 2.2). Further evaluation of less recreational
intensity during 2018 demonstrated that May, June, July, and August total RRIs were less than
57
identical months in 2016 and 2017 (Table 2.2). Additionally, the total MN6HR for June (0.345),
July (0.328), and August (0.536) of 2018 were less (p < 0.05) than 2016 (i.e., June = 0.927, July
= 0.622, August = 0.962) and 2017 (i.e., June = 0.848, July = 1.038, August = 1.099).
Equestrian use had the greatest mean RRIs among recreation types during May, June,
September, and October, while hiking/foot-traffic was greatest during July and August (Figure
2.2). Notably, the RRI for equestrian use and hiking/foot-traffic were always greater than
mountain biking and ORV use which had the least RRIs, respectively for each month during
each year of our study. Further evaluation of recreation types during each month of each year
indicated September had the greatest RRIs for each recreation type, followed by May and
October for equestrian use and July and October for hiking/foot traffic (Table 2.4). Monthly
mountain biking RRIs were inconsistent among years apart from September having the greatest
RRI each year. We found similar patterns among recreation types for 2016 and 2017 with
noticeable differences in 2018. For example, the monthly RRI for hiking/foot-traffic and
mountain biking for each month during 2018 was less than 2016 and 2017 (Table 2.4). For
mountain biking, the MN6HR for all months in 2018 were less (p < 0.05) than 2016 and 2017.
Additionally, monthly 2018 RRIs for equestrian use were less than 2016 and 2017 in May, June,
and July, before increasing in August and September to the greatest RRIs for each type during
each respective month between study years (Table 2.4).
58
Table 2.2. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., RRI, MN6HR) captured by remote digital trail cameras on public lands within the
Pigeon River Country (PRC) State Forest and Atlanta State Forest (ASF) Management Units
during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Camera PRC ASF
Year/month hours events RRI1 MN6HR2 events RRI1 MN6HR2
2016 65,832 3,105 0.047 0.944 966 0.021 0.229
May3 3,552 229 0.064 0.921 31 0.072 0.621
June 10,248 392 0.038 0.927 68 0.010 0.208
July 11,280 368 0.033 0.622 75 0.010 0.197
August 12,936 350 0.027 0.962 91 0.012 0.090
September 13,680 1,205 0.088 1.401 400 0.037 0.278
October3 14,136 561 0.040 301 0.027
2017 105,504 4,878 0.046 1.084 2,294 0.021 0.264
May4 6,048 317 0.052 0.780 104 0.021 0.199
June 19,488 484 0.025 0.848 188 0.010 0.065
July 20,832 680 0.033 1.038 295 0.013 0.228
August 20,832 607 0.029 1.099 263 0.012 0.309
September 20,160 1,573 0.078 1.984 830 0.038 1.413
October4 18,144 1,217 0.067 614 0.033
2018 92,328 3,429 0.037 0.556 1,774 0.016 0.355
May5 3,816 157 0.041 0.623 68 0.015 0.312
June 21,672 340 0.016 0.345 211 0.008 0.280
July 22,608 358 0.016 0.328 220 0.008 0.234
August 22,728 572 0.025 0.536 285 0.010 0.200
September 21,504 2,002 0.093 1.401 990 0.038 1.374
Total 263,664 11,412 0.043 0.829 5,034 0.019 0.278
May3–5 13,416 703 0.052 0.765 203 0.021 0.338
June 51,408 1,216 0.024 0.647 467 0.009 0.156
July 54,720 1,406 0.026 0.596 590 0.010 0.219
August 56,496 1,529 0.027 0.828 639 0.011 0.177
September 55,344 4,780 0.086 1.597 2,220 0.038 0.814
October3–5 32,280 1,778 0.055 915 0.031
59
Table 2.2. (cont’d)
1
Number of recreation events per camera hour (i.e., RRI = Rec. events/Camera hours).
2
Mean number of recreation events during all 6-hour intervals for a given time period. Values
were obtained by back-transformation of estimated logmean number of events using a
generalized linear mixed model.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
Table 2.3. Independent variable F-values produced by generalized linear mixed models with
Poisson distributions to determine which variables had the most effect on detection of a
recreation event within a 6-hour period in the Michigan elk range (i.e., Atlanta [ASF] and Pigeon
River Country [PRC] State Forest management units) during 2016–2018.
Variables Df Sum sq. F-value
PRC
Recreation type1 2 2030.3 1015.1
Year2 2 44.1 22.0
Month3 4 2927.3 731.8
Day of week 6 1442.6 240.4
Time of day4 2 2275.9 1137.9
ASF
Recreation type1 3 1130.8 376.9
Year2 2 12.9 6.4
Month3 4 1178.1 294.5
Day of week 6 675.3 112.6
Time of day4 2 497.6 248.8
1
Equestrian use, hiking/foot-traffic, mountain biking, ORV use; Off-road vehicle use (ORV)
was not included in the PRC model due to insufficient sample size.
2
2016–2018
3
May–September
4
Six-hour intervals (i.e., 5:00–10:59, 11:00–16:59, 17:00–23:00)
60
Table 2.4. Trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a camera in the same
direction within a 5-minute period) and recreational intensity (i.e., RRI, MN6HR) for recreation types (i.e., equestrian use,
hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public lands within the Pigeon River
Country State Forest management unit during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Camera Equestrian use Hiking/foot-traffic Mountain-biking ORV use
Year/month hours events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2
2016 65,832 1,785 0.027 2.005 968 0.015 1.056 341 0.005 0.397 11 0.000
May3 3,552 181 0.051 4.406 30 0.008 0.631 11 0.003 0.281 7 0.002
June 10,248 210 0.020 1.514 126 0.012 1.146 55 0.005 0.460 1 0.000
July 11,280 125 0.011 0.665 175 0.016 1.080 66 0.006 0.335 2 0.000
August 12,936 117 0.009 1.358 174 0.013 1.352 59 0.005 0.486 0 0.000
September 13,680 846 0.062 5.377 255 0.019 1.242 103 0.008 0.472 1 0.000
October3 14,136 306 0.022 208 0.015 47 0.003 0 0.000
2017 105,504 2,712 0.026 2.370 1,797 0.017 1.659 338 0.003 0.324 31 0.000
May4 6,048 235 0.039 3.843 60 0.010 0.732 18 0.003 0.169 4 0.001
June 19,488 244 0.013 1.425 201 0.010 1.434 39 0.002 0.299 0 0.000
July 20,832 190 0.009 1.142 408 0.020 2.466 73 0.004 0.398 9 0.000
August 20,832 168 0.008 1.596 359 0.017 2.112 77 0.004 0.394 3 0.000
September 20,160 1,070 0.053 7.490 414 0.021 2.299 80 0.004 0.454 9 0.000
October4 18,144 805 0.044 355 0.020 51 0.003 6 0.000
2018 92,328 2,208 0.024 1.917 1,090 0.012 1.005 99 0.001 0.089 32 0.000
May5 3,816 135 0.035 4.842 19 0.005 0.691 2 0.001 0.072 1 0.000
June 21,672 106 0.005 0.915 200 0.009 0.689 34 0.002 0.065 0 0.000
July 22,608 72 0.003 0.570 256 0.011 0.922 16 0.001 0.067 14 0.001
August 22,728 331 0.015 1.229 222 0.010 1.218 11 0.000 0.103 8 0.000
September 21,504 1,564 0.073 8.349 393 0.018 1.920 36 0.002 0.172 9 0.000
61
Table 2.4. (cont’d)
Total 263,664 6,705 0.025 2.088 3,855 0.015 1.208 778 0.003 0.226 74 0.000
May3–5 13,416 551 0.041 4.344 109 0.008 0.683 31 0.002 0.151 12 0.001
June 51,408 560 0.011 1.254 527 0.010 1.042 128 0.002 0.207 1 0.000
July 54,720 387 0.007 0.756 839 0.015 1.349 155 0.003 0.208 25 0.000
August 56,496 616 0.011 1.386 755 0.013 1.515 147 0.003 0.270 11 0.000
September 55,344 3,480 0.063 6.953 1,062 0.019 1.764 219 0.004 0.332 19 0.000
October3-5 32,280 1,111 0.034 563 0.017 98 0.003 6 0.000
1
Number of recreation events per camera hour (i.e., RRI = Rec. events/Camera hours).
2
Mean number of recreation events during all 6-hour intervals for a given time period. Values were obtained by back-transformation
of estimated logmean number of events using a generalized linear mixed model. Off-road vehicle use (ORV) is prohibited in the
PRC and was not included due to insufficient sample size.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
62
Figure 2.2. Mean relative recreational intensity (i.e. RRI = number of recreation events per
camera hour) of equestrian use, hiking/foot-traffic, mountain biking, and ORV use during peak
summer–fall recreational periods (i.e., May–October), 2016–2018. Data were captured by remote
digital trail cameras in the northern lower peninsula of Michigan: (A) Pigeon River Country
State Forest management unit, (B) Atlanta State Forest management unit. Variation depicted by
error bars is SD of mean.
63
Evaluation of recreational intensity by day of week in the PRC
Evaluation of recreational intensity by day of week revealed Saturday, Sunday, and Friday
had greater RRIs (Sat = 0.080, Sun = 0.060, Fri = 0.049) and MN6HR (p < 0.05; Sat = 1.937,
Sun = 1.297, Fri = 1.170) among days of the week during 2016–2018, respectively (Table 2.5).
Notably, Saturday had greater RRI and MN6HR (p < 0.05) among days of the week for all years
and months. Mean RRI’s for each day of the week from 2016–2018 had similar distributions
among recreation types, with Saturday, Sunday, and Friday having the highest mean intensity for
all recreation types, respectively (Figure 2.3). Additionally, we found greater (p < 0.05) MN6HR
during Friday, Saturday, and Sunday than other days for each recreation type (Table 2.6). Further
evaluation for differences among years revealed MN6HR (p < 0.05) for Monday, Wednesday,
Thursday, and Friday was least during 2018 for equestrian use and hiking/foot-traffic, and all
days of the week were least during 2018 for mountain biking.
Evaluation of recreational intensity by time of day in the PRC
Evaluation of recreational intensity by time of day demonstrated that 60.8% (6,941 events) of
all recreation events during 2016–2018 occurred between 11:00–16:59, 22.8% (2,596 events)
occurred between 17:00–23:00, and 16.4% (1,870 events) occurred between 5:00–10:59 (Table
2.7). Only 5 events (i.e., < 0.001%) occurred from 23:01–4:59 during 2016–2018, which were
censored from GLMM analyses. Additionally, the MN6HR was greater (p < 0.05) during 11:00–
16:59 than 5:00–10:59 and 17:00–23:00 during all years, months, and days. Peak times of day
for recreation types in the PRC during 2016–2018 was 10:00–15:59 and 18:00–19:59 for
equestrian use (i.e., 75.7% of use), 10:00–16:59 for hiking/foot-traffic (i.e., 73.1% of use),
11:00–17:59 for mountain biking (i.e., 75.6% of use), and 13:00–19:59 for ORV use (77% of
use) (Figure 2.4). The majority of events for each recreation type occurred from 11:00–16:59
64
(Table 2.8), during which the MN6HR was greater (p < 0.05) for each type during all years and
months.
No differences (p < 0.05) were detected for time of day use among years or months for any
recreation type. However, there were differences in time of day use among days of the week for
each recreation type. For equestrian use, mean time of day use occurred earlier in the day (X2 =
172.94, df = 6, p < 0.001) during days at the beginning of the week (i.e., Monday = 13:00 ± 0:08
[SE], Sunday = 13:06 ± 0:06, Tuesday = 13:30 ± 0:09, Wednesday = 14:00 ± 0:09), and later in
the day during days at the end of the week (i.e., Thursday = 14:24 ± 0:07, Friday = 14:24 ± 0:06,
Saturday = 14:00 ± 0:05). For hiking/foot-traffic, differences in time of day use among days of
the week were primarily between weekend days (i.e., Fri–Sat, Fri–Sun, Sat–Sun) (X2 = 43.53, df
= 6, p < 0.001), with mean time of day use occurring latest on Friday (14:12 ± 0:09) and earlier
on Saturday (13:30 ± 0:05) and Sunday (13:00 ± 0:06). The mean time of day for mountain
biking was earlier (X2 = 49.93, df = 6, p < 0.001) on Saturday (12:54 ± 0:11) than all other days
of the week (Monday = 13:18 ± 0:17, Sunday = 14:00 ± 0:13, Tuesday = 14:12 ± 0:26,
Wednesday = 14:24 ± 0:23, Thursday = 14:42 ± 0:20, Friday = 14:48 ± 0:17).
Evaluation of recreational intensity during peak days of use in the PRC
Evaluation of the outliers (i.e., n = 25 days) for daily RRI values during 2016-2018
demonstrated: (a) the distribution of days with the greatest RRIs was relatively even among years
with 7 occurring in 2016, 9 in 2017, and 9 in 2018; (b) 20 of 25 days with greatest RRIs were
during September, including the 6 greatest days; (c) 21 of 25 days with the greatest RRIs were
Friday-Sunday; (d) 7 of the 25 days with the greatest RRIs were during holiday weekends (i.e.,
2-Memorial Day [i.e., late May], 5-Labor Day [i.e., early September]); and (e) equestrian use
65
accounted for 71.8% (1,927 of 2,684 events) of the recreational activity during days with greatest
RRI.
Characteristics of group size for recreation types in the PRC
Mean group sizes for recreation types were 3.14 ± 0.027 (SE) for equestrian use, 2.37 ±
0.044 for hiking/foot-traffic, 1.65 ± 0.041 for mountain biking, and 1.30 ± 0.069 for ORV use
during 2016-2018 (Table 2.9). No significant differences in mean group sizes occurred among
years or months for hiking/foot-traffic, mountain biking, and ORV use. However, mean group
size for equestrian use was greater (X2 = 40.61, df = 2, p < 0.001) in 2016 (3.431, ± 0.057) than
2017 (3.013, ± 0.038) and 2018 (3.067, ± 0.046), and greater (X2 = 75.37, df = 5, p < 0.001)
during May (3.48 ± 0.089) and September (3.22 ± 0.036) (i.e., months of greatest recreational
intensity) than June (3.12 ± 0.124), July (3.06 ± 0.093), October (3.05 ± 0.072), and August
(2.64 ± 0.053). Additionally, larger group sizes (X2 = 62.51, df = 6, p < 0.001) occurred during
weekends (i.e., days of greatest recreational intensity; Friday = 3.42 ± 0.081, Saturday = 3.38 ±
0.057, Sunday = 3.10 ± 0.056) than other days of the week (i.e., Thursday = 3.04 ± 0.070,
Monday = 2.90 ± 0.072, Tuesday = 2.83 ± 0.063, Wednesday = 2.68 ± 0.066).
While group sizes were relatively small and consistent for each recreation type from 2016-
2018, we observed notable outliers (i.e., n = 831 events) in median group size representing
substantially larger group sizes for each recreation type (i.e., equestrian use ≥ 8, hiking/foot-
traffic ≥ 4, mountain biking ≥ 4). For example, the largest group sizes were 26 for equestrian use,
41 for hiking/foot-traffic, and 15 for mountain biking. The largest groups for hiking/foot-traffic
appeared to be organized recreation events for youth accompanied by adults (e.g., school groups,
clubs). September had the highest number (n = 312) of such events among all months.
66
Additionally, Friday through Sunday accounted for 71.81% (n = 596) of such events among days
of the week. Lastly, 65% (n = 540) of such events occurred from 11:00-16:59.
67
Table 2.5. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5-minute period) and recreational
intensity (i.e., RRI, MN6HR) by days of the week captured by remote digital trail cameras on
public lands within the Pigeon River Country (PRC) and Atlanta (ASF) State Forest
Management Units during peak summer–fall recreational periods (i.e., May–October), 2016–
2018.
Camera PRC ASF
Year/day hours events RRI1 MN6HR2 events RRI1 MN6HR2
2016a 65,832 3,105 0.047 0.944 966 0.021 0.229
Monday 9,600 304 0.032 0.842 79 0.012 0.181
Tuesday 9,168 230 0.025 0.450 79 0.013 0.127
Wednesday 9,144 277 0.030 0.588 98 0.016 0.151
Thursday 9,216 410 0.044 0.803 118 0.019 0.136
Friday 9,504 548 0.058 1.393 198 0.031 0.266
Saturday 9,576 746 0.078 1.953 226 0.036 0.596
Sunday 9,624 590 0.061 1.370 168 0.026 0.447
2017b 105,504 4,878 0.046 1.084 2,294 0.021 0.264
Monday 14,952 434 0.029 0.816 184 0.012 0.161
Tuesday 15,168 408 0.027 0.635 181 0.012 0.148
Wednesday 15,168 343 0.023 0.558 189 0.012 0.141
Thursday 15,168 508 0.033 0.766 259 0.016 0.240
Friday 15,240 790 0.052 1.467 387 0.024 0.291
Saturday 15,240 1,411 0.093 2.898 691 0.043 0.776
Sunday 14,568 984 0.068 1.873 403 0.026 0.490
2018c 92,328 3,429 0.037 0.556 1,774 0.016 0.355
Monday 12,768 329 0.026 0.456 175 0.026 0.295
Tuesday 12,912 342 0.026 0.369 145 0.026 0.228
Wednesday 12,960 293 0.023 0.315 140 0.023 0.200
Thursday 13,344 360 0.027 0.362 211 0.027 0.287
Friday 13,464 532 0.040 0.783 322 0.040 0.392
Saturday 13,488 902 0.067 1.284 443 0.067 0.738
Sunday 13,392 671 0.050 0.849 338 0.050 0.638
68
Table 2.5. (cont’d)
Total 263,664 11,412 0.043 0.829 5,034 0.019 0.278
Monday 37,320 1,067 0.029 0.679 438 0.012 0.205
Tuesday 37,248 980 0.026 0.473 405 0.011 0.162
Wednesday 37,272 913 0.024 0.469 427 0.011 0.162
Thursday 37,728 1,278 0.034 0.606 588 0.016 0.211
Friday 38,208 1,870 0.049 1.170 907 0.024 0.312
Saturday 38,304 3,059 0.080 1.937 1,360 0.035 0.699
Sunday 37,584 2,245 0.060 1.297 909 0.024 0.519
1
Number of recreation events per camera hour (i.e., RRI = Rec. events/Camera hours).
2
Mean number of recreation events during all 6-hour intervals for a given time period. Values
were obtained by back-transformation of estimated logmean number of events using a
generalized linear mixed model.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
69
Figure 2.3. Mean relative recreational intensity (i.e. RRI = number of recreation events per
camera hour) of equestrian use, hiking/foot-traffic, mountain biking, and ORV use by day of
week during peak summer–fall recreational periods (i.e., May–October), 2016–2018. Data were
captured by remote digital trail cameras in the northern lower peninsula of Michigan: (A) Pigeon
River Country state forest management unit, (B) Atlanta state forest management unit. Variation
depicted by error bars is SD of mean.
70
Table 2.6. Trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a camera in the same
direction within a 5 minute period) and recreational intensity (i.e., RRI, MN6HR) by days of the week for recreation types (i.e.,
equestrian use, hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public lands within the
Pigeon River Country State Forest management unit during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Camera Equestrian use Hiking/foot-traffic Mountain-biking ORV use
Year/day hours events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2
2016a 65,832 1,785 0.027 2.005 968 0.015 1.056 341 0.005 0.397 11 0.000
Monday 9,600 151 0.016 1.597 119 0.012 0.911 34 0.004 0.411 0 0.000
Tuesday 9,168 140 0.015 1.145 73 0.008 0.511 14 0.002 0.156 3 0.000
Wednesday 9,144 153 0.017 1.314 92 0.010 0.647 31 0.003 0.240 1 0.000
Thursday 9,216 275 0.030 2.008 86 0.009 0.886 49 0.005 0.291 0 0.000
Friday 9,504 372 0.039 3.269 125 0.013 1.343 51 0.005 0.616 0 0.000
Saturday 9,576 371 0.039 3.073 269 0.028 2.367 102 0.011 1.024 4 0.000
Sunday 9,624 323 0.034 2.684 204 0.021 1.722 60 0.006 0.556 3 0.000
2017b 105,504 2,712 0.026 2.370 1,797 0.017 1.659 338 0.003 0.324 31 0.000
Monday 14,952 223 0.015 1.592 166 0.011 1.207 41 0.003 0.283 4 0.000
Tuesday 15,168 232 0.015 1.662 147 0.010 0.987 27 0.002 0.156 2 0.000
Wednesday 15,168 184 0.012 1.283 138 0.009 0.840 21 0.001 0.161 0 0.000
Thursday 15,168 318 0.021 1.971 172 0.011 1.155 16 0.001 0.197 2 0.000
Friday 15,240 486 0.032 3.541 243 0.016 1.933 56 0.004 0.461 5 0.000
Saturday 15,240 711 0.047 4.692 573 0.038 4.804 114 0.007 1.080 13 0.001
Sunday 14,568 558 0.038 3.778 358 0.025 3.221 63 0.004 0.540 5 0.000
71
Table 2.6. (cont’d)
2018c 92,328 2,208 0.024 1.917 1,090 0.012 1.005 99 0.001 0.089 32 0.000
Monday 12,768 221 0.017 1.405 90 0.007 0.798 9 0.001 0.085 9 0.001
Tuesday 12,912 252 0.020 1.525 86 0.007 0.678 3 0.000 0.049 1 0.000
Wednesday 12,960 200 0.015 1.141 86 0.007 0.560 5 0.000 0.049 2 0.000
Thursday 13,344 252 0.019 1.470 99 0.007 0.645 6 0.000 0.050 3 0.000
Friday 13,464 350 0.026 2.984 163 0.012 1.221 19 0.001 0.132 0 0.000
Saturday 13,488 515 0.038 3.281 342 0.025 2.517 34 0.003 0.256 11 0.001
Sunday 13,392 418 0.031 2.704 224 0.017 1.727 23 0.002 0.131 6 0.000
Total 263,664 6,705 0.025 2.088 3,855 0.015 1.208 778 0.003 0.226 74 0.000
Monday 37,320 595 0.016 1.529 375 0.010 0.957 84 0.002 0.214 13 0.000
Tuesday 37,248 624 0.017 1.426 306 0.008 0.699 44 0.001 0.106 6 0.000
Wednesday 37,272 537 0.014 1.244 316 0.008 0.673 57 0.002 0.124 3 0.000
Thursday 37,728 845 0.022 1.799 357 0.009 0.871 71 0.002 0.142 5 0.000
Friday 38,208 1,208 0.032 3.257 531 0.014 1.469 126 0.003 0.335 5 0.000
Saturday 38,304 1,597 0.042 3.617 1,184 0.031 3.059 250 0.007 0.657 28 0.001
Sunday 37,584 1,299 0.035 3.015 786 0.021 2.124 146 0.004 0.340 14 0.000
1
Number of recreation events per camera hour (i.e., RRI = Rec. events/Camera hours).
2
Mean number of recreation events during all 6-hour intervals for a given time period. Values were obtained by back-transformation
of estimated logmean number of events using a generalized linear mixed model. Off-road vehicle use (ORV) is prohibited in the
PRC and was not included due to insufficient sample size.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
72
Table 2.7. Trail-based recreation events (i.e., any number of individuals of the same recreation
type passing by a camera in the same direction within a 5 minute period) and recreational
intensity (i.e., MN6HR) by time of day (i.e., 3, 6-hour intervals within 5:00–23:00) captured by
remote digital trail cameras on public lands within the Pigeon River Country (PRC) State Forest
and Atlanta State Forest (ASF) Management Units during peak summer–fall recreational periods
(i.e., May–October), 2016–2018.
PRC ASF
Year/time period events MN6HR1 events MN6HR1
2016a 3,102 0.944 964 0.229
05:00–10:59 472 0.549 167 0.119
11:00–16:59 1,931 2.074 424 0.388
17:00–23:00 699 0.739 373 0.261
2017b 4,877 1.084 2,279 0.264
05:00–10:59 775 0.656 348 0.130
11:00–16:59 3,024 2.325 1,135 0.537
17:00–23:00 1,078 0.836 796 0.264
2018c 3,428 0.556 1,767 0.355
05:00–10:59 623 0.363 324 0.209
11:00–16:59 1,986 1.137 814 0.655
17:00–23:00 819 0.416 629 0.327
Total 11,407 0.829 5,010 0.278
05:00–10:59 1,870 0.507 839 0.148
11:00–16:59 6,941 1.763 2,373 0.515
17:00–23:00 2,596 0.636 1,798 0.282
1
Mean number of recreation events during all 6-hour intervals for a given time period. Values
were obtained by back-transformation of estimated logmean number of events using a
generalized linear mixed model.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
73
Figure 2.4. Human recreation use (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV
use) events (i.e., any number of individuals of the same recreation type passing by a camera in
the same direction within a 5 minute period) by hourly time intervals during peak summer–fall
recreational periods (i.e., May–October), 2016–2018. Data were captured by remote digital trail
cameras in the northern lower peninsula of Michigan: (A) Pigeon River Country State Forest
management unit, (B) Atlanta State Forest management unit.
74
Table 2.8. Trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a camera in the same
direction within a 5-minute period) and recreational intensity (i.e., MN6HR) by time of day for recreation types (i.e., equestrian use,
hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public lands within the Pigeon River
Country State Forest management unit during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Equestrian use Hiking/foot-traffic Mountain-biking ORV use
Year/time period events MN6HR1 events MN6HR1 events MN6HR1 events MN6HR1
2016a 1,785 2.005 965 1.056 341 0.397 11
5:00–10:59 265 1.114 161 0.665 45 0.223 1
11:00–16:59 1,082 3.698 626 2.473 217 0.976 6
17:00–23:00 438 1.955 178 0.716 79 0.289 4
2017a 2,712 2.370 1,796 1.659 338 0.324 31
5:00–10:59 438 1.371 292 1.087 45 0.189 0
11:00–16:59 1,554 4.266 1,209 3.792 235 0.777 26
17:00–23:00 720 2.275 295 1.108 58 0.232 5
2018a 2,208 1.917 1,089 0.324 99 0.089 32
5:00–10:59 403 1.197 202 0.711 13 0.056 5
11:00–16:59 1,221 3.292 683 2.192 72 0.203 10
17:00–23:00 584 1.788 204 0.652 14 0.062 17
Total 6,705 2.088 3,850 1.208 778 0.226 74
5:00–10:59 1,106 1.223 655 0.801 103 0.133 6
11:00–16:59 3,857 3.731 2,518 2.739 524 0.536 42
17:00–23:00 1,742 1.996 677 0.803 151 0.161 26
1
Mean number of recreation events during all 6-hour intervals for a given time period. Values were obtained by back-transformation
of estimated logmean number of events using a generalized linear mixed model. Off-road vehicle use (ORV) is prohibited in the
PRC and was not included due to insufficient sample size.
a
Trail cameras operated from 20–May to 31–October, 2016; 16–May to 28–October, 2017; 24–May to 30–September, 2018.
75
Table 2.9. Mean group sizes of trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a
camera in the same direction within a 5 minute period) for recreation types (i.e., equestrian use, hiking/foot-traffic, mountain biking,
ORV use) captured by remote digital trail cameras on public lands in the Pigeon River Country State Forest management unit during
peak summer–fall recreational periods (i.e., May–October), 2016–2018. Trail cameras operated from 20–May to 31–October, 2016;
16–May to 28–October, 2017; 24–May to 30–September, 2018.
Temporal Equestrian use Hiking/foot-traffic Mountain-biking ORV use
category events mean SE events mean SE events mean SE events mean SE
Overall 6,705 3.14 0.027 3,855 2.37 0.044 778 1.65 0.041 74 1.30 0.069
2016 1,785 3.43 0.057 968 2.50 0.110 341 1.75 0.076 11 1.00 0.000
2017 2,712 3.01 0.038 1,797 2.34 0.059 338 1.55 0.047 31 1.35 0.989
2018 2,208 3.07 0.047 1,090 2.30 0.072 99 1.64 0.100 32 1.34 0.124
May 551 3.48 0.089 109 2.19 0.141 31 1.74 0.113 12 1.08 0.083
June 560 3.12 0.124 527 2.08 0.066 128 1.93 0.167 1 1.00 0.000
July 387 3.06 0.093 839 2.83 0.014 155 1.57 0.086 25 1.16 0.095
August 616 2.64 0.053 755 2.60 0.011 147 1.62 0.074 11 1.36 0.152
September 3,480 3.22 0.036 1,062 2.15 0.060 219 1.61 0.063 19 1.58 0.192
October 1,111 3.05 0.072 563 2.10 0.078 98 1.51 0.087 6 1.33 0.211
Monday 595 2.90 0.072 375 2.14 0.137 84 1.56 0.080 13 1.23 0.166
Tuesday 624 2.83 0.063 306 2.48 0.178 44 1.48 0.105 6 1.17 0.167
Wednesday 537 2.68 0.066 316 2.47 0.176 57 1.33 0.068 3 1.00 0.000
Thursday 845 3.04 0.071 357 2.36 0.190 71 2.03 0.304 5 1.80 0.583
Friday 1,208 3.42 0.081 531 2.52 0.153 126 1.50 0.076 5 1.00 0.000
Saturday 1,597 3.38 0.057 1,184 2.44 0.064 250 1.70 0.054 28 1.36 0.106
Sunday 1,299 3.10 0.056 786 2.20 0.071 146 1.74 0.096 14 1.29 0.125
5:00–10:59 1,106 3.19 0.067 655 2.21 0.116 103 1.73 0.113 6 1.00 0.000
11:00–16:59 3,857 3.09 0.036 2,518 2.45 0.056 524 1.61 0.041 42 1.38 0.108
17:00–23:00 1,742 3.22 0.049 677 2.25 0.080 151 1.74 0.140 26 0.23 0.084
76
Relative Intensities, Temporal Characteristics, and Group Sizes of Recreational Users in
the ASF
Evaluation of recreational intensity in the ASF
We captured 5,034 recreation events during 266,184 hours of monitoring in the ASF during
2016–2018 (Table 2.2). We censored 21 events due to inability to determine recreation type. The
overall RRI was 0.019 and the overall MN6HR was 0.278. Recreation type had the most effect
on detection of a recreation event within a 6-hour period followed by month, time of day, day of
week, and year (Table 2.3). Off-road vehicle use was the most frequently detected type of trail-
based recreation in ASF during 2016–2018 (i.e., RRI = 0.010, MN6HR = 0.988), followed by
hiking/foot-traffic (i.e., RRI = 0.005, MN6HR = 0.635), equestrian use (i.e., RRI = 0.004,
MN6HR = 0.294), and mountain biking (i.e., RRI = 0.001, MN6HR = 0.032) (Table 2.10).
Evaluation of recreational intensity by year and month in the ASF
Annual total RRI was less during 2018 (0.016) than 2016 (0.021) and 2017 (0.021) (Table
2.2). However, no differences (p < 0.05) in MN6HR were detected among years for total trail-
based recreational activity or for any recreation types. Off-road vehicle use had the greatest RRI
and MN6HR each year, followed by hiking/foot-traffic, equestrian use, and mountain biking,
respectively (Table 2.10). Additionally, we found differences (p < 0.05) in MN6HR among all
recreation types each year. May had the greatest RRI in 2016 (0.072) and September had the
greatest RRIs in 2017 (0.038) and 2018 (0.038) (Table 2.2). However, there were only 432
camera hours for 3 days (i.e., 29–31) of a holiday weekend in the ASF during May 2016, which
likely inflated the RRI. Additionally, September had the second greatest RRI in 2016 (0.037)
which was consistent with 2017 and 2018 (Table 2.2). Moreover, the MN6HR was greater (p <
77
0.05) in September (0.814) than all other months (May = 0.338, June = 0.156, July = 0.219,
August = 0.177) during 2016–2018.
Off-road vehicle use had the greatest mean RRIs among recreation types during all months
(Figure 2.10). While hiking/foot-traffic had greater monthly mean RRIs than equestrian use from
June through August, and October, equestrian use had greater mean RRIs during May and
September. Mountain biking had the least mean RRIs each month. We observed similar trends
for total monthly MN6HR with ORV use being greatest (p < 0.05) and mountain biking being
least (p < 0.05) for all months (Table 2.10). While hiking/foot-traffic had greater total monthly
MN6HR than equestrian use during all months except for September, we only found differences
(p < 0.05) during July and August. We found similar trends among months for each recreation
type, with September having the greatest mean RRIs for hiking/foot-traffic (0.008, SD = 0.0008),
mountain biking (0.001, SD = 0.0003), and ORV use (0.019, SD = 0.004) during 2016–2018
(Figure 2.2). For equestrian use, May had the greatest mean RRI (0.012, SD = 0.016) among
months during 2016–2018, however, the RRI for May 2016 was likely inflated due to the short
monitoring period and timing. As a result, the SD (0.016) was larger than the estimated mean
RRI (0.012) for equestrian use in May 2016. September had the second greatest mean RRI
(0.009, SD = 0.002) for equestrian use during 2016–2018, which is consistent with the trend of
greater recreational use during September for other types. Additionally, the MN6HR was
significantly higher during September for equestrian use (1.507), hiking/foot-traffic (1.364), and
ORV use (2.630) during 2016–2018 (Table 2.10).
Evaluation of recreational intensity by day of week in the ASF
Evaluation of recreational intensity by day of week revealed Saturday, Sunday, and Friday
had the greatest total RRIs (Table 2.5) and total MN6HR (Sat = 0.699, Sun = 0.519, Fri = 0.312)
78
during 2016–2018, respectively. Saturday had the greatest RRI among days of the week for all
years and months. Additionally, the total MN6HR for Saturday was greater (p < 0.05) than
weekdays during all years and months, with only Sunday having no difference (p < 0.05) from
Saturday for 2016, 2018, and May, July, and August during 2016–2018.
Mean RRI’s for each day of the week during 2016–2018 had similar patterns among
equestrian use, hiking/foot-traffic, and ORV use, with each having the greatest daily RRIs on
Saturday, Friday, and Sunday, respectively (Figure 2.3). Mountain biking had its greatest daily
RRIs on Saturday, Sunday, and Thursday, respectively. Additionally, all recreation types had
greater (p < 0.05) MN6HR on Saturday than all other days except for Sunday (Table 2.11).
Evaluation of recreational intensity by time of day in the ASF
Evaluation of recreational intensity by time of day revealed that 47.1% of all events (2,373)
occurred between 11:00–16:59, 35.7% (i.e., 1,798 events) occurred between 17:00–23:00, and
16.7% (i.e., 839 events) between 5:00–10:59 (Table 2.7). Notably, we censored 24 events (i.e., <
0.005%) from our GLMM analysis that occurred from 23:01–4:59, during 2016–2018.
Additionally, the MN6HR was greater (p < 0.05) between 11:00–16:59 from 5:00–10:59 and
17:00–23:00 during all study years, months, and days of week. Peak times of day for recreation
types during 2016–2018 were 10:00–14:59 and 17:00–20:59 for equestrian use (i.e., 80.7% of
use), 8:00–10:59, 12:00–16:59, and 18:00–20:59 for hiking/foot-traffic (i.e., 78.2% of use),
11:00–16:59 for mountain biking (i.e., 77.6% of use), and 12:00–19:59 for ORV use (70.6% of
use) (Figure 2.4). Similar to the PRC, the MN6HR was greater (p < 0.05) during 11:00–16:59 for
each recreation type during all study years and months (Table 2.12).
No differences (p < 0.05) were detected for time of day use among years for equestrian use
and hiking/foot-traffic. For mountain biking, mean time of day use was later (X2 = 12.23, df = 2,
79
p = 0.002) during 2017 (14:06 ± 0:24 [SE]) than 2016 (12:48 ± 0:26) and 2018 (12:24 ± 0:16).
For ORV use, mean time of day use was later (X2 = 8.39, df = 2, p = 0.015) during 2016 (15:30 ±
0:11) than 2017 (15:06 ± 0:06) and 2018 (14:54 ± 0:08). We found differences in time of day use
among study months for equestrian use, hiking/foot-traffic, and ORV use. For equestrian use,
mean time of day use was earliest (X2 = 17.98, df = 5, p = 0.003) during July (12:48 ± 0:34) and
latest during June (15:24 ± 0:50), while being relatively consistent (i.e., means were 14:00–
14:48) among other study months. For hiking/foot-traffic, mean time of day use was earlier (X2 =
47.68, df = 5, p < 0.001) during May (12:42 ± 0:25), October (13:12 ± 0:12), and September
(13:00 ± 0:16) than other study months (i.e., July = 14:12 ± 0:22, August = 14:54 ± 0:18, June =
15:06 ± 0:21). For ORV use, mean time of day use was earliest (X2 = 49.07, df = 5, p < 0.001)
during October (14:24 ± 0:09), May (14:48 ± 0:21), and September (14:54 ± 0:06), and latest
during July (15:54 ± 0:11), June (15:42 ± 0:14), and August (15:24 ± 0:12). Additionally, we
found an earlier mean time of day use (X2 = 46.99, df = 6, p < 0.001) for ORV use during
Sunday (14:18 ± 0:10) from all other days of the week (i.e., means were 15:18–15:54).
Evaluation of recreational intensity during peak days of use in the ASF
Evaluation of the outliers (i.e., n = 41 days) for daily RRI values during 2016–2018
demonstrated: (a) there were more days (20) with the greatest RRIs during 2017 than 2016 (15)
and 2018 (6); (b) 21 of 41 days with greatest RRIs were during September; (c) 34 of 41 days
with the greatest RRIs occurred from Friday–Sunday; (d) 15 of the 41 days with the greatest
RRIs were during holiday weekends (i.e., 4-Memorial Day, 2-Independence Day [4th of July], 5-
Labor Day); and (e) ORV use accounted for 52.9% (937 of 1,770 events) of recreational activity
during days with greatest RRIs.
80
Table 2.10. Trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a camera in the same
direction within a 5-minute period) and recreational intensity (i.e., RRI, MN6HR) for recreation types (i.e., equestrian use,
hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public lands within the Atlanta State Forest
management unit during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Camera Equestrian use Hiking/foot-traffic Mountain-biking ORV use
Year/month hours events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2
2016 44,472 264 0.006 0.281 265 0.006 0.433 33 0.001 0.032 404 0.009 0.710
May3 432 13 0.030 1.012 2 0.005 0.930 0 0.000 0.100 16 0.037 1.579
June 6,816 4 0.001 0.107 10 0.001 0.374 9 0.001 0.079 45 0.007 0.585
July 7,440 22 0.003 0.180 14 0.002 0.431 0 0.000 0.028 39 0.005 0.700
August 7,824 11 0.001 0.152 30 0.004 0.263 1 0.000 0.005 49 0.006 0.357
September 10,800 112 0.010 0.596 92 0.009 0.384 16 0.001 0.033 180 0.017 0.781
October3 11,160 102 0.009 117 0.010 7 0.001 75 0.007
2017 109,872 324 0.003 0.219 586 0.005 0.755 50 0.000 0.025 1,334 0.012 1.186
May4 5,064 20 0.004 0.219 40 0.008 0.452 2 0.000 0.022 42 0.008 0.734
June 19,632 3 0.000 0.023 71 0.004 0.177 10 0.001 0.017 104 0.005 0.265
July 22,440 15 0.001 0.141 114 0.005 0.756 6 0.000 0.021 160 0.007 1.176
August 22,320 20 0.001 0.351 71 0.003 1.365 3 0.000 0.011 169 0.008 1.774
September 21,600 148 0.007 2.051 159 0.007 2.968 20 0.001 0.113 503 0.023 5.776
October4 18,816 118 0.006 131 0.007 9 0.000 356 0.019
2018 111,840 381 0.003 0.414 459 0.004 0.783 64 0.001 0.043 870 0.008 1.146
May5 4,392 5 0.001 0.484 11 0.003 0.547 7 0.002 0.043 45 0.010 0.828
June 27,000 23 0.001 0.138 59 0.002 0.591 20 0.001 0.092 109 0.004 0.824
July 27,120 17 0.001 0.204 55 0.002 0.601 9 0.000 0.028 139 0.005 0.870
August 27,432 60 0.002 0.320 102 0.004 0.682 1 0.000 0.009 122 0.004 0.825
September 25,896 276 0.011 2.803 232 0.009 2.226 27 0.001 0.142 455 0.018 4.034
81
Table 2.10. (cont’d)
Total 266,184 969 0.004 0.294 1,310 0.005 0.635 147 0.001 0.032 2,608 0.010 0.988
May3–5 9,888 38 0.004 0.475 53 0.005 0.613 9 0.001 0.045 103 0.010 0.986
June 53,448 30 0.001 0.069 140 0.003 0.340 39 0.001 0.049 258 0.005 0.503
July 57,000 54 0.001 0.173 183 0.003 0.581 15 0.000 0.026 338 0.006 0.895
August 57,576 91 0.002 0.257 203 0.004 0.626 5 0.000 0.008 340 0.006 0.805
September 58,296 536 0.009 1.507 483 0.008 1.364 63 0.001 0.081 1,138 0.020 2.630
October3–5 29,976 220 0.007 248 0.008 16 0.001 431 0.014
1
Number of recreation events per camera hour (i.e., RRI = Rec. events/Camera hours).
2
Mean number of recreation events during all 6-hour intervals for a given time period. Values were obtained by back-transformation
of estimated logmean number of events using a generalized linear mixed model.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
82
Table 2.11. Trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a camera in the same
direction within a 5 minute period) and recreational intensity (i.e., RRI, MN6HR) by days of the week for recreation types (i.e.,
equestrian use, hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail cameras on public lands within the
Atlanta State Forest management unit during peak summer–fall recreational periods (i.e., May–October), 2016–2018.
Camera Equestrian use Hiking/foot-traffic Mountain-biking ORV use
Year/day hours events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2 events RRI1 MN6HR2
2016a 44,472 264 0.006 0.281 265 0.006 0.433 33 0.001 0.032 404 0.009 0.710
Monday 6,576 19 0.003 0.333 25 0.004 0.411 1 0.000 0.013 34 0.005 0.579
Tuesday 6,216 24 0.004 0.224 19 0.003 0.235 5 0.001 0.014 31 0.005 0.351
Wednesday 6,216 35 0.006 0.197 32 0.005 0.290 3 0.000 0.019 28 0.005 0.476
Thursday 6,216 34 0.005 0.142 42 0.007 0.253 0 0.000 0.026 42 0.007 0.362
Friday 6,336 57 0.009 0.342 56 0.009 0.625 0 0.000 0.021 85 0.013 1.097
Saturday 6,360 59 0.009 0.517 53 0.008 0.916 14 0.002 0.149 100 0.016 1.790
Sunday 6,552 36 0.005 0.378 38 0.006 0.702 10 0.002 0.114 84 0.013 1.319
2017b 109,872 324 0.003 0.219 586 0.005 0.755 50 0.000 0.025 1,334 0.012 1.186
Monday 15,480 43 0.003 0.201 51 0.003 0.556 1 0.000 0.008 89 0.006 0.751
Tuesday 15,720 31 0.002 0.176 45 0.003 0.413 1 0.000 0.011 104 0.007 0.592
Wednesday 15,720 37 0.002 0.124 43 0.003 0.409 1 0.000 0.012 108 0.007 0.643
Thursday 15,720 30 0.002 0.170 79 0.005 0.678 8 0.001 0.031 142 0.009 0.929
Friday 15,936 41 0.003 0.253 85 0.005 1.038 5 0.000 0.016 256 0.016 1.747
Saturday 16,008 84 0.005 0.455 171 0.011 1.808 19 0.001 0.130 417 0.026 3.386
Sunday 15,288 58 0.004 0.280 112 0.007 1.167 15 0.001 0.084 218 0.014 2.100
83
Table 2.11. (cont’d)
2018c 111,840 381 0.003 0.414 459 0.004 0.783 64 0.001 0.043 870 0.008 1.146
Monday 15,408 49 0.004 0.516 60 0.004 0.783 2 0.000 0.019 64 0.004 0.985
Tuesday 15,888 62 0.002 0.381 38 0.002 0.491 1 0.000 0.022 44 0.003 0.655
Wednesday 15,912 29 0.003 0.248 40 0.003 0.449 6 0.000 0.022 65 0.004 0.656
Thursday 15,888 53 0.003 0.286 45 0.003 0.627 10 0.001 0.047 103 0.006 0.800
Friday 15,224 62 0.006 0.479 93 0.006 1.076 4 0.000 0.027 163 0.010 1.686
Saturday 16,272 74 0.006 0.609 99 0.006 1.327 25 0.002 0.159 245 0.015 2.313
Sunday 15,248 52 0.005 0.512 84 0.005 1.171 16 0.001 0.141 186 0.011 1.962
Total 266,184 969 0.004 0.294 1,310 0.005 0.635 147 0.001 0.032 2,608 0.010 0.988
Monday 37,464 111 0.003 0.326 136 0.004 0.564 4 0.000 0.013 187 0.005 0.754
Tuesday 37,824 117 0.003 0.247 102 0.003 0.362 7 0.000 0.015 179 0.005 0.514
Wednesday 37,848 101 0.003 0.182 115 0.003 0.376 10 0.000 0.017 201 0.005 0.585
Thursday 37,824 117 0.003 0.191 166 0.004 0.475 18 0.000 0.034 287 0.008 0.646
Friday 38,496 160 0.004 0.346 234 0.006 0.887 9 0.000 0.021 504 0.013 1.478
Saturday 38,640 217 0.006 0.523 323 0.008 1.300 58 0.002 0.145 762 0.020 2.412
Sunday 38,088 146 0.004 0.379 234 0.006 0.986 41 0.001 0.111 488 0.013 1.758
1
Number of recreation events per camera hour (i.e., RRI = Rec. events/Camera hours).
2
Mean number of recreation events during all 6-hour intervals for a given time period. Values were obtained by back-transformation
of estimated logmean number of events using a generalized linear mixed model.
3
Trail cameras operated from 20–May to 31–October, 2016.
4
Trail cameras operated from 16–May to 28–October, 2017.
5
Trail cameras operated from 24–May to 30–September, 2018.
84
Table 2.12. Trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a camera in the same
direction within a 5 minute period) and recreational intensity (i.e., MN6HR) by time of day (i.e., 3, 6-hour intervals within 5:00–
23:00) for recreation types (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) captured by remote digital trail
cameras on public lands within the Atlanta State Forest management unit during peak summer–fall recreational periods (i.e., May–
October), 2016–2018.
Equestrian use Hiking/foot-traffic Mountain-biking ORV use
Year/time period events MN6HR1 events MN6HR1 events MN6HR1 events MN6HR1
2016a 263 0.281 265 0.433 33 0.032 403 0.710
5:00–10:59 42 0.124 79 0.353 4 0.018 42 0.259
11:00–16:59 125 0.404 80 0.438 26 0.114 193 1.133
17:00–23:00 96 0.444 106 0.525 3 0.016 168 1.219
2017a 324 0.219 576 0.755 50 0.025 1,329 1.186
5:00–10:59 40 0.091 170 0.581 4 0.013 134 0.408
11:00–16:59 184 0.377 235 0.917 37 0.105 679 2.274
17:00–23:00 100 0.304 171 0.807 9 0.011 516 1.796
2018a 381 0.414 455 0.783 64 0.043 867 1.146
5:00–10:59 62 0.207 140 0.723 10 0.027 112 0.473
11:00–16:59 161 0.647 165 0.863 51 0.165 437 1.994
17:00–23:00 158 0.529 150 0.769 3 0.018 318 1.594
Total 968 0.294 1,296 0.635 147 0.032 2,599 0.988
5:00–10:59 144 0.133 389 0.529 18 0.018 288 0.368
11:00–16:59 470 0.462 480 0.702 114 0.126 1,309 1.725
17:00–23:00 354 0.415 427 0.688 15 0.015 1,002 1.517
1
Mean number of recreation events during all 6-hour intervals for a given time period. Values were obtained by back-transformation
of estimated logmean number of events using a generalized linear mixed model.
a
Trail cameras operated from 20–May to 31–October, 2016; 16–May to 28–October, 2017; 24–May to 30–September, 2018.
85
Characteristics of group size for recreation types in the ASF
Mean group sizes for recreation types were 2.83 ± 0.070 (SE) for equestrian use, 1.91 ±
0.036 for hiking/foot-traffic, 1.47 ± 0.090 for mountain biking, and 1.28 ± 0.014 for ORV use
during 2016–2018 (Table 2.13). No differences (p < 0.05) were detected in mean group sizes
among years or months for any recreation type. However, mean group size was greatest on
Saturday for equestrian use (3.55 ± 0.211; X2 = 48.71, df = 6, p < 0.001), hiking/foot traffic (2.25
± 0.092; X2 = 46.99, df = 6, p < 0.001), and ORV use (1.38 ± 0.032; X2 = 46.99, df = 6, p <
0.001). Additionally, mean group size was greater on Friday, Saturday, and Sunday than other
days for equestrian use, hiking/foot-traffic, and mountain biking (Table 2.13). Similar to the
PRC, we observed notable outliers (n = 791 events) in median group size representing
substantially larger group sizes for each recreation type (i.e., equestrian use ≥ 5, hiking/foot-
traffic ≥ 4, mountain biking ≥ 2, ORV use ≥ 2). The largest group sizes for each recreation type
were 22 for equestrian use, 12 for hiking/foot-traffic, 8 for mountain biking, and 13 for ORV use.
September had the highest number (n = 362) of such events among months. Additionally, Friday
through Sunday accounted for 73.7% (n = 583) of such events among days of the week. Lastly,
52.7% (n = 417) of such events occurred from 11:00–16:59.
86
Table 2.13. Mean group sizes of trail-based recreation events (i.e., any number of individuals of the same recreation type passing by a
camera in the same direction within a 5 minute period) for recreation types (i.e., equestrian use, hiking/foot-traffic, mountain biking,
ORV use) captured by remote digital trail cameras on public lands in the Atlanta State Forest management unit during peak summer–
fall recreational periods (i.e., May–October), 2016–2018. Trail cameras operated from 20–May to 31–October, 2016; 16–May to 28–
October, 2017; 24–May to 30–September, 2018.
Temporal Equestrian use Hiking/foot-traffic Mountain-biking ORV use
category events mean SE events mean SE events mean SE events mean SE
Overall 969 2.83 0.070 1,310 1.91 0.036 147 1.47 0.090 2,608 1.28 0.014
2016a 264 2.92 0.139 265 1.92 0.077 33 1.82 0.248 404 1.32 0.046
2017b 324 2.76 0.127 586 1.88 0.051 50 1.24 0.079 1,334 1.30 0.018
2018c 381 2.83 0.102 459 1.94 0.067 64 1.47 0.146 870 1.24 0.023
May 38 2.97 0.265 53 1.91 0.146 9 1.44 0.242 103 1.27 0.065
June 30 2.40 0.163 140 1.80 0.120 39 1.62 0.225 258 1.19 0.031
July 54 2.48 0.154 183 2.19 0.115 15 1.07 0.067 338 1.48 0.061
August 91 2.66 0.135 203 2.01 0.103 5 1.00 0.000 340 1.21 0.028
September 536 2.87 0.086 483 1.89 0.056 63 1.56 0.143 1,138 1.27 0.020
October 220 2.93 0.208 248 1.74 0.068 16 1.31 0.176 431 1.27 0.028
Monday 111 2.35 0.093 136 1.57 0.074 4 1.00 0.000 187 1.31 0.052
Tuesday 117 2.38 0.108 102 1.79 0.134 7 1.00 0.000 179 1.10 0.024
Wednesday 101 2.23 0.105 115 1.69 0.077 10 1.00 0.000 201 1.11 0.023
Thursday 117 2.18 0.095 166 1.66 0.065 18 1.17 0.090 287 1.36 0.044
Friday 160 3.41 0.223 234 1.91 0.078 9 1.22 0.222 504 1.23 0.029
Saturday 217 3.55 0.211 323 2.25 0.092 58 1.93 0.201 762 1.38 0.032
Sunday 146 2.80 0.142 234 1.98 0.094 41 1.24 0.084 488 1.25 0.027
5:00–10:59 144 2.45 0.163 389 1.67 0.051 18 3.00 0.485 288 1.19 0.030
11:00–16:59 470 2.87 0.104 480 1.86 0.056 114 1.19 0.056 1,309 1.33 0.022
17:00–23:00 354 2.94 0.113 427 2.20 0.076 15 1.73 0.206 1,002 1.23 0.020
87
Comparisons between the PRC and ASF
Overall RRI was greater for the PRC (0.043) than ASF (0.019), with more than twice the
number of events (11,412 PRC, 5,034 ASF) captured during fewer overall camera hours
(263,664 PRC, 266,184 ASF) (Table 2.2). Despite differences in recreation intensity between
study regions, we found similar patterns for recreational intensity by year, month, day of week,
and time of day in both study regions. Both study regions had similar RRI during 2016 and 2017,
and lower RRI in 2018 (Table 2.2). September, October, and May had the highest overall RRIs
by month for both study regions, respectively (Table 2.2). Saturday, Sunday, and Friday had the
highest overall RRIs by day of week for both study regions, respectively (Table 2.4). The
majority of events occurred between 11:00–16:59 in both study regions (Table 2.7).
Additionally, temporal patterns for peak days of use were relatively similar between regions with
September and weekend days accounting for most of peak days during the study.
While recreational intensity was similar between regions in temporal patterns of annual,
monthly, day of week, time of day, and peak days of use, we found differences in which
recreation types were most prominent between regions. Equestrian use accounted for the most
(58.8%) recreation events in the PRC, while ORV use accounted for the most (51.8%) events in
the ASF. Notably, equestrian use only accounted for 19.2% of events in the ASF, while ORV use
predictably accounted for < 1% of events in the PRC due to being prohibited. Additionally, ORV
use was the only recreation type to have greater RRIs in the ASF than the PRC.
Our study regions also differed by which variables had the most effect on detection of a
recreation event. Recreation type, month, and time of day had the most effect on detection of a
recreation event within a 6-hour period for both regions, not respectively (Table 2.3). However,
time of day had the most effect for the PRC, while recreation type had the most effect for ASF.
88
In both regions, day of week and year had the least effect on detection of a recreation event
within a 6-hour period. Lastly, mean group sizes and the largest group sizes for each recreation
type besides ORV use were smaller in the ASF than the PRC (Tables 2.9 and 2.13).
89
DISCUSSION
Our study demonstrated consistent summer-fall use of two public land areas for trail-based
recreational activities in the Michigan elk range during 2016-2018. Although variation in
recreational intensity among study months, days of week, and time of day were expected based
on previous literature (Ladle et al. 2017, Longshore et al. 2013, Reilly et al. 2017), we did not
expect to find differences among study years. However, we believe our findings of less
recreational intensity during 2018 may be due in part to greater daily temperatures. For example,
2018 had greater mean daily high temperatures for May-August, which may explain the
decreased recreational intensity in both study regions during all months besides September in
2018 (NOAA 2020). Notably, 2018 had more days of ≥ 26.7 ◦C (60 days) and ≥ 32.2 ◦C (13
days) than 2016 (i.e., ≥ 26.7 ◦C = 50 days, ≥ 32.2 ◦C = 7 days) and 2017 (i.e., ≥ 26.7 ◦C = 38
days, ≥ 32.2 ◦C = 3 days).
Our findings of greater recreational intensities in May and September each year in each
region may be due in part to lower daily high temperatures. Mean daily high temperatures for our
study area were lower in May and September than June-August each year (NOAA 2020). While
September had the highest recreational intensities among months for all trail-based recreation
types, the differences among months was most prominent for equestrian use. For example, the
RRI for equestrian use in September was 9 times greater than July in both regions during 2016-
2018, while only being 1.3-3.3 times greater for other recreation types. We believe the disparity
in recreational intensity between September and other months for equestrian use is likely due in
part to preferences for riding during cooler temperatures. A national online poll that surveyed
equestrian users for preference of riding season showed that 57.2% of 764 respondents preferred
riding in fall, 22% spring, 15.2% summer, and 5.6% winter (Whittle 2017). Approximately 21%
90
of the respondents commented their primary reasons for preferring fall riding were cooler
temperatures, fall colors, and fewer insects. Moreover, the University of Minnesota Extension
recommends avoiding riding horses during periods when the combined air temperature (◦F) and
relative humidity is > 150 due to reduced cooling efficiency and potential for overheating
(Martinson et al. 2020). Evaluation of climatological data during our study showed that 55.4% of
days during June-August had periods when the combined air temperature (◦F) and relative
humidity was > 150, while only 30.6% of days with such conditions occurred during May and
September (NOAA 2020). Although hotter temperatures (i.e., > 32 ◦C) have been found to affect
recreational patterns for hiking and bicycling, responses are typically to change to earlier or later
times of the day when temperatures are cooler (Li and Lin 2012, Chan and Whichman, 2020).
Greater recreational intensity in May and September may also be due to annual US federal
holiday weekends each month (i.e., Memorial Day, a Monday in May; Labor Day, a Monday in
September). Our findings of greater recreational intensity during Memorial Day and Labor Day
weekends (i.e., 25% of our highest daily RRI values) are consistent with previous research
identifying greater visitor numbers in natural areas during holidays (Dwyer 1988, Remacha et al.
2016). Further evaluation of mean RRI values during Memorial Day and Labor Day weekend
days (i.e., Friday–Monday) showed RRIs were 2.4 times greater than other weekends. Hence, we
believe the greater observed recreational intensities during May and September were primarily
due to the combination of lower daily high temperatures and occurrences of Memorial Day and
Labor Day holiday weekends.
Our findings of greater recreational intensity during weekends and mid-day (i.e., 11:00–
17:00) for all recreation types in both regions are similar to previous research on public lands
(Ladle et al. 2017, Longshore et al. 2013). Our findings of earlier (i.e., 2:36) time of day use for
91
equestrian users during July from June in the ASF may be due to greater mean high
temperatures. Evaluation of average monthly high temperatures during 2016-2018 showed July
had a 2.8 ◦C greater mean high temperature than June. Equestrian users may have modified
riding times for earlier in the day to avoid hotter temperatures. Our findings of earlier use for all
recreation types during weekend days is likely due to increased use during weekends. Increased
presence of recreational users that are camping during weekends may explain our findings of
later use on days of arrival (i.e., Thursday, Friday) and earlier use on days of departure (i.e.,
Saturday–Tuesday). Our findings of later use for ORV users may be attributed to headlights and
speed of travel for returning to camping sites.
The differences we documented in recreational intensity, type of use, and group size between
the PRC and ASF may be related to availability of recreational trails and amenities. Previous
research has suggested that trail attributes (e.g., trail markers, loops, lengths > 24 km) and site
amenities (e.g., water sources, camping sites) may contribute to increases in annual visits by
equestrian users ≥ 4 visits per available attribute or amenity (Blackwell et al. 2009). The PRC
had greater recreational intensity than ASF during all months and years while providing 2.5
times more designated recreational trails per km2 (i.e., PRC = 0.46 km/km2, ASF = 0.18 km/km2)
and designated camping and horse camping sites that offer numerous amenities. Although
equestrian use was the most frequently detected type of trail-based recreation in the PRC (i.e.,
58.7% of events) given our objectives, it only accounted for 19.2% of events in the ASF. We
believe the disparity in intensity of equestrian use between regions is primarily due to lack of the
above-mentioned amenities and designated trails in the ASF. The PRC has 69.34 km of
designated equestrian trails while the ASF has none. Equestrian users have 6 equestrian
campgrounds in the PRC, of which 5 have direct connections to designated trails. The primary
92
horse campground in the PRC (i.e., Elk Hill Trail Campground) provides amenities for first-
come, first-served camping and reserved group camping which accommodates up to 5 equestrian
groups and has a maximum capacity of 100 individuals. The PRC also provides 6 rustic
campgrounds with amenities that provide easy access for numerous recreational opportunities,
including 79.77 km of trails designated to bicycling and hiking, while the ASF has none. Hence,
the PRC provides numerous attributes that may be appealing to some recreational users and that
are not available in the ASF.
In contrast, the absence of designated trails and amenities in the ASF may be appealing to
different users who prefer fewer or no restrictions (e.g., off-trail riding, open camping). For
example, in-person communications with equestrian users in the ASF revealed a preference for
off-trail riding and discovering new areas to camp. Conversely, in-person communications with
equestrian users in the PRC revealed a preference for riding in familiar areas that were close to
horse camps that provided amenities for large groups. Our findings of larger mean group size for
equestrian users in the PRC (3.14 ± 0.027) from ASF (2.83 ± 0.070) may support observed user
preferences for allowance of larger equestrian groups in the PRC. The presence of relatively
larger hiking/foot-traffic group sizes (e.g., 41 individuals) of adult-led youth groups only
occurring in the PRC are likely due to the presence of marked designated trails for hiking.
Marked designated trails may provide security for recreational users, especially when leading
large groups or when hikers are unfamiliar with large tracts of public land.
Greater equestrian use in the PRC than the ASF may also be due to avoidance of areas with
ORVs and mountain biking. Others have reported user conflicts between motorized and non-
motorized forms of recreation (Adams and McCool 2009, Shilling et al. 2012, Vaske et al. 2007).
An online survey used by Shilling et al. (2012) to determine trail user preferences and
93
experiences in the Tahoe National Forest reported 63% of non-motorized recreational users (i.e.,
equestrian users, hikers, mountain bikers) that wrote additional comments reported conflicts with
motorized forms (e.g., quads, motorcycles) of recreation. Approximately 54% of equestrian users
and 27% of hikers reported opposition against multiple-use trails (Shilling et al. 2012). Personal
communications with equestrian users in the PRC indicated preferences for avoiding areas
allowing ORVs and mountain bikes, primarily due to avoidance of disturbances to horses caused
by loud and faster moving recreational users.
In addition to user preferences based on recreational restrictions, designated trails, and
amenities within each forest, differences in recreational intensity may be related to the proximity
of Interstate-75 and 2 communities (i.e., Gaylord, MI and Indian River, MI) that provide access
to external amenities (e.g., lodging, restaurants, consumer goods) along the interstate. Previous
research documented that presence of interstate highways increases economic activity in counties
they pass through while causing a decrease in adjacent counties (Chandra and Thompson 2000).
Additional research has documented a decreased likelihood of visitation to protected nature
conservation sites with distance from local housing (Neuvonen et al. 2010, Rossi et al. 2015,
Weitowitz et al. 2019). Approximately 97% of the PRC is within Otsego and Cheboygan
counties which have a combined population of 50,100, and incorporate Interstate-75, the city of
Gaylord (i.e., approximately 3,700 residents), and the community of Indian River (i.e.,
approximately 1,960 residents) that provide goods and services such as lodging and dining.
Notably, the western border of the PRC can be reached from the closest Interstate-75 off-ramp
(i.e., exit 290) with a 5 to 10-minute (i.e., 6.4 km) drive using the only paved road (i.e., East
Sturgeon Valley Road) within the PRC. Additionally, there is signage located 0.5 km from the
Interstate off-ramp reading “Sturgeon Valley Road gateway to Pigeon River Country” to direct
94
visitors. Conversely, approximately 88% of the ASF within the elk range is located within
Montmorency County which has an estimated population of 9,265 and its largest communities
are Lewiston (i.e., approximately 1,400 residents) and Atlanta (i.e., approximately 760
residents). Atlanta is only 1.3 km from the southern edge of the ASF, but only provides two
motels for lodging and few options for dining and goods. The shortest drive to reach the ASF
from the closest Interstate-75 off-ramp (i.e., exit 282) takes approximately 30 minutes (i.e., 32.5
km) along Highway-32 and several county roads, which have no signs directing visitors or
advertising the ASF (C. Stevens, MDNR, personal communication). The close proximity of the
PRC to the only interstate through the northern half of the lower peninsula of Michigan and
nearby population centers may be contributing factors for greater visitor use than the ASF.
Differences in recreational intensity between the two state forests may also be attributed to
the degree of familiarity users had and differences in tourism promotion. Previous research
documented increased use and greater preferences for recreation sites that were highly promoted,
provided more tourism infrastructure, and were perceived as being popular (Hallmann et al.
2014, Schägner et al. 2016, Schirpke et al. 2018). Additionally, Neuvonen et al. (2010)
documented that older national parks in Finland had greater visitor numbers than newer parks,
which was attributed to greater public awareness. Awareness and interest in the PRC has
increased since the early 20th century following the introduction of 7 elk in 1918, elk hunts
occurring in the 1960’s, controversy of drilling for oil and natural gas in the 1970’s, and
increased use of the forest for recreational opportunities in the last 50 years (MDNR 2007).
Hence, the PRC is arguably more popular and promoted due to its historical nature, unique
characteristics that have been conserved through guidelines provided by the COM, and
provisions for more tourism infrastructure than the ASF. The PRC provides numerous amenities
95
for recreational users with a staffed headquarters that provides information on local plants and
wildlife (e.g., brochures, specimens, taxidermy displays); maps for wildlife viewing, recreational
trails, and camping; and an award-winning “Discovery Center” which hosts environmental
education programs, promotes natural resources stewardship through displays and an interpreter,
describes the history of conservation in the PRC, and provides visitors with information for
recreational opportunities and regulations (PRCA 2020). We believe the lower observed
recreational intensity in the ASF may be due in part to its juxtaposition to the larger, more
publicized, and well-developed tourism infrastructure of the PRC.
Although we captured more than twice the number of events in the PRC than the ASF during
fewer overall camera hours and documented greater overall RRI, direct comparison between
regions were limited due to differing camera placement strategies necessitated by differences in
recreational regulations and associated visitor use patterns. Allowance of off-trail activity creates
challenges for assessing recreational activities and patterns. For example, equestrian users in the
ASF may avoid all forest roads and edges to have a more challenging or primitive riding
experience. However, while estimating recreational use in the ASF required additional trail
cameras and time to identify recreational user patterns for camera placements, our different
sampling strategies allowed us to successfully quantify trail-based recreational intensity, group
sizes, and temporal use patterns for both regions. We documented that differences in regulations,
trail systems, amenities, proximity to human development, storied history, and marketing of
these state forests provide different recreational opportunities that ultimately lead to differences
in intensities, group sizes, and primary types of trail-based recreational use.
96
CHAPTER 3: ELK SPACE-USE AND RESOURCE SELECTION PATTERNS IN
RESPONSE TO SUMMER TRAIL-BASED RECREATION
INTRODUCTION
The increased use of public lands for a growing diversification of outdoor recreational
activities in the early 21st century has resulted in rising concerns of the effects to wildlife and
their habitat (Steven et al. 2011, Cordell 2012, Larson et al. 2016). A growing body of research
has examined numerous perceived effects from recreational use on wildlife populations and
communities such as changes in behavior (Jayakody et al. 2008, Naylor et al. 2009, Fortin et al.
2016), physiology (Thiel et al. 2011, Harris et al. 2014, Arlettaz et al. 2015), abundance (Mallord
et al. 2007, Patthey et al. 2008, Wolf et al. 2013), and community composition (Miller et al.
1998). Although wildlife responses to recreational activity may range from relatively short-term
behavioral reactions (e.g., flight responses; Papouchis et al. 2001, Stankowich 2008) to longer
term effects (e.g., avoidance of disturbed areas; Neumann et al. 2009, Coppes et al. 2017b), both
can have direct negative effects such as increased stress, reduced reproductive success, and
decreased foraging (Shively et al. 2005, Arlettaz et al. 2015, Spitz et al. 2019).
Long-term effects of consistent interactions between wildlife and recreational users, such as
avoidance of frequently used areas, create challenges for managers attempting to provide wildlife
habitat components. Wildlife avoidance of areas used by recreational users can lead to indirect
habitat loss and has been observed in large herbivores such as red deer (Cervus elaphus),
mountain caribou (Rangifer tarandus caribou), and elk (Cervus elaphus nelsoni) (Sibbald et al.
2011, Lesmerises et al. 2018, Wisdom et al. 2018). Additionally, wildlife responses to
recreational activity may lead to numerous indirect effects appearing as human-wildlife conflicts
such as vehicle collisions and agricultural crop depredation (Coppes et al. 2017a). Wildlife
97
professionals attempting to manage wildlife population sizes and their spatial distributions
should consider the potential direct and indirect effects of encounters with recreational users
occurring within or near wildlife habitats (Neumann et al. 2011, Coppes et al. 2018, Heinemeyer
et al. 2019). Furthermore, it is vital to understand the differences in the effects caused by
different types of recreational use occurring at varying intensities and temporal periods (Boyle
and Samson 1985, Larson et al. 2016).
In Michigan, there has been an increase in trail-based recreation (e.g., equestrian use,
mountain biking, ORV use) on public lands within the elk range during the last 50 years (MDNR
2007, MDNR 2012). Increased reports of elk causing agricultural depredation outside of their
core range has raised concerns among natural resource managers over the potential impacts of
recreational activities on elk movements and behaviors (B. Mastenbrook, MDNR, personal
communications). Consequently, natural resources managers have hypothesized elk may be
selecting areas with less recreational activity. The Michigan elk population has remained
relatively stable since approximately 2006 (i.e., mean=1,065, 95% CI = 931–1,200; S. Adams,
MDNR, personal communication). Hence, elk occurrences outside of the current range may not
be due to population growth. Research examining elk habitat suitability and potential on public
and private lands within the range indicates an abundance of cover types supporting elk habitat
requirements, suggesting elk may not be selecting sites outside of their range for habitat
components (Williamson et al. 2021). Notably, 30% of the public land within the elk range was
composed of high-quality food sources for elk (e.g., 16% northern hardwoods/maple, 11%
openings, 3% regenerating aspen). While elk habitat is managed relatively consistently across
public lands within the Michigan Department of Natural Resources (MDNR) defined elk range,
recreational guidelines differ between the Pigeon River Country (PRC) State Forest and Atlanta
98
State Forest (ASF) Management Units. The PRC’s recreational guidelines were established in its
Concept of Management (COM), which are specific to the PRC due to its designation as a
Special Management Unit (MDNR 2007). These guidelines led to regulations limiting equestrian
use and mountain biking to designated trails, while prohibiting ORV use (MDNR 2016).
Our objectives were to: 1) evaluate and compare space-use patterns and resource selection for
Michigan elk in response to habitat suitability and the intensity of summer trail-based recreation
types (i.e., equestrian use, hiking/foot-traffic, mountain biking, ORV use) at different temporal
periods (i.e., year, month, day, hour) for the PRC and ASF; and 2) provide recommendations to
natural resource managers challenged with balancing objectives for managing elk, their habitat,
and diverse recreational opportunities on state forests. Based on previous findings of elk
avoiding frequently used areas by recreational users (Rogala et al. 2011, Wisdom et al. 2018), we
predicted that during peak periods of summer trail-based recreational intensity (i.e., May,
September, weekends, mid-day; Williamson et al. in review) elk would: 1) have greater monthly
home-range sizes and daily movement distances; 2) exhibit increased activity during typical
periods of inactivity (e.g., mid-day); 3) use areas of relatively lower habitat suitability; and 4) use
areas farther from roads and recreational trails. Finally, based on previous findings of ORV use
having greater effects on elk than other types of recreation (e.g., equestrian use, hiking, mountain
biking; Naylor et al. 2009, Wisdom et al. 2018), we predicted that the above-mentioned
predictions would be more pronounced in the ASF since ORV use is prohibited in the PRC.
99
METHODS
Elk Capture
Personnel from Wildlife Helicopter Services (Austin, TX, USA), MDNR, and Michigan
State University (MSU) captured elk in the PRC and ASF regions of the elk range using net-
gunning techniques during February 2016 (Schemnitz 1996, Walsh 2007). Our goal was to
capture an equal sex ratio and spatial distribution of elk between and within each region.
Captured elk were ear tagged and fitted with a collar (Vectronic Aerospace GmbH, Berlin,
Germany; VERTEX Plus Iridium) that included a Global Positioning System (GPS) receiver and
a very high frequency (VHF) beacon transmitter. Collars deployed on bull elk included
polyurethane foam inserts to accommodate neck swelling during rutting periods. To redeploy
collars collected after mortality events in 2016, personnel from the MDNR immobilized an elk
by administering 2 ml of butorphanol-azaperone-medetomidine via a 2-ml Pneu-Dart Type ‘P’
RDD (Pneu-Dart, Williamsport, PA, USA) dart and a Pneu-Dart X-caliber dart gun and used 0.5
ml of Naltrexone and 4 ml of Atipamezole via syringe for reversal. To redeploy refurbished and
additional new collars following 2017 mortality events and collar failures, we used the same
personnel and methods described for our 2016 capture during winter 2018. All capture and
handling procedures were developed and reviewed by the MDNR’s wildlife veterinarian and
approved by the MDNR.
Elk Locations
Collars calculated locations every 30 minutes during 1–May to 30–September (i.e., peak
summer trail-based recreation period), 2016–2018. Collars were programmed to drop-off on 15–
January, 2019. Locations were stored on-board and on a server maintained by Vectronic
Aerospace that was accessible using a password login. Collar locations were retained if at least 4
100
satellites were used and the dilution of precision (DOP) was <10.0 (i.e., estimated error is ≤15 m
for approximately 99.6% of locations; C. Kochanny, Vectronic Aerospace, personal
communication). We estimated collar accuracy by placing collars in open and medium–heavy
forest canopy cover areas and recorded GPS locations using a handheld GPS receiver (GPSMAP
64s, Garmin International, Olathe, KS, USA).
Elk Movement and Behavior
We investigated elk movement and behavioral responses during peak periods (i.e., May,
September, weekends, mid-day) of summer trail-based recreational intensity in the PRC and ASF
(Williamson et al. in review). To evaluate cow and bull elk movement and behavior patterns at
multiple spatial and temporal scales, we quantified elk summer (i.e., 1–May to 30–September)
home ranges, monthly (i.e., May–September) home ranges, daily movement distances, and
hourly patterns of circadian movement behavior. We used a dynamic Brownian bridge
movement model (dBBMM; Kranstauber et al. 2012) to estimate summer and monthly home
range sizes (i.e., 50% and 95% utilization distributions [UD]) for each sex in each region. The
dBBMM is an extension of the standard Brownian bridge movement model (BBMM; Horne et
al. 2007) that allows for changes in movement patterns across temporal periods. Specifically, the
dBBMM uses a sliding window, maximum likelihood estimation, and the Bayesian information
2
criterion to determine Brownian motion variance (𝜎𝑚 ) at different time steps, and subsequently
2
applies the traditional BBMM using separate estimates of 𝜎𝑚 (Kranstauber et al. 2012; Horne et
al. 2007). We used a location error of 15 m, a moving window size of 13, and a margin of 3 in
the dBBMM computation (Byrne et al. 2014).
We calculated daily linear movement distances for elk using the sum of consecutive 30-
minute increment GPS locations within a 24-hour period as an estimate of the total distance
101
moved each day for individual elk (Devore et al. 2016). Days with missing locations (i.e., < 48)
were not included in our analysis. Median daily linear movement distances for each elk and each
day of the week were averaged across months and years and one-way analysis of variances
(ANOVA) were used to assess differences between cows and bulls in the ASF and PRC and
across the days of the week. We examined the greatest linear daily elk movement distances by
extracting outliers to identify patterns between sexes, regions, and among days of the week.
Outliers were identified by extracting linear daily elk movement distances > Q3 + 1.5 ×
interquartile range [IQR].
We evaluated changes in elk circadian movement behavior throughout the day (i.e., 0:00 –
2
24:00) by estimating 𝜎𝑚 for individual elk at 1-hour intervals for each day of the week averaged
2
across months and years. Median 𝜎𝑚 was calculated for each hour interval for each day and
Kruskal–Wallis tests were used to assess for differences among sexes, regions, and days of the
week during each interval. An alpha level of 0.05 was used for all tests for significance. The
move package in R was used to calculate dBBMMs (Kranstauber et al. 2020).
Elk Resource Use
To investigate the effects of recreational intensity on elk resource use, we examined resource
selection of cows and bulls at different spatial and temporal scales. The primary goal of our
landscape–scale analyses was to determine which cover types had the greatest probability of use
by cows and bulls in the ASF and PRC during May–September, 2016–2018. The goal of our
home range–scale analyses was to determine if use of cover types within cow and bull home
ranges varied during peak periods of recreational intensity (i.e., May, September, weekends,
mid-day).
102
Landscape-scale resource use
We performed multiple resource selection analyses (Nielson and Sawyer 2013) to evaluate
use of cover types by cows and bulls within the ASF and PRC. Michigan DNR forest inventory
data was used in ArcGIS version 10.6.1 (Environmental Systems Research Institute, Redlands,
CA, USA) to delineate polygon layers for 7 cover type categories (i.e., openings, aspen, northern
hardwoods/maple, oak, other hardwoods, upland conifers, lowland conifers) for all public lands
within each region. To account for declines in browse use as age of aspen increases (Campa
1989, Campa et al. 1993, Raymer 2000), we divided our aspen cover type polygon layer into 2
sublayers based on age (i.e., regenerating aspen [<7 years of age], aspen [≥7 years of age)
(Williamson et al. 2021). We accounted for changes in the spatial distribution and structure of
cover types due to ongoing forest management practices (e.g., clearcutting) by evaluating
MDNR forest inventory data each year for 2016–2018. The mean size of openings and
regenerating aspen stands in the ASF and PRC were calculated to evaluate for differences
between the regions due to different management guidelines in the COM (e.g., restrictions for
even-aged forest management treatments [i.e., clearcuts] to be no greater than approximately 16
ha in the PRC; MDNR 2007).
We overlaid cover type polygon layers onto a 30 x 30 m grid (Williamson et al. 2021) and
determined the number of elk locations occurring within each 30 x 30 m grid cell for each
temporal period (i.e., year, month, day of week, hour of day). To model the probability of elk use
for each cover type during each year, generalized linear mixed models (GLMM) were used with
a negative binomial distribution to account for overdispersion in the data that resulted from a
large number of pixels that did not contain elk locations (Breslow and Clayton, 1993). Models
were fitted separately for cows and bulls in each region (Ruhl 1984, Beyer 1987, Walsh 2007).
103
Our response variable was the number of times an individual elk was detected in a 30 x 30 m
grid cell. We included a random effect of individual elk in each model to account for non-
independence between locations from the same elk (Devore et al. 2016). Cover types were
included as indicator variable fixed effects in each GLMM, taking value 1 if that 30 x 30 m pixel
was of the specific cover type and value 0 if the pixel was a different cover type. The natural log
of the total number of elk GPS locations was included as an offset term in each GLMM to
interpret model coefficients for the probability of elk use rather than for the number of elk
locations (Nielson and Sawyer 2013, Devore et al. 2016).
Home range-scale resource use
The patterns of cover type selection within elk home ranges were quantified using the 50%
and 95% UDs calculated in our dBBMMs. We considered cover types within the 95% UDs as
available and 50% UDs as selected to evaluate changes in cover type selection patterns during
different time periods (i.e., month, day, hour) (Byrne et al. 2014, Silva et al. 2018). Proportions
of cover types occurring in elk 95% UDs were averaged across years (i.e., May–September,
2016–2018) for individuals (i.e., to avoid pseudoreplication) and then averaged across
individuals for each sex in each region. To compare elk selection patterns to proportions of
available cover types occurring in 95% UDs, we calculated the proportions of elk locations
occurring in each cover type within 50% UDs for each individual elk during each time period.
We averaged the proportions of elk locations for individual elk across time periods (e.g., year,
month, day of week) during each respective temporal category, and calculated the mean
proportional 50% UD cover type use for cows and bulls in the ASF and PRC. Analyses were
conducted in R (R Core Team 2018).
104
Influence of Habitat Suitability and Recreational use of Roads and Trails
Elk use of suitable areas
To further investigate how elk resource use may be affected by recreational intensity, we
sought to determine if elk used areas of greater habitat suitability within their range and if
suitability of elk locations varied during peak periods of recreational intensity (i.e., May,
September, weekends, mid-day). We used a Michigan elk habitat suitability model for spring
food created by Williamson et al. (2021) to identify suitability values of each 30 x 30 m grid cell
on public lands across the elk range and for each elk location. Although our spring food model
was designed to evaluate habitat requirements for elk occurring in spring (i.e., March–May), elk
feeding behavior in Michigan is similar throughout the summer and fall months (i.e., June–
November) and suitability ratings are assumed to relate directly to quality of habitat from spring
to fall (Beyer 1987, Williamson et al. 2021). Suitability values were estimated based on each grid
cell’s value to provide spring food based on its cover types, stand-level attributes (e.g., age,
canopy closure, size, shape), and spatial juxtaposition to other cover types (Williamson et al.
2021). To determine whether elk selected areas with greater habitat suitability within the elk
range, we compared mean habitat suitability of all grid cells found within bull and cow elk home
ranges to all grid cells within the PRC and ASF. To determine whether elk selected for portions
of their home ranges with greater habitat suitability, we compared the mean habitat suitability of
all grid cells within 95% (i.e., available) and 50% (i.e., used) UDs for bulls and cows in each
region. To determine whether habitat suitability of elk locations changed during peak periods of
trail-based recreation (i.e., May, September, weekends, mid-day; Williamson et al. in review),
we used ANOVAs to assess for differences in mean habitat suitability among temporal periods
(i.e., month, day, hour) for cows and bulls in each region.
105
Elk use of areas near roads and trails
To further investigate how use of cover types may be affected by recreational intensity, we
examined the spatial relationships between elk locations and recreational trails and roads on
public lands during peak periods of recreational intensity (May, September, weekends, mid-day).
We examined distances to roads and trails at landscape– and home range–scales to determine if
elk were disproportionately using locations farther from roads and trails and if distances differed
during peak periods of recreational intensity. Thus, we quantified distances to roads and trails
within the elk range (i.e., landscape–scale) and within elk 95% UDs (i.e., home range–scale).
Previous research documented responses of wildlife to human recreation may vary depending on
cover type (van der Zande et al. 1984, Thiel et al. 2007, Coppes et al. 2018). Additionally, cover
types were not evenly distributed across the elk range and provide different habitat components
for elk (e.g., food, hiding cover, thermal cover). Thus, we performed separate analyses for
regenerating aspen, openings, and northern hardwoods/maple based on their greater observed use
within our resource selection analyses and their potential to provide greater habitat suitability
than other cover types (Williamson et al. 2021). Although each cover type provides highly
suitable food sources for elk, openings provide no cover, regenerating aspen stands provide
horizontal (i.e., hiding) cover, and northern hardwood/maple stands provide horizontal and
vertical (i.e., thermal) cover (Beyer, 1987, Williamson et al. 2021). For analyses in the ASF, we
combined roads and trails since there was only 1 primary multi-use trail in the region (Figure
3.1). Roads and trails (i.e., multi-use, biking, equestrian) were evaluated separately in the PRC.
For our landscape–scale analyses, Mann-Whitney U tests were used to compare the average
median distance of cow and bull locations to the median distance of all possible locations (i.e.,
each 30 x 30 m grid cell) to roads and trails within each cover type (i.e., regenerating aspen,
106
openings, northern hardwoods/maple) in each region. For our home range–scale analyses, we
used Mann-Whitney U tests to compare the average median distance of cow and bull locations
(i.e., averaged across individuals) within 50% UDs to the median distance of all possible
locations (i.e., each 30 x 30 m grid cell) to roads and trails for each cover type within 95% UDs.
To further evaluate for differences in elk median distances to roads and trails within home ranges
during peak summer trail-based recreational periods (i.e., September, weekends), Kruskal-Wallis
tests were used to evaluate for differences among months (i.e., May–September) and days of the
week. We did not evaluate for differences among hours since elk use of cover types varied with
time of day resulting in insufficient data for comparisons. For example, in openings in the ASF
we recorded >1000 cow locations each hour from 3:00–8:00, while recording <100 locations
each hour from 15:00–20:00. An alpha level of 0.05 was used for all tests for significance. All
statistical analyses were performed in R statistical software (R Core Team 2018) using the
pgirmess (Giraudoux 2018) package and visualized using the ggplot2 (Wickham 2016) package.
107
Figure 3.1. Elk capture and GPS-collaring locations and recreational trails and campgrounds within the Michigan Department of
Natural Resources designated elk range in the northern lower peninsula of Michigan. Elk capture and collaring was during 15–16
February, 2016 (n=40, 18 PRC, 22 ASF), 9 April, 2017 (n=1, ASF), and 22 February, 2018 (n=12, 7 PRC, 5 ASF).
108
RESULTS
Elk Capture
During 15–16 February, 2016, 40 adult elk (20 m, 20 f) were captured in the PRC (7 M, 11
F) and ASF (13 M, 9 F) (Figure 3.1, APPENDIX A). On 9 April, 2017, MDNR personnel
immobilized and collared an additional bull elk in the ASF. On 22 February, 2018, 8 refurbished
and 4 new collars were deployed on 7 adult elk in the PRC (5 M, 2 F) and 5 cow elk in the ASF.
Elk Locations
We recorded 764,758 elk locations during 1–May to 30–September, 2016–2018. We
removed 467 locations that had DOPs ≥10.0 or <4 satellites were used, and retained 764,291
locations of which 98.6% (753,961) were within the elk range (i.e., 48.9% = PRC, 33.9% = ASF,
17.2% = private land; Table 3.1). In 2016, 2017, and 2018, we recorded 277,913 (PRC = 46.5%,
ASF = 36.4%, private = 17.1%), 217,162 (PRC = 46.1%, ASF = 35.9%, private = 18%), and
258,886 (PRC = 53.7%, ASF = 29.7%, private = 16.6%) locations within the elk range,
respectively.
Two collar failures (i.e., stopped providing GPS fixes, intermittent) occurred in 2016, 7 in
2017, and 16 in 2018 (APPENDIX B). The mean fix success rate for collars that did not fail (n =
28) was 99.5% (SD = 0.007, range = 95.7–99.8%). We recorded 108 locations to evaluate
precision and accuracy of GPS collars in open and full forest canopy cover testing sites. Mean
linear error was 3.61 m (SD = 1.94, range = 0.18–9.13) in open areas (n = 59) and 7.48 m (SD =
5.92, range = 0.93–27.28) in full canopy cover areas (n = 49). Fourteen collars were collected
from 2016–2020 following mortality events primarily through hunter harvests (APPENDIX B).
109
Elk Movement and Behavior
Mean summer elk home range estimates (dBBMM 95% UD) averaged across 2016–2018
were larger (P < 0.01) in ASF than in PRC and were larger (P < 0.01) for bulls in both regions
(Table 3.2). Mean summer core area estimates (dBBMM 50% UD) averaged across years were
similar between cows and bulls while being larger (P < 0.01) in ASF than PRC (Table 3.2). No
significant differences (P < 0.05) in mean home range or core area sizes were detected among
years for either sex or region. Mean monthly (i.e., May–September) home range sizes were
larger (P < 0.05) in May than June–September for cows in both regions and bulls in the ASF
(Table 3.3). Mean monthly home range size for bulls in the PRC was largest in May, but we
found no significant differences (P < 0.05) among months (Table 3.3). Mean monthly core area
sizes were: 1) larger (P = 0.03) in May than September for cows in the ASF; 2) larger (P < 0.01)
in May than June–September for bulls in the ASF and cows in the PRC; and 3) larger (P < 0.01)
in May than July–September for bulls in the PRC (Table 3.3).
Average median daily linear movement distances were larger in the ASF than the PRC for
cows (+211 m) and bulls (+86 m), but differences between regions were not significant (α =
0.05). We found no significant differences among days of the week for cow or bull elk mean
daily linear movement distances in the ASF or PRC averaged across all time periods (i.e., May–
September, 2016–2018) and during each month averaged across years. However, average median
daily linear movement distance was greatest on Friday and Saturday for cows in ASF and
Saturday and Sunday for bulls in ASF and cows and bulls in PRC (Table 3.4). We extracted 347
outliers representing the longest daily linear distances traveled by elk (i.e., >7.5 km [> Q3 + 1.5
× IQR]), of which: 1) 63% were from bulls; 2) 55% were in ASF; 3) 42% were in May, 23% in
June, 18% in September, 13% in July, and 4% in August; and 4) 70 occurred on Saturday, 54 on
110
Sunday, 49 on Friday, 46 on Monday, 45 on Thursday, and 42 on Tuesday, and 41 on
Wednesday.
2
Average median hourly elk Brownian motion variance (𝜎𝑚 ) was greatest between 6:00–8:00
2
and 19:00–21:00 for bulls and cows in the ASF and PRC (Figure 3.2). Notably, 𝜎𝑚 typically
increased between 3:00–6:00 and 16:00–20:00 and decreased between 8:00–11:00 and 21:00–
2
1:00. No differences were detected in hourly 𝜎𝑚 between bulls and cows or between regions.
2
Few differences (P < 0.05) were detected in hourly 𝜎𝑚 among days of the week and time of day
for either sex in either region, of which none occurred during peak periods (i.e., weekends, mid-
day) of recreational intensity or were ecologically relevant.
111
Table 3.1. Locations of GPS-collared elk on public (i.e., Atlanta State Forest [ASF], Pigeon
River Country State Forest [PRC]) and private lands inside and outside of the Michigan
Department of Natural Resources designated elk range in northern lower Michigan from 1–May
to 30–September, 2016–2018.
Inside of elk range Outside of elk range
Year ASF PRC private public private
2016 101,097 129,367 47,449 0 1,345
2017 77,927 100,165 39,070 2,431 668
2018 76,841 139,054 42,991 2,145 3,741
Total 255,865 368,586 129,510 4,576 5,754
Table 3.2. Mean summer (i.e., 1–May to 30–September) elk home ranges (i.e., 95% utilization
distribution [UD]) and core areas (i.e., 50% UD) as estimated by dynamic Brownian bridge
movement models in Atlanta State Forest (ASF) and Pigeon River Country (PRC) State Forest in
the northern lower peninsula of Michigan, 2016–2018.
95% UD size (ha) 50% UD size (ha)
Region/sex N x̄ SD Range x̄ SD Range
ASF
Cows 12 936 215 478–1,248 115 31 70–183
Bulls 14 1,160 187 726–1,723 120 23 67–179
PRC
Cows 13 676 160 363–1,057 93 14 63–124
Bulls 10 886 209 452–1,770 96 13 54–127
112
Table 3.3. Mean monthly elk home ranges (i.e., 95% utilization distribution [UD]) and core areas (i.e., 50% UD) as estimated by
dynamic Brownian bridge movement models in Atlanta State Forest (ASF) and Pigeon River Country (PRC) State Forest in the
northern lower peninsula of Michigan during summer 2016–2018.
Cows Bulls
95% UD size (ha) 50% UD size (ha) 95% UD size (ha) 50% UD size (ha)
Region/month N x̄ SD x̄ SD N x̄ SD x̄ SD
ASF
May 13 602 221 46 12 14 799 214 65 14
June 12 355 115 35 8 14 484 131 44 9
July 12 395 136 40 8 14 504 137 40 9
August 12 384 92 40 11 14 395 82 36 6
September 12 366 110 34 10 14 460 176 36 9
PRC
May 13 452 161 53 10 10 489 192 53 15
June 13 246 79 31 7 10 442 219 41 15
July 13 308 69 33 5 10 354 92 33 9
August 12 286 81 33 7 11 299 60 30 6
September 12 340 139 34 7 10 449 235 35 11
113
Table 3.4. Average median elk daily linear movement distances (km) in Atlanta State Forest
(ASF) and Pigeon River Country (PRC) State Forest in the northern lower peninsula of
Michigan, 2016–2018.
ASF PRC
Cows (n = 13) Bulls (n = 14) Cows (n = 13) Bulls (n = 11)
Day x̄ SD x̄ SD x̄ SD x̄ SD
Monday 4.08 1.34 4.20 1.53 3.96 1.38 4.00 1.54
Tuesday 4.14 1.43 4.16 1.51 3.94 1.33 4.10 1.83
Wednesday 4.04 1.37 4.02 1.40 3.95 1.27 4.12 1.66
Thursday 4.09 1.41 4.19 1.47 3.85 1.14 4.12 1.68
Friday 4.22 1.49 4.26 1.44 4.01 1.33 4.06 1.59
Saturday 4.15 1.50 4.38 1.66 4.17 1.45 4.40 1.85
Sunday 4.13 1.50 4.39 1.51 4.04 1.38 4.22 1.79
114
2
Figure 3.2. Average median hourly Brownian motion variance (𝜎𝑚 ) for GPS-collared elk in the Atlanta State Forest (A = cows [n =
13], B = bulls [n = 14]) and Pigeon River Country (C = cows [n = 13], D = bulls [n = 11]) State Forest in the northern lower peninsula
of Michigan from May–September, 2016–2018. Gray-filled areas represent mean times of day between sunset and sunrise during our
monitoring periods.
115
Elk Resource Use
Landscape-scale resource use
Delineations of cover types for the PRC (45,840 ha) and portions of the ASF (16,800 ha) within
the elk range revealed minor changes in proportions (i.e., 0.1–0.6%) of cover types among years
in the PRC and ASF. For example, area of regenerating aspen (<7 years of age) increased in the
PRC from 2016 (880 ha) to 2017 (990 ha) and 2018 (1,100 ha) due to clearcutting mature aspen
stands. Similar proportions (i.e., within <3%) of openings, regenerating aspen, northern
hardwoods, oak, and lowland conifers occurred between the PRC and ASF. The PRC had
proportionally more (7.9–8.4%) aspen (≥7 years of age), while the ASF had proportionally more
other hardwoods (i.e., 3.2–3.3%) and upland conifers (i.e., 4.8%). In the PRC, the proportional
distribution of cover types was 22.4% (SD = 0.24%) aspen (≥7 years of age), 21% (SD = 0.06%)
upland conifers, 17.3% lowland conifers, 15.9% northern hardwoods/maple, 10.7% (SD =
0.05%) openings, 5.6% other hardwoods, 2.4% oak, and 2.3% (SD = 0.25%) regenerating aspen.
In the ASF, the proportional distribution of cover types was 25.8% (SD = 0.06%) upland
conifers, 16.9% lowland conifers, 14.2% aspen (≥7 years of age), 13.4% (SD = 0.11%) northern
hardwoods/maple, 12.2% (SD = 0.06%) openings, 8.9% (SD = 0.05%) other hardwoods, 4.9%
oak, and 1.5% regenerating aspen. The mean size of openings was 3 ha (n=667, min=0.004 ha,
max=57 ha, median=1 ha) in the ASF and 3 ha (n=1,204, min=0.059 ha, max=130 ha, median=1
ha) in the PRC. The mean size of regenerating aspen stands was 7 ha (n=36, min=1.4 ha,
max=52 ha, median=4 ha) in the ASF and 8 ha (n=120, min=0.3 ha, max=33 ha, median=7 ha) in
the PRC.
The greatest relative probability of cow and bull elk use was in either openings or
regenerating aspen each year (i.e., May–September) in each region (Figure 3.3). In the ASF,
116
openings had the greatest relative probability of use in 2016 (0.269) and 2017 (0.547) for cows
and in 2016 (0.299) and 2018 (0.233) for bulls (Figure 3.3). Regenerating aspen had the greatest
probability of use in 2017 (0.258) for bulls and 2018 (0.306) for cows in the ASF. The only
cover type with greater probability of use than regenerating aspen was upland conifers used by
cows in 2016 (0.192) and 2017 (0.263) and other hardwoods used by bulls in 2016 (0.171) in the
ASF. We found no use of regenerating aspen (< 7 years old) by cow elk in the ASF in 2017
(Figure 3.3). However, probability of cows in the ASF using openings and upland conifers
increased from 2016 to 2017 by 103% and 37%, respectively. Conversely, probability of cows in
the ASF using openings and upland conifers decreased from 2017 to 2018 by 53% and 81%,
respectively, while probability of using regenerating aspen (0.306) increased to a greater
probability of use than openings (0.259) (Figure 3.3). For bulls in the ASF, probability of use
was relatively consistent among years for all cover types except for regenerating aspen which
increased 95% from 2016 to 2017. In the PRC, regenerating aspen had the greatest relative
probability of use for cows and bulls each year, followed by openings and northern
hardwoods/maple, respectively (Figure 3.3). Probability of using regenerating aspen increased
for bulls in the PRC by 17% from 2016 to 2017 and 16% from 2017 to 2018. Lowland conifers
had the lowest probability of use each year in each region, except for regenerating aspen by cows
in the ASF in 2017.
117
Figure 3.3. Relative probability of use of cover types by GPS-collared elk in the Atlanta State Forest (A = cows [n = 13], B = bulls [n
= 14]) and Pigeon River Country (C = cows [n = 13], D = bulls [n = 11]) State Forest in the northern lower peninsula of Michigan
from May–September, 2016–2018.
118
Home range-scale resource use
We found differences between cow and bull cover type proportions, and which cover types
composed the majority of 95% UDs in the ASF (Table 3.5). Cow elk 95% UDs in the ASF were
primarily composed of upland conifers (28.7%), openings (16.8%), and northern
hardwoods/maple (15.7%). Bull 95% UDs in the ASF were primarily composed of mature aspen
(20.7%), northern hardwoods/maple (20.6%), and openings (16.4%). In the ASF, cow 95% UDs
had proportionally more upland conifers (13.4%), whereas bull 95% UDs had proportionally
more mature aspen (9.2%), northern hardwoods/maple (4.9%), and other hardwoods (2.8%). In
the PRC, cow and bull mean cover type proportions within 95% UDs were similar with only
mature aspen (i.e., 5.9% greater for bulls) and openings (i.e., 5.5% greater for cows) having
proportional differences >2% (Table 3.5). Northern hardwoods/maple (cows = 32.2%, bulls =
33.0%), mature aspen (cows = 20.2%, bulls = 26.1%), and openings (cows = 19.9%, bulls =
14.4%) were the most prominent cover types found within cow and bull 95% UDs in the PRC,
respectively.
From May–September cow and bull use (i.e., proportion of locations within 50% UDs) of
openings and regenerating aspen was greater than availability (i.e., proportion of area within
95% UDs) in both regions during 2016–2018 (Table 3.5). Bulls used mature aspen greater than
their availability in both regions from 2016–2018. In the PRC, cows and bulls used other
hardwoods greater than their availability from 2016–2018. Openings was the only cover type
used greater than their availability during all months (i.e., May–September) by cows and bulls in
both regions (Figure 3.4). Cow and bull use of openings was greatest during August in both
regions. Elk use of regenerating aspen was greater than availability during all months except for
cows in the ASF during August and bulls in the PRC during May, although use was within 1% of
119
availability during both months (Figure 3.4). Cow use of mature and regenerating aspen was
greatest in September in both regions. No notable differences were found in use versus
availability of cover types by elk among days of the week (Figure 3.5).
Cows and bulls in both regions tended to use regenerating aspen and openings greater than
their availability during crepuscular and nocturnal periods (e.g., 18:00–8:00) and less than their
availability during daytime hours (e.g., 8:00–18:00) (Figure 3.6). Conversely, elk in both regions
tended to use mature aspen, northern hardwoods/maple, and upland conifers greater than their
availability during daytime hours and less than their availability during crepuscular and nocturnal
hours (Figure 3.6). In the PRC, elk use of other hardwoods was greater than availability during
all times of day for bulls and most times of day for cows (Figure 3.6).
120
Table 3.5. Mean proportions of cover types and elk locations found within GPS-collared elk home ranges (i.e., 95% utilization
distribution [UD]) and core areas (i.e., 50% UD) in the Atlanta State Forest (ASF) and Pigeon River Country (PRC) State Forest in the
northern lower peninsula of Michigan from May–September, 2016–2018.
ASF PRC
Female Male Female Male
Cover type N Prop.95a Loc.50b N Prop.95 a Loc.50 b N Prop.95 a Loc.50 b N Prop.95 a Loc.50 b
Aspen (mature) 12 0.115 0.088 14 0.207 0.221 13 0.202 0.150 10 0.261 0.275
Aspen (regenerating) 11 0.042 0.055 11 0.028 0.054 13 0.055 0.148 10 0.047 0.113
Lowland conifers 12 0.012 0.008 14 0.015 0.005 12 0.015 0.011 10 0.034 0.019
N. hardwoods/maple 12 0.157 0.107 14 0.206 0.137 13 0.322 0.228 10 0.330 0.230
Oak 12 0.084 0.045 14 0.066 0.038 10 0.021 0.008 8 0.011 0.008
Openings 12 0.168 0.325 14 0.164 0.326 13 0.199 0.318 10 0.144 0.236
Other hardwoods 12 0.126 0.084 14 0.154 0.104 13 0.051 0.052 10 0.047 0.059
Upland conifers 12 0.287 0.285 14 0.154 0.119 12 0.133 0.087 10 0.122 0.060
a
Proportion of cover types found within elk 95% UDs. Mean proportions of cover types were calculated by averaging individual elk
cover type proportions across years before averaging across each group (i.e., females and males in each region).
b
Proportion of elk locations found within elk 50% UDs. Mean proportions of elk locations were calculated by averaging individual
elk location proportions across years before averaging across each group (i.e., females and males in each region).
121
Figure 3.4. Proportional monthly core area (i.e., 50% utilization distribution [UD]) use of cover
types within elk home ranges (i.e., 95% UD) in the Atlanta State Forest (A = cows [n = 13], B =
bulls [n = 14]) and Pigeon River Country (C = cows [n = 13], D = bulls [n = 11]) State Forest in
the northern lower peninsula of Michigan from May–September, 2016–2018. Dashed horizontal
lines represent proportional availability of each cover type within elk 95% UDs. Oak and lowland
conifer cover types were omitted due to low proportional use during the sampling period.
122
Figure 3.5. Proportional daily core area (i.e., 50% utilization distribution [UD]) use of cover
types within elk home ranges (i.e., 95% UD) in the Atlanta State Forest (A = females, B = males)
and Pigeon River Country (C = females, D = males) State Forest in the northern lower peninsula
of Michigan from May–September, 2016–2018. Dashed horizontal lines represent proportional
availability of each cover type within elk 95% UDs.
123
Figure 3.6. Proportional hourly core area (i.e., 50% utilization distribution [UD]) use of cover types within elk home ranges (i.e., 95%
UD) in the Atlanta State Forest (A = cows [n = 13], B = bulls [n = 14]) and Pigeon River Country (C = cows [n = 13], D = bulls [n =
11]) State Forest in the northern lower peninsula of Michigan from May–September, 2016–2018. Dashed horizontal lines represent
proportional availability of each cover type within elk 95% UDs. Oak and lowland conifer cover types were omitted due to low
proportional use during the sampling period. Gray-filled areas represent mean times of day between sunset and sunrise during our
monitoring periods.
124
Influence of Habitat Suitability and Recreational use of Roads and Trails
Elk use of suitable areas
Mean elk habitat suitability for spring food was 0.254 (SD = 0.153) in the PRC and 0.258
(SD = 0.138) in the ASF (Williamson et al. 2021). Elk selected areas of greater habitat suitability
for home ranges (i.e., 95% UD) in both regions (ASF = 0.311 ± 0.058, PRC = 0.372 ± 0.086).
We found greater (P < 0.05) mean habitat suitability within UDs for cows in the PRC (95% =
0.406 ± 0.081, 50% = 0.408 ± 0.091) than for cows in the ASF (95% = 0.280 ± 0.067, 50% =
0.282 ± 0.090), bulls in the ASF (95% = 0.333 ± 0.038, 50% = 0.344 ± 0.052), and bulls in the
PRC (95% = 0.323 ± 0.069, 50% = 0.315 ± 0.087). However, no differences (P < 0.05) in mean
habitat suitability between 95% and 50% UDs were detected for either sex in each region. No
differences (α = 0.05) in mean habitat suitability within 50% UDs were detected during peak
periods of recreational intensity (i.e., May, September, weekends, mid-day) among months, days,
and hours for either sex in each region.
Elk use of areas near roads and trails
In the ASF, bull locations in openings had a closer (P = 0.02) median distance (61.3 m) to
roads or trails than the median distance (108.7 m) of all openings to roads or trails (Table 3.6). In
the PRC, cow locations in northern hardwoods/maple had a closer (P = 0.04) median distance
(459 m) to roads than the median distance (269.2) of all northern hardwoods/maple to roads
(Table 3.6). Cow locations in regenerating aspen had a closer (P < 0.01) median distance (229.8
m) to trails than the median distance (1.17 km) of all regenerating aspen to trails in the PRC.
Cow and bull locations in regenerating aspen, openings, and northern hardwoods/maple had
closer (P < 0.05) median distances to equestrian trails than the median distances of each cover
type to equestrian trails in the PRC (Table 3.6). No differences (α = 0.05) were detected between
125
median distances of elk locations to roads and trails in 50% UDs and median distances to roads
and trails in elk 95% UDs for cows or bulls in each cover type in each region (Table 3.7).
Additionally, we found no differences in median distance to roads and trails during peak periods
of recreational intensity (i.e., May, September, weekends, mid-day) among months and days for
either sex in each cover type in each region.
126
Table 3.6. Distances of key cover types (i.e., regenerating aspen, openings, northern hardwoods/maple) and elk locations to the nearest
road and trail (i.e., multi-use trail [trail], biking trail [bike], equestrian trail [horse]) in the Atlanta State Forest (ASF) and Pigeon River
Country (PRC) State Forest in the northern lower peninsula of Michigan. Elk locations were recorded from GPS-collared elk (n=53,
25 PRC [13 cows, 12 bulls], 28 ASF [14 cows, 14 bulls]) from 1–May to 30–September, 2016–2018.
ASF PRC
Dist. (road/trail) Dist. (road) Dist. (trail) Dist. (bike) Dist. (horse)
Cover type/sex N1 median2 N1 median 2 median 2 median 2 median 2
Aspen (regenerating) 8,073 84 33,094 170 1,167 2,748 3,915
Cow 3,663 134 20,830 201 230 1,211 711
Bull 7,834 103 10,311 200 789 1,667 1,044
Openings 22,421 109 50,406 373 1,411 2,276 3,073
Cow 18,565 76 63,395 381 1,071 1,906 1,256
Bull 55,647 61 30,551 368 1,053 1,917 1,387
Northern hardwoods 77,009 126 222,733 269 1,159 2,051 2,893
Cow 8,035 185 49,457 459 1,211 2,221 1,447
Bull 30,616 181 34,810 302 966 2,151 1,408
1
Number of 30 x 30 m grid cells or elk locations within regenerating aspen, openings, or northern hardwoods/maple cover types.
2
Median distances are the median distance of individual elk median distances (m).
127
Table 3.7. Distances of primary cover types (i.e., regenerating aspen, openings, northern hardwoods/maple) within elk home ranges
(i.e., 95% UD) and elk locations within core areas (i.e., 50% UD) to the nearest road and trail (i.e., multi-use trail [trail], biking trail
[bike], equestrian trail [horse]) in the Atlanta State Forest (ASF) and Pigeon River Country (PRC) State Forest in the northern lower
peninsula of Michigan. Elk locations were recorded from GPS-collared elk (n=53, 25 PRC [13 F, 12 M], 28 ASF [14 F, 14 M]) from
1–May to 30–September, 2016–2018.
ASF PRC
Dist. (road/trail) Dist. (road) Dist. (trail) Dist. (bike) Dist. (horse)
Cover type/sex N1 median2 N1 median 2 median 2 median 2 median 2
Openings
Cow elk 95% UD 14,658 68 42,610 427 922 1,881 1,100
Cow elk 50% UD 12,163 71 38,120 386 985 2,029 1,238
Bull elk 95% UD 37,851 60 23,247 354 941 2,027 1,281
Bull elk 50% UD 34,969 63 16,346 418 1,324 1,860 1,467
Regenerating aspen
Cow elk 95% UD 3,135 107 11,039 216 257 1,406 874
Cow elk 50% UD 2,520 144 14,629 223 280 1,283 870
Bull elk 95% UD 5,749 93 7,540 193 375 1,153 1,068
Bull elk 50% UD 4,308 78 6,429 207 610 1,940 1,008
Northern Hardwoods
Cow elk 95% UD 12,621 121 57,593 410 1,005 2,500 1,154
Cow elk 50% UD 3,596 163 23,555 411 1,183 2,034 1,509
Bull elk 95% UD 43,760 159 53,212 282 884 2,085 1,212
Bull elk 50% UD 13,699 173 15,145 299 755 1,893 1,372
1
Number of 30 x 30 m grid cells or elk locations within regenerating aspen, openings, or northern hardwoods/maple cover types.
2
Median distances are the median distance of individual elk median distances (m).
128
DISCUSSION
Our results demonstrated home range-scale changes in elk space-use and resource selection
patterns in response to peak periods of summer trail-based recreation in northern lower
Michigan. We found no evidence of landscape-level elk avoidance of areas with recreational
activity during our study. Our results were similar to those for red deer which found that
recreational use strongly affected selection of resources within home ranges but had no effect on
selection of home ranges (Coppes et al. 2017a). Our prediction of greater monthly elk home
range sizes and daily movement distances during peak periods of recreational intensity (i.e.,
May, September, weekends) was partially supported by our findings of greater monthly home
range sizes in May than June–September and greater daily movement distances during
weekends. Williamson et al. (in review) found that recreational intensity of ORV users in the
ASF during May was 1.6–2 times greater than June–August, 2016–2018. In the PRC, intensity of
equestrian users during May was 3.7–5.9 times greater than June–August (Williamson et al. in
review). Notably, 42% of our longest daily linear distances traveled by elk occurred in May. Our
observed differences in home range sizes among months were evident in both regions but were
most evident in the ASF where cow and bull home ranges were 1.5–1.7 and 1.6–2 times larger in
May than June–September, respectively (Table 3.3). Although we did not find greater home
range sizes during September (i.e., peak month for summer recreational intensity [Williamson et
al. in review]) for either sex or region, elk may have modified movement patterns to use areas
with less recreational use during May or habituated to predictable recreational activity
throughout the summer months (Lyon and Ward 1982, Thompson and Henderson 1998, Taylor
and Knight 2003). Hence, our findings may suggest elk space-use patterns were, in part, affected
by the late-spring resurgence of summer trail-based recreational users in the elk range.
129
Although greater home range sizes and daily movement distances in May could be due to
changes in movement and resource selection related to calving and spring green-up (Beyer 1987,
Walsh 2007, Lehman et al. 2016), we found no studies that reported greater home range sizes or
movement distances by bull elk specifically in May. Notably, Ruhl (1984) reported that cow
home ranges in Michigan were larger in spring (2,344 ± 134 ha, n = 2) than summer (1,621 ±
638 ha, n = 6), but bull home ranges were larger in summer (3,717 ± 2,331 ha, n = 3) than spring
(3,533 ± 3,754 ha, n = 3). Although bull home ranges were larger in May than June–September
in both regions, bull home range sizes in the PRC were only 47 ha larger in May than June while
being 315 ha larger in the ASF (Table 3.3). The disparity between regions in observed home
range size differences among months may be due to ORV use being prohibited in the PRC.
Williamson et al. (in review) found that the intensity of ORV use in the ASF in May ranged from
1.6–5.3 times greater than June from 2016–2018. Thus, our findings of significantly greater (P <
0.05) home range sizes for bulls in May than June–August in the ASF may be due to increased
recreational intensity of ORV use during May.
Due to the relatively even distribution of openings throughout our study regions, we do not
believe larger home range sizes in May were related to bulls expanding ranges to include new
food sources provided by openings in late spring. Our findings of bulls in the ASF using
proportionally less regenerating aspen and more upland conifers during May and September may
suggest that bulls modified their selection of cover types in response to ORV use (Figure 3.4).
Notably, May and September were the only months that bulls in the ASF used upland conifers
greater than or equal to their availability, respectively. While mature aspen and upland conifer
stands provide thermal cover during summer months, upland conifer stands can also provide
hiding cover (Beyer 1987). Bulls in the ASF may be expanding home ranges to select areas with
130
hiding cover to limit interactions with ORVs. Similarly, our findings of cows in both regions
using proportionally less openings and more regenerating aspen during September may be due to
cows using horizontal cover for hiding while feeding. Prior research on elk habitat selection
patterns in Michigan documented increased use of openings in September (Ruhl 1984, Moran
1973, Beyer 1987). However, we found a decline in proportional use of openings from August to
September for cows and bulls in both regions from 2016–2018 (Figure 3.4). Our observations of
cows switching food sources from openings to regenerating aspen during September in both
regions may have been attributed to increased trail-based recreational intensity. Previous
research has documented increased elk use of hiding cover in response to logging, human
disturbances, and recreational use (Edge and Marcum 1985, Buchanan et al. 2014, Wisdom et al.
2018). Williamson et al. (in review) found that overall intensity of summer trail-based recreation
(i.e., equestrian use, hiking, mountain biking, ORV use) in the ASF and PRC was greatest in
September from 2016–2018. Although other types of recreation (e.g., hunting, wildlife viewing)
occur frequently in September, recent research within Michigan’s elk range using surveys found
that path activities (i.e., hiking, mountain biking, equestrian use) accounted for more visits (n =
140) than hunting (n = 133) or wildlife viewing (n = 88) from June–November, 2018 (Hunt
2019). Changes in cover type selection during September may also be attributed to rutting
behavior, however, we found no changes in cover type selection patterns that suggested bulls or
cows were selecting cover types to seek or avoid each other during courting.
Although elk home range sizes were not larger during September, greater daily movement
distances on weekends from May–September suggests a consistent response to recreational
activity despite the changes in intensity of use across months. Although we found no differences
(α = 0.05) among days of week for elk daily movement distances, distances from Friday–Sunday
131
were greater and accounted for 50% of our longest daily movement distances. Thus, elk may be
selecting habitat components within their home ranges that are relatively farther from
recreational activities during weekend days than weekdays (i.e., Monday–Thursday).
Although elk exhibited greater daily movement distances during weekends, our findings of
2
no differences in 𝜎𝑚 among days of the week suggest elk travel farther to preferred cover types
but do not exhibit irregular behavior during days of greater recreational intensity (i.e., weekends)
or at times of the day when elk are typically inactive (i.e., 0:00–4:00, 10:00–16:00). Elk use of
preferred cover types for food and cover remained relatively consistent across days of the week,
which we attributed to minimal overlap of the crepuscular behavior of elk and typical peak
activity periods of trail-based recreation during 2016–2018. Williamson et al. (in review) found
that 61% of recreation events occurred between 11:00–17:00 during 2016–2018. Based on our
2
findings of elk 𝜎𝑚 throughout the day, elk exhibited typical crepuscular behavior that resulted in
the lowest mid-day periods of inactivity between 10:00–16:00. Thus, human-elk interactions
were unlikely during periods of the day when recreational users were most active. These findings
may also explain why elk were found at greater distances from roads when in northern
hardwoods/maple during mid-day (i.e., 8:00–18:00) bedding periods and at closer distances to
trails when in regenerating aspen and openings between evening and early morning hours (i.e.,
18:00–8:00).
2
Our findings for hourly elk resource selection patterns and elk 𝜎𝑚 revealed different patterns
of use for preferred food sources during nocturnal and diurnal periods of inactivity. Primary
periods of inactivity for cows and bulls in both regions were between 0:00–4:00 and 10:00–
16:00. However, elk use of preferred cover types (i.e., openings, regenerating aspen) remained
approximately 2–4 times greater than availability between 0:00–4:00 while remaining lower than
132
availability between 10:00–6:00. Thus, elk remained in openings and regenerating aspen during
nocturnal periods of inactivity and no recreational activities, but moved to vegetation types
providing cover (e.g., upland conifers, northern hardwoods) for diurnal periods of inactivity and
increased recreational intensity (Figure 3.6). Our findings were consistent with previous research
in Michigan that found similar crepuscular and nocturnal patterns of elk using openings and
regenerating aspen (Ruhl 1984). Notably, Beyer and Haufler (1994) found that elk use of
openings was 4.5 times greater during nocturnal than diurnal periods. Although our findings may
suggest elk are not using preferred food sources due to mid-day periods of increased recreational
intensity, previous research documented that forest stands with greater canopy cover reduces
direct sunlight thereby providing thermal relief for elk during higher mid-day summer
temperatures (Marcum 1975, Skovlin et al. 2002). In Custer State Park in South Dakota (i.e., a
region with similar summer temperatures and humidity), elk selected bed sites that favored
conditions (i.e., greater canopy cover and lower ambient temperatures) providing thermal cover
instead of hiding cover in an area with an extensive network (i.e., 341 km) of roads and trails
(Millspaugh et al. 1998). However, Coppes et al. (2017a) found that red deer avoided preferred
summer food sources in areas where recreation occurred. Hence, similar to Ruhl (1984), we
believe elk may be using mature hardwood and conifer stands between 10:00–6:00 for the
combined benefits of thermal and hiding cover that may limit interactions with recreational
users.
Our prediction of elk using areas with relatively lower habitat suitability during peak periods
of recreational intensity was not supported by our findings. Elk use of areas with greater habitat
suitability during peak periods of recreational intensity may be attributed to the abundance of
high suitability areas distributed throughout the elk range (Williamson et al. 2021). Coppes et al.
133
(2018) found that areas of greater habitat suitability reduced the effects of recreational activity to
the western capercaillie (Tetrao urogallus). Due to the relatively large size and even distribution
of cover types providing areas of food and cover for elk in the PRC and ASF, elk likely did not
have to travel far (e.g., daily movement distances; Table 3.4) within home ranges to remain in
high-suitability areas during intense periods of trail-based recreation. Thus, we attribute the
minimized effects of recreational intensity, in part, to an abundance of cover types that provide
food (e.g., openings, regenerating aspen) and cover (northern hardwoods/maple, upland conifers)
for elk.
Our prediction of elk using areas farther from roads and recreational trails during peak
periods of recreational intensity was not supported by our results. While research has
demonstrated avoidance of roads and trails by elk during periods of increased human activity
(Rowland et al. 2000, McCorquodale 2003, Spitz et al. 2019), others documented elk do not
avoid roads with light hunting pressure and may become habituated to consistent use of roads
and trails (Millspaugh et al. 1998, Baasch et al. 2010). Roads and trails create edges in forested
landscapes which may serve as food sources and travel corridors for elk moving among cover
types (Anderson et al. 2005). Although elk selected for openings, regenerating aspen, and
northern hardwoods closer to equestrian trails at the landscape–scale, we found no evidence of
elk selecting for areas closer to equestrian trails within core areas of their home ranges. Notably,
while differences were not significant (p<0.05), median distances of elk locations to equestrian
trails within openings and northern hardwoods in core areas were greater than median distances
within home ranges for cows and bulls. Furthermore, our observed landscape–scale selection of
areas closer to equestrian trails was likely not due to elk selecting areas of greater suitability that
were coincidentally located near equestrian trails. Mean habitat suitability of areas within 60 m
134
(i.e., mean flight distance for elk in Michigan [Bender et al. 1999]) of an equestrian trail was
0.215, which was lower than the mean habitat suitability within elk home ranges (i.e., cows =
0.406, bulls = 0.323) and for the entire PRC (0.254). Elk selection of home ranges in close
proximity to equestrian trails may be related to use of trails as travel corridors during times of the
day (i.e., 18:00–8:00) when recreational activity is low. Elk may have also demonstrated a
greater tolerance for areas with equestrian use than areas with other types of trail-based
recreation. Naylor et al. (2009) found that elk showed some evidence of habituation to equestrian
use, but no evidence of habituation to mountain biking, hiking, or ORV (i.e., all-terrain vehicle
[ATV]) use. We attribute the selection of home ranges closer to equestrian trails and lack of
avoidance of roads, multi-use trails, and mountain biking trails to: 1) a lack of trail-based
recreation during typical activity periods for elk (Williamson et al. in review); 2) potentially, a
greater tolerance for equestrian use than other types of trail-based recreation (Table 3.6); and 3)
the juxtaposition of trails and roads to an abundance of high suitability areas that provided
preferred cover types.
Our prediction of more pronounced effects from trail-based recreational intensity in the ASF
due to lack of designated trails and ORV use was supported by our results on elk space-use. Cow
and bull home range sizes and daily movement distances were greater in the ASF than the PRC.
Our findings for elk resource selection patterns did not indicate differences between regions. We
believe the lack of differences in elk resource selection between regions provides further support
for mitigation of effects due to an abundance of high habitat suitability and differences in
recreational regulations between regions. The potential differences in elk responses to the PRC’s
designation and restriction of recreational activity to trails may partially explain the differences
in elk space-use patterns between the regions. Off-road vehicle use is prohibited in the PRC and
135
was the most commonly detected type of recreation during summer months in the ASF
(Williamson in review). In Oregon, Priesler et al. (2006) found that elk demonstrated strong
patterns of movements to hiding areas during ORV (i.e., ATV) use from mid-April to October,
and Wisdom et al. (2018) found that ORV (i.e., ATV) use had a greater effect on elk distances to
trails than mountain biking, hiking, and equestrian use, respectively. We attribute greater elk
home range sizes and daily movement distances in the ASF to a lack of designated trails and
presence of ORVs.
136
CHAPTER 4: ELK RESPONSES TO EXPERIMENTAL EQUESTRIAN USE AND
MOUNTAIN BIKING EVENTS ON PUBLIC LANDS IN MICHIGAN
INTRODUCTION
Human-wildlife interactions have become a focus of wildlife management issues in the early
21st century with human population growth and growing participation in outdoor nature-based
recreation (Taylor and Knight 2003, Larson et al. 2016). Although participation in hunting and
fishing has declined in recent decades, participation in other types of outdoor recreation (e.g.
equestrian use, mountain biking, off-road vehicle [ORV] use, snowmobiling, wildlife viewing)
has increased (Cordell 2012). While much research has focused on the negative effects (e.g.,
changes in habitat use, abundance, physiology) of such recreation types on various taxa, few
studies have compared the responses of wildlife to different types of outdoor recreation (Larson
et al. 2016).
Recent studies have implemented experimental forms of recreational activity to measure the
response and effects of these activities on large mammals such as elk and moose (Alces alces)
(Naylor et al. 2009, Neumann et al. 2011, Wisdom et al. 2018). In Oregon, Naylor et al. (2009)
applied 4 types of recreational activities (all-terrain vehicle use [ATV], mountain biking, hiking,
equestrian use) during a 5-day period to 13 GPS radio-collared adult female elk after 14 days of
no human activity. Elk activity increased during all 4 recreation types with ATV use causing the
most impact, followed by mountain biking, hiking, and horseback riding (Naylor et al. 2009).
Similarly, Neumann et al. (2011) exposed 29 adult female moose with GPS radiocollars to off-
trail hiking and snowmobiling activity in northern Sweden. Both experimental treatments led to
increased moose movements lasting 1-2 hrs and increased movement speed 4-8 times. Energetic
costs of moose were estimated to increase by 16% in response to hiking, and by 19% in response
137
to snowmobiling (Neumann et al. 2011). Although both studies demonstrated effects on elk and
moose in response to experimental recreation events, replication of studies comparing different
types of recreational activity on public lands with different recreational regulations and land use
objectives is lacking.
In Michigan, the Pigeon River Country (PRC) State Forest and portions of the Atlanta State
Forest (ASF) Management Units are considered the core of elk range and provide numerous
outdoor recreational opportunities. In the last 50 years, natural resource managers have observed
an increase in trail-based recreation types such as equestrian use and mountain biking (MDNR
2007, MDNR 2012). Mountain biking and equestrian use are among the recreation types that are
projected to increase the most per capita in the U.S. in the next 40 years (Cordell 2012). The
increase in trail-based recreation in the PRC and ASF has led to concerns over the potential
negative effects (e.g., indirect habitat loss, increased human-wildlife conflicts) it may have on
the elk population and habitat (B. Mastenbrook, MDNR, personal communications). Although
the elk population has remained relatively stable (mean=1,065, 95% CI = 931–1,200; S. Adams,
MDNR, personal communication) since 2006, elk occurrences outside of the MDNR’s
designated range have led wildlife managers to hypothesize that elk may be moving outside of
the PRC and ASF in response to periods of increased recreational intensity.
In response to the growing use of public lands in the Michigan elk range by trail-based
recreational users, we studied the behavioral responses of elk to equestrian use and mountain
biking activity during the peak summer month (September) for trail-based recreational activity
(See RESULTS, Chapter 2). Our objectives were to: 1) quantify and compare the behavioral
responses of radiocollared elk to experimental equestrian use and mountain biking events in the
ASF and PRC; 2) document the typical riding behaviors of equestrian users and mountain bikers
138
within the elk range; and 3) provide recommendations to natural resource managers challenged
with balancing objectives for managing elk, their habitat, and trail based recreation on state
forests. Based on previous findings of increased elk activity in response to experimental
equestrian use and mountain biking events (Naylor et al. 2009, Wisdom et al. 2018), we
predicted that: 1) elk would show increased hourly movement behavior in response to
interactions with experimental recreation events; 2) recreation events occurring within the mean
flight distance for elk in Michigan (i.e., 60 m; Bender et al. 1999) would result in elk movements
away from recreation; and 3) movement distance and duration would be greater for mountain
biking than equestrian use (Naylor et al. 2009, Wisdom et al. 2018).
139
METHODS
Experimental Recreation Events
Based on our findings that the greatest recreational intensity of all trail-based recreation in
the PRC and ASF occurred in September (See RESULTS, Chapter 2), we planned experimental
equestrian use and mountain biking events during September, 2018. We contacted members of
the Pigeon River Country Equestrian Committee, Tri-County Horse Association, and Alpena
County Horseman’s Club in early August, 2018 to recruit riders to participate in the study. We
also presented study objectives and methods, distributed informational fliers, and answered
questions at a public equestrian users meeting in the PRC equestrian campground on 28–
September, 2018 to encourage participation. Mountain biking volunteers were recruited by
contacting 7 key individuals that were known to ride in the PRC (S. Whitcomb, MDNR, personal
communication) in July–August, 2018. Volunteer equestrian users and mountain bikers were
encouraged to exhibit “normal” riding behavior (e.g., ride times, ride areas, ride pace, group size)
and were provided with handheld Global Positioning System (GPS; GPSMAP 64s, Garmin
International, Olathe, KS, USA) receivers to carry during rides. Handheld GPS units were
programmed to record date, time, and locations in 1–minute intervals. Volunteers were asked to
record group size and encounters with elk.
Elk Movements and Behavior
We investigated elk activity and movement patterns by monitoring 19 (i.e., PRC = 5 cows, 5
bulls; ASF = 4 cows, 5 bulls) of 34 GPS-collared (Vectronic Aerospace GmbH, Berlin,
Germany; VERTEX Plus Iridium) elk (See RESULTS, Chapter 3) during experimental
recreation events. Monitored elk were chosen based on proximity to recreational trails (i.e.,
140
within the mean maximum daily movement distance [1,690 m; See METHODS, Chapter 1])
during August of 2018. Prior to experimental recreation events, GPS collar programming was
modified to record locations in 5-minute intervals to obtain finer resolution elk movements in
relation to recreation events than would be available using our standard 30-minute interval
programming (See METHODS, Chapter 3). We considered elk encounters with recreational
users to be any 5-minute period where elk were within 120 m (i.e., 2 times the average flight
distance [60 m] for elk in Michigan [Bender et al. 1999]) of an equestrian user or mountain biker
carrying a handheld GPS receiver. To quantify elk movements in response to encounters, we
measured the linear movement distance between elk locations at 5-minute intervals from 30
minutes prior to an encounter to 30 minutes after (i.e., 13 5-minute intervals). Linear movement
distances for each 5-minute interval were averaged across all elk, and a one-way analysis of
variance (ANOVA) was used to assess for differences in mean linear movement distances among
5-minute intervals that occurred before, during, and after elk encounters with recreation events.
To identify changes in hourly elk activity in response to encounters with recreational users,
2
we examined elk Brownian motion variance (𝜎𝑚 ) at 1-hour intervals using a dynamic Brownian
bridge movement model (dBBMM; See METHODS, Chapter 3). Based on our previous findings
2
of typical elk inactivity during diurnal periods (See Results, Chapter 3), we expected elk 𝜎𝑚 to
remain relatively low during typical mid-day riding periods (See Results, Chapter 2) with
2
increases in elk 𝜎𝑚 during hours with encounters with recreational users. One-way analysis of
2
variance (ANOVA) was used to assess for differences in elk 𝜎𝑚 among hours of the day
averaged across all elk encounters.
141
RESULTS
Experimental Recreation Events
We recorded 69 (PRC = 65, ASF = 4) equestrian and 3 mountain biking (PRC = 1, ASF = 2)
events from 31–August to 30–September, 2018 (Table C1, APPENDIX C). Mean group sizes
were larger in the ASF (equestrian = 6, mountain biking = 3) than in the PRC (equestrian use =
3, mountain biking = 2). The largest group size for equestrian users was 12 in the ASF and 6 in
the PRC (Table 4.1). Ride duration, distance, and average speed for equestrian users was similar
between the PRC and ASF (Table 4.1). Although the average equestrian use event lasted longer
than mountain biking events in both regions, the average speed of mountain bikers was greater
than equestrian users (Table 4.1). Start times of equestrian trials ranged from 7:43 to 18:55
(median = 11:48) and end times ranged from 9:14 to 20:07 (median = 14:37). Equestrian groups
reported seeing at least one elk during 12 of 65 rides in the PRC and during 3 of 4 rides in the
ASF. Mountain biking groups did not report seeing elk in either region during rides.
Elk Movements and Behavior
During 72 experimental equestrian user and mountain biking events, we only recorded 4
encounters (i.e., a GPS-collared elk within 120 m of a recreation event) that occurred between
equestrian users and the same cow elk (collar ID = 29868) in the PRC (Table 4.2, Figures 4.2–
4.5). Notably, there were only 2 additional equestrian events where riders were within 180 m
(i.e., 3 times the average flight distance for elk in Michigan) of a radio-collared elk, and in both
cases it was the same cow elk as mentioned above (collar ID = 29868). Although we had GPS-
collared elk throughout the PRC, the majority of elk locations were not in close proximity to the
core network of recreational trails in the PRC (Figure 4.6). The closest distance of a mountain
142
biking event to a radio-collared elk was 162 m in the ASF. We were unable to calculate the
average linear distance between 5-minute intervals between equestrian users and cow elk 29868
due to a delay in the cow’s collar receiving the change from 30-minute to 5-minute location
programming. However, the collar received the update for 5-minute programming before our
final recorded encounter that occurred at 17:40 on 9-27-18 (Figure 4.5). We were not able to
2
perform statistical analyses or examine elk Brownian motion variance (𝜎𝑚 ) due to our results
being limited to one elk with 4 encounters that occurred during different collar programming
schedules (30-minute locations, 5-minute locations). However, the range of linear distances
between 30-minute locations occurring from 30 minutes before to 30 minutes after our first 3
encounters was only 1–32 m. The range of linear distances between 5-minute locations from 30
minutes before to 30 minutes after our last encounter was only 8–35 m (Figure 4.2). Thus, we
found no evidence of increased movements for cow elk 29868 in response to encounters within
120 m of equestrian users. Furthermore, there was no change in cover type selection of cow elk
29868 during the 30-minute periods following each encounter.
143
Figure 4.1. Location of experimental equestrian use (n = 69) and mountain biking (n = 3) events
within the Pigeon River Country and Atlanta State Forests from 31–August to 30–September,
2018. User locations were recorded in 1-minute intervals using handheld Global Positioning
System receivers carried during recreation events.
144
Table 4.1. Summary of experimental equestrian use and mountain biking events within the Pigeon River Country (PRC) and Atlanta
State Forests (ASF) from 31–August to 30–September, 2018.
Recreation type/ Group size Duration (hr:min) Distance (km) Speed (kph)
region N x̄ range x̄ SD x̄ SD x̄ SD
Equestrian use
ASF 4 6 2–12 2:35 1:10 9.7 4.8 3.8 0.8
PRC 65 3 2–6 2:31 1:13 11.1 5.1 4.5 0.7
Mountain biking
ASF 2 3 3 0:44 0:14 5.6 1.8 7.6 0.1
PRC 1 2 1:57 13.0 6.7
Table 4.2. Encounters (i.e., any 5-minute period where elk were within 120 m of an equestrian user or mountain biker carrying a
handheld GPS receiver) between equestrian users and elk during experimental recreation events within the Pigeon River Country State
Forest from 31–August to 30–September, 2018.
Group Elk
Event type size Region Date Time Collar ID Sex Distance (m) Cover type
Equestrian1 3 PRC 08-31-2018 11:30 29868 cow 114 opening
Equestrian2 2 PRC 09-02-2018 11:30 29868 cow 42 northern hardwoods
Equestrian3 2 PRC 09-27-2018 08:30 29868 cow 86 regenerating aspen
Equestrian4 2 PRC 09-27-2018 17:40 29868 cow 92 upland conifers
1
Figure 4.2
2
Figure 4.3
3
Figure 4.4
4
Figure 4.5
145
Figure 4.2. Encounter between equestrian users and a cow elk (collar ID = 29868) during an experimental recreation event within the
Pigeon River Country State Forest on 31–August, 2018. Cow elk locations shown occurred in 30-minute intervals from 10:00 to
13:00.
146
Figure 4.3. Encounter between equestrian users and a cow elk (collar ID = 29868) during an experimental recreation event within the
Pigeon River Country State Forest on 2–September, 2018. Cow elk locations shown occurred in 30-minute intervals from 10:00 to
13:00.
147
Figure 4.4. Encounter between equestrian users and a cow elk (collar ID = 29868) during an experimental recreation event within the
Pigeon River Country State Forest on 27–September, 2018. Cow elk locations shown occurred in 30-minute intervals from 07:00 to
10:00.
148
Figure 4.5. Encounter between equestrian users and a cow elk (collar ID = 29868) during an experimental recreation event within the
Pigeon River Country State Forest on 27–September, 2018. Cow elk locations shown occurred in 5-minute intervals from 30 minutes
before to 30 minutes after the closest linear distance (i.e., 92 m at 17:40) between the equestrian users and the cow.
149
Figure 4.6. Locations of GPS-collared elk and elk capture locations in relation to experimental equestrian use (n = 69) and mountain
biking (n = 3) events within the Pigeon River Country and Atlanta State Forests from 31–August to 30–September, 2018.
150
Figure 4.7. Locations of GPS-collared elk in relation to habitat suitability and experimental equestrian use (n = 69) and mountain
biking (n = 3) events within the Pigeon River Country and Atlanta State Forests from 31–August to 30–September, 2018.
151
DISCUSSION
Although we did not find evidence of behavioral responses (i.e., increased movement
2
distances between locations, increased 𝜎𝑚 ) to experimental recreation events during our study,
results were limited to the responses of 1 cow elk to equestrian users in the PRC. We attribute
the lack of encounters between GPS-collared elk and experimental recreation events to the
absence of GPS-collared elk from the core region of equestrian trails in the PRC where people
were riding (Figure 4.6) and limited participation by equestrian users in the ASF and mountain
bikers in both regions. During the time of our recreation events (i.e., 31–August to 30–
September) we only recorded 1 cow elk (in the PRC) and 1 bull elk (in the ASF) within 3 times
the mean flight distance (60 m; Bender et al. 1999) of elk in Michigan to a recreation event.
Although we collared a relatively even distribution of elk throughout the study area, 4 of 11 elk
that were collared in the PRC within relatively close proximity (< 1 km) to areas where
recreation events occurred died (1) or had collar failures (3) before our recreation event period
(31–August to 30–September) (Figure 4.6). Thus, the reduced number of collared elk in the core
region of the PRC where our riding events were concentrated likely limited the number of
encounters between recreational users and elk during our events.
Based on our landscape-level findings of cows and bulls having closer median distances to
equestrian trails while in preferred cover types (i.e., regenerating aspen, openings, northern
hardwoods; See RESULTS, Chapter 3), we do not believe the absence of GPS-collared elk from
areas with recreation events in the PRC during September was due to elk avoiding areas with
recreational trails. Additionally, we found no changes in elk selection of areas for home ranges in
response to increased recreational intensity (See RESULTS, Chapter 3). We did find that areas
within 60 m of an equestrian trail in the PRC had lower habitat suitability (0.215) than the mean
152
habitat suitability of elk home ranges (cows = 0.406, bulls = 0.323) and for the entire PRC
(0.254) (See DISCUSSION, Chapter 3; Figure 4.7). Thus, we believe the absence of GPS-
collared elk from areas with equestrian trails in the PRC was likely due to elk not using areas
with lower habitat suitability.
Although our lack of responses from a cow elk in the PRC during 4 encounters with
equestrian users suggests that equestrian use may not elicit behavioral responses from elk within
the distances they occurred from trails, the small sample size does not provide sufficient data to
evaluate our predictions. Furthermore, the cow elk was in a different cover type (i.e., openings,
northern hardwoods, regenerating aspen, upland conifers) during each encounter, which likely
provided varying amounts of horizontal cover that has been found to partially mitigate the
negative responses of elk and red deer (Cervus elaphus) from interactions with humans (Lyon
1983, Sibbald et al. 2011, Wisdom et al. 2018). For example, the only encounter we recorded
within the mean flight distance (60 m) for elk in Michigan had an encounter distance of 42 m
with the cow being in a mature (80-89 years-old) northern hardwoods stand with hardwoods
regeneration in the understory that likely provided horizontal cover (Table 4.2). In contrast, the
only encounter between the cow and equestrian users with no cover occurred in an opening at a
distance of 114 m, which was nearly 2 times the mean flight distance (60 m) of elk in Michigan.
Thus, ample horizontal cover may have prevented elk responses during the 3 encounters that
were closest to equestrian users (42 m in northern hardwoods, 86 m in regenerating aspen, 92 m
in upland conifers).
Our lack of encounters between equestrian users and elk in the ASF was likely due, in part,
to our findings for greater recreational intensity by equestrian users in the PRC than the ASF
(See RESULTS, Chapter 2). In September 2018, recreational intensity of equestrian users was
153
6.6 times greater in the PRC than the ASF (See Tables 2.4 and 2.10, Chapter 2). During
conversations with volunteers at the public equestrian users meeting on 28–September, 2018,
users communicated their preference for riding in the PRC due to the group camping experience
(i.e., a minimum of 10 people per group with a maximum campground capacity of 100
individuals) and variety of amenities (e.g., fire rings, tables, toilets, potable water, and manure
bunkers). We attribute the disparity in equestrian use participation in our experimental events
between regions to less equestrian use in the ASF and an abundance of equestrian groups (i.e.,
Pigeon River Country Equestrian Committee, Tri-County Horse Association, and Alpena County
Horseman’s Club) in the PRC.
Our lack of volunteer participation from mountain bikers in both regions was primarily due
to a relative lack of use in the PRC and ASF during our study events. Although mountain biking
intensity was greater during September than May–August, intensity of equestrian use was 37
times greater than mountain biking in the PRC and 11 times greater in the ASF (See Tables 2.4
and 2.10, Chapter 2). During conversations with mountain bikers in July of 2018, 6 of 7
volunteer contacts commented that use of the PRC was rare and inconsistent due to riding
elsewhere. Although there are approximately 78 km of mountain biking trails in the PRC, the
Michigan Division of Parks and Recreation provides approximately 2,250 km of mountain
biking trails statewide (MDNR 2018). Notably, there is a 100 km rail trail (North Central State
Trail) that is within 1 km of Interstate-75, which mountain bikers would need to pass if traveling
to the PRC from Interstate-75. Thus, interactions between mountain bikers and elk in the PRC
and ASF are likely to be infrequent and less common than interactions between equestrian users
and elk due to the abundance of opportunities for riding elsewhere.
154
Despite our lack of encounters between recreational users and GPS-collared elk, 18% of
equestrian user event participants reported seeing at least 1 elk that was not collared. Notably, we
participated in an equestrian use event on 28–September, 2018, during which we observed an
uncollared bull elk feeding in an opening approximately 100 m from the forest road that we used.
The bull elk did not flee during the encounter with 6 equestrian users and 2 dogs. However, we
observed a car using the forest road < 5 minutes after our encounter, which elicited a flight
response from the bull. Although anecdotal, our observation was consistent with other reported
sightings by equestrian users during our events. In a similar study in Oregon, elk demonstrated
some evidence of habituation to equestrian use but not mountain biking, hiking, or ATV use
(Naylor et al. 2009). Although our limited results and observations of no changes in elk behavior
or movements in response to encounters with equestrian users during our study suggest that elk
in the PRC may have a tolerance to the consistent presence of equestrian users, we caution
against using results limited to 4 encounters with 1 cow in the PRC and anecdotal reports and
observations.
155
CONCLUSIONS AND MANAGEMENT IMPLICATIONS
The processes we used to develop and integrate habitat suitability and habitat potential
models demonstrated a multifocal approach for understanding the spatiotemporal dynamics of
wildlife habitat. While each model provides a different lens to examine the habitat value for elk,
we believe the integration of habitat suitability and habitat potential models provides added value
for managers. A primary advantage in the simultaneous use of these models is the ability for
managers to identify areas with low habitat suitability and high habitat potential to determine
which areas will respond to management treatments by providing habitat life requisites for elk.
Areas with high suitability can be managed to remain suitable, and areas with low suitability and
high potential can be managed to provide high suitability in later years. For example, mature
aspen stands have low habitat suitability for winter and spring food due to the inability of elk to
reach browse, but have high potential for food if they are clearcut to promote regeneration.
Although elk managers in Michigan currently manage for elk food by promoting regeneration
through harvesting mature stands and maintaining the proportion of aspen for no net loss, the
selection of stands that are cut may be important for managing the spatial distribution of elk
within the PRC and ASF. Selecting mature aspen for harvest in areas near the core of the PRC
and ASF instead of the boundary near private lands may reduce elk movements beyond the
designated elk range. Areas with low habitat potential for elk, especially near the edges of the
range, could become focal areas to meet other wildlife management goals or land uses (e.g.,
recreation opportunities). We also recommend consideration of clearcuts that are greater in size
than the current 0.16 km2 (40-acre) harvest size limitation found in the PRC’s COM. Clear-cuts
greater than 0.16 km2 in size may be more likely to regenerate successfully by dispersing elk
browsing throughout the area.
156
Our findings of consistent temporal patterns of recreational intensity by month, day of week,
and time of day in the PRC and ASF suggests that managers may be able to develop land
management strategies irrespective of recreational use regulations. For example, managers
desiring to provide recreational opportunities for specific recreation types (e.g., equestrian use,
ORV use) can expect similar patterns of weekend and mid-day use regardless of target user type.
Designating trail use for specific types and maintenance or enhancements of existing trail
systems (e.g., PRC) may directly affect recreational intensity and influence which types of
recreation are likely to occur. We encourage natural resource managers to consider trail attributes
(e.g., proximity to camping areas) and amenities (e.g., potable water) when developing land use
plans involving management of recreational use patterns. For example, providing horse-related
amenities such as manure bunkers in camping areas near designated equestrian trails will likely
result in increased intensity of use and spatial partitioning of equestrian users to those areas.
Conversely, multi-use trails, off-trail riding, and a lack of amenities will likely disperse users
across the landscape but result in less recreational use (e.g., ASF). Understanding the intensity,
temporal patterns, and spatial extent of recreational users on public lands is vital for managers
attempting to achieve multiple management objectives for long-term sustainable use and
recreational opportunities. For example, managers could provide trail enhancements and user
amenities to focus or re-direct recreational use to areas where other management objectives (e.g.,
wildlife habitat enhancement, timber management) are less of a priority. State agencies may be
able to use the presence or absence of recreational use regulations (e.g., designated trails) and
amenities to market different potential recreational experiences for the public. Areas with a well-
developed system of designated trails in close proximity to user-related amenities (e.g., PRC)
could be marketed to new outdoor recreational users of the area, or those users looking for a
157
more comfortable and group-friendly experience, while areas that permit off-trail use without
amenities (e.g., ASF) may be marketed to users that desire a more primitive or less structured
experience. We recommend that future research focus on other recreation types (e.g.,
snowmobiling, cross-country skiing) occurring during winter and early spring months when elk
and other species may be nutritionally stressed to further evaluate and compare recreational use
patterns under differing regulations, and evaluate potential impacts to wildlife populations,
communities, and habitat. Consideration of user preferences for multiple user groups is vital for
natural resources managers attempting to understand and manage the intensity, temporal
patterns, and spatial extent of recreational use on public lands while conserving natural resources
for other management objectives.
Increasing participation of trail-based recreational activities creates challenges and
opportunities for wildlife and land managers attempting to achieve multiple management
objectives on public lands around the world. Our findings indicated some changes in home-range
scale space-use patterns and proportional use of habitat components of elk in response to periods
of relatively greater intensity of ORV use. Although we found little evidence to suggest trail-
based recreational activity is having direct or indirect negative effects on the Michigan elk herd,
we attribute our findings to wildlife managers creating an abundance and interspersion of cover
types providing high habitat suitability throughout the elk range. Designating areas for specific
types of trail use provides managers opportunities to limit negative human-wildlife interactions
spatially and temporally. Although we found little evidence that summer trail-based recreation
had any effects on elk space-use and resource selection in the PRC, we suggest managers
alternate use of connector trails seasonally to reduce the frequency and volume of interactions to
decrease the potential for impacts due to future changes in recreational intensity or types of use.
158
For example, temporarily closing a mountain biking or equestrian trail that travels through or
within close proximity to an opening during late-spring (April–May), would likely limit late-
morning or early-evening interactions between elk and mountain bikers or equestrian users.
Concurrently, alternate trails that travel through hardwood or conifer stands could be opened
during late-spring to provide new opportunities for trail users that reduce the likelihood of
interactions with elk. Although these recommendations seek to limit negative interactions
between elk and recreational users, objectives for providing elk viewing opportunities could be
achieved by establishing elk viewing areas at strategic trail locations such as near large openings
that would limit numerous encounters leading to disturbances of elk. Habitat and user
management strategies that limit negative interactions between elk and trail-based recreational
users will maintain long-term sustainability of elk, quality habitat, and diverse recreational
opportunities for the future. We also recommend periodic monitoring of all recreation types in
the elk range to identify potential changes in the types, patterns, and cumulative effects of use in
the future. According to the most recent (22-April, 2021) meeting minutes of the Pigeon River
Country Advisory Council, use of campgrounds and areas for recreational use was greater in
spring 2021 than previous years and the trend of increasing use is expected to continue in the
future (PRCAC 2021). Although we found very few ORV users in the PRC during our study
(i.e., 74 events), recent discussions with MDNR personnel revealed that conservation officers are
using ORVs to monitor illegal ORV use in the PRC. Thus, intensity of ORV use in the PRC has
likely increased since the completion of data collection period in 2018.
Although we demonstrated consistent use of the PRC’s designated equestrian trails during
September 2018, the absence of collared elk from areas with equestrian use and mountain biking
events during our trials limited our results to 4 encounters between elk and equestrian users. Our
159
limited results showing elk tolerance for equestrian use was consistent with findings in other
research. Based on our limited results being consistent with other research for encounters
between elk and equestrian users and other research demonstrating negative elk responses to
mountain biking and ORV use, we believe natural resource managers can use the varying
responses of elk to different types of recreation when managing recreational use on public lands.
For example, designated trails for ORV use and mountain biking may be placed, managed, and
promoted in areas with low habitat potential to decrease the frequency of negative interactions
with elk. Although we do not recommend placing equestrian trails in areas of high elk habitat
suitability, areas of medium suitability may be considered to increase opportunities for wildlife
viewing and limit the potential for negative effects to elk. Although we found no evidence of elk
responses to equestrian or mountain biking events, we caution against drawing conclusions based
on our experimental recreation results and observations. We recommend future research that
focuses on quantifying and comparing elk responses to a greater variety of recreation types (e.g.,
ORV use, snowmobiling) that have been found to have different effects on elk and other large
ungulates. We also recommend monitoring recreation events throughout the year to monitor elk
responses to recreational activity during different seasons. For example, monitoring
snowmobiling during winter when horizontal cover is limited and elk are more nutritionally
stressed may provide insights on the value of hiding cover for elk during encounters with
recreational users. Identifying differences in how elk respond to a variety of recreation types
throughout the year would inform wildlife managers attempting to develop elk habitat
management strategies and predict population responses.
160
APPENDICES
161
APPENDIX A
Elk Collaring and Capture Data
162
Table A1. Elk collaring and capture data from 3 capture events in the Pigeon River Country (PRC) and Atlanta State Forests (ASF)
from 15–16 February, 2016, 9–April, 2017, and 22–February, 2018.
Collar Collar Ear Capture BC Samples Antibiotic
2
ID frequency tag date Site Latitude Longitude Sex Age score3 Blood Fecal Hair injection4
20172 150.010 77 2/15/16 PRC 45.17597 -84.37998 Cow A 4 Y Y Y Y
20173 150.070 78 2/16/16 ASF 45.08769 -84.30280 Cow A NA Y Y Y N
20174 150.200 54 2/16/16 ASF 45.10657 -84.33894 Bull A 4 Y Y Y Y
20175 150.270 86 2/16/16 ASF 45.10610 -84.31988 Cow A 4.5 Y Y Y Y
20176 150.310 80 2/15/16 PRC 45.20316 -84.39331 Bull Y 7 N N N N
20177 150.330 88 2/15/16 PRC 45.20944 -84.35055 Bull A 3 Y Y Y Y
20177b1 150.330 127/128 4/09/17 ASF 45.08429 -84.35522 Bull A 2 Y Y Y Y
20178 150.360 66 2/15/16 PRC 45.07901 -84.41920 Cow NA 4.5 Y Y Y Y
20179 150.420 52 2/16/16 ASF 45.08781 -84.32435 Cow A 4 Y Y Y Y
20180 150.440 79 2/16/16 ASF 45.09666 -84.34527 Bull NA NA Y Y Y Y
20181 150.460 69 2/16/16 ASF 45.08746 -84.32471 Bull A 4.5 Y Y Y Y
20182 150.490 68 2/16/16 PRC 45.22780 -84.43920 Bull A 4 Y Y Y N
20182b1 150.490 107/124 2/22/18 PRC 45.06119 -84.23547 Bull A 3.5 Y Y Y Y
20183 150.510 89 2/16/16 ASF 45.08611 -84.30277 Bull NA 3 Y Y Y N
20184 150.530 70 2/16/16 ASF 45.10431 -84.33684 Bull Y 4.5 Y Y Y N
20185 150.570 75 2/15/16 PRC 45.10407 -84.37108 Cow A 4 Y Y Y N
20186 150.590 59 2/15/16 PRC 45.20186 -84.39114 Cow Y 3 Y Y Y N
20187 150.610 81 2/16/16 PRC 45.08733 -84.38811 Bull Y 3 Y Y Y N
20188 150.640 64 2/16/16 ASF 45.10521 -84.33646 Cow A 4 Y Y Y Y
20189 150.660 61 2/15/16 PRC 45.10440 -84.37095 Cow A 3.5 Y Y Y Y
20190 150.690 83 2/16/16 PRC 45.17944 -84.51417 Bull A NA Y Y Y N
20191 150.730 60 2/15/16 PRC 45.12155 -84.39676 Cow A 4 Y Y Y Y
20192 151.100 71 2/16/16 ASF 45.10600 -84.33600 Bull Y 2.5 Y Y Y Y
20193 151.120 56 2/16/16 ASF 45.10536 -84.33824 Bull NA 4 Y Y Y Y
20194 151.140 84 2/16/16 ASF 45.08710 -84.30396 Cow A 4.5 Y Y Y Y
20194b1 151.140 139/140 2/22/18 PRC 45.22634 -84.43938 Cow A 3.5 Y Y Y Y
20195 151.160 90 2/16/16 ASF 45.08710 -84.30396 Bull NA 4 Y Y Y Y
20196 151.190 72 2/15/16 PRC 45.07870 -84.41810 Bull Y 4 Y Y Y N
163
Table A1. (cont’d)
20197 151.210 82 2/15/16 PRC 45.10411 -84.37133 Cow A 5 Y Y Y Y
20197b1 151.210 105/106 2/22/18 ASF 45.08611 -84.30297 Cow A 4 Y Y Y Y
20198 151.280 87 2/15/16 PRC 45.12200 -84.39512 Cow NA 5 Y Y Y Y
20199 151.360 74 2/16/16 ASF 45.10411 -84.33537 Bull Y 3 Y Y Y Y
20200 151.450 57 2/16/16 ASF 45.10561 -84.31893 Cow NA 4.5 Y Y Y Y
20201 151.490 73 2/16/16 PRC 45.19777 -84.44694 Bull NA 3 Y Y Y N
20202 151.520 58 2/16/16 ASF 45.10518 -84.33632 Bull Y 4 Y Y Y Y
20203 151.560 76 2/16/16 ASF 45.08718 -84.30396 Cow A 5 Y Y Y Y
20204 151.590 85 2/16/16 ASF 45.08673 -84.30253 Bull NA 4 Y Y Y Y
20205 151.630 55 2/16/16 ASF 45.10442 -84.33601 Cow A 3.5 Y Y Y Y
20206 151.660 51 2/15/16 PRC 45.20165 -84.39032 Cow A 4.5 Y Y Y Y
20207 151.690 62 2/16/16 ASF 45.08762 -84.30214 Bull Y 4 Y Y Y Y
20419 151.720 63 2/16/16 ASF 45.10681 -84.32018 Cow A 4.5 Y Y Y Y
20419b1 151.720 092 2/22/18 ASF 45.08791 -84.30495 Cow A 4 Y Y Y Y
20420 151.760 67 2/16/16 ASF 45.08750 -84.30444 Bull A 3 Y Y Y N
20421 151.790 65 2/15/16 PRC 45.20176 -84.39057 Cow A 4 N Y Y Y
20422 151.830 53 2/15/16 PRC 45.07737 -84.41600 Cow Y 2 Y Y Y Y
23364 151.680 097 2/22/18 ASF 45.08636 -84.29931 Cow A 4 Y Y Y Y
29865 150.040 098 2/22/18 ASF 45.08597 -84.30271 Cow A 4 Y Y Y Y
29866 150.100 141/142 2/22/18 PRC 45.10241 -84.39116 Bull A 3 Y Y Y Y
29867 150.170 133/134 2/22/18 PRC 45.22562 -84.43699 Bull A 4 Y Y Y Y
29868 150.770 100 2/22/18 PRC 45.17760 -84.40125 Bull A 2.5 Y Y Y Y
29869 150.830 093 2/22/18 PRC 45.22626 -84.43648 Cow A 3.5 Y Y Y Y
29871 150.910 135/136 2/22/18 PRC 45.10060 -84.39149 Bull A 3.5 Y Y Y Y
29872 150.870 091 2/22/18 ASF 45.08691 -84.29824 Cow A 4 Y Y Y Y
1
Collar was collected following mortality and refurbished before being placed on elk.
2
Age was categorized as either adult (A) or yearling (Y).
3
A visual body condition score (BC) was used to evaluate the status and condition of the elk (Gerhart et al. 1996).
4
Elk were given intramuscular antibiotic (Flocillin, Bristol Laboratories, Syracuse, N.Y.) injections to minimize risk of infection.
164
APPENDIX B
Elk Collar Events History
Collar Deployments, Elk Mortalities, Collar Failures, Collar Retrievals
165
Table B1. Elk collar events history (i.e., collar deployments, elk mortalities, collar failures, collar retrievals) in northern lower
Michigan from 15–February, 2016 to 2–February, 2020.
Collar Deployment Retrieval1 Elk Mortality Collar Failure
ID date date Date Cause Date Notes
20172 02/15/2016 04/10/2017 Stopped providing GPS fixes
20173 02/16/2016 12/10/2017 12/10/2017 Hunter harvest 05/31/2016 Stopped providing GPS fixes
20174 02/16/2016 05/01/2017 Stopped providing GPS fixes
20175 02/16/2016 03/07/2018 Stopped providing GPS fixes
20176 02/15/2016 02/13/2019
20177 02/15/2016 04/03/2016 03/30/2016 Meningeal worm2
20177b 04/09/2017 02/13/2019 09/18/2018 Stopped providing GPS fixes
20178 02/15/2016 02/13/2019
20179 02/16/2016 02/12/2019
20180 02/16/2016 02/12/2018 Stopped providing GPS fixes
20181 02/16/2016 09/25/2017 Stopped providing GPS fixes
20182 02/16/2016 10/01/2016 10/01/2016 Hunter harvest 09/27/2016 Stopped providing GPS fixes
20182b 02/22/2018 02/13/2019
20183 02/16/2016 02/13/2019
20184 02/16/2016 02/12/2019
20185 02/15/2016 01/03/2018 01/03/2018 Hunter harvest 06/25/2017 Stopped providing GPS fixes
20186 02/15/2016 02/12/2019
20187 02/16/2016 12/16/2018 12/16/2018 Hunter harvest
20188 02/16/2016 01/12/2019 01/15/2019 Stopped providing GPS fixes
20189 02/15/2016 07/17/2018 GPS fixes were very intermittent
20190 02/16/2016 11/30/2018 Stopped providing GPS fixes
20191 02/15/2016 02/12/2019
20192 02/16/2016 12/11/2017 12/11/2017 Hunter harvest
20193 02/16/2016 02/12/2019
20194 02/16/2016 12/10/2016 12/10/2016 Illegal kill3
20194b 02/22/2018 02/17/2020 04/25/2019 Stopped providing GPS fixes
20195 02/16/2016 09/01/2018 Stopped providing GPS fixes
20196 02/15/2016 02/12/2019
166
Table B1. (cont’d)
20197 02/15/2016 12/10/2016 12/10/2016 Hunter harvest
20197b 02/22/2018 02/13/2019
20198 02/15/2016 02/13/2019
20199 02/16/2016 02/13/2019
20200 02/16/2016 09/23/2018 Stopped providing GPS fixes
20201 02/16/2016 09/27/2018 Stopped providing GPS fixes
20202 02/16/2016 02/23/2017 Stopped providing GPS fixes
20203 02/16/2016 12/15/2018 12/15/2018 Hunter harvest 04/04/2017 Stopped providing GPS fixes
20204 02/16/2016 12/17/2018 12/17/2018 Hunter harvest 06/04/2018 Stopped providing GPS fixes
20205 02/16/2016 02/12/2019
20206 02/15/2016 11/16/2017 Stopped providing GPS fixes
20207 02/16/2016 02/12/2019
20419 02/16/2016 12/10/2016 12/10/2016 Illegal kill3
20419b 02/22/2018 02/13/2019 12/26/2020 Hunter harvest
20420 02/16/2016 12/19/2018 12/19/2018 Hunter harvest
20421 02/15/2016 02/13/2019
20422 02/15/2016 02/12/2019
23364 02/22/2018 03/06/2018 02/23/2018 Capture related4
29865 02/22/2018 02/13/2019 02/28/2019 Stopped providing GPS fixes
29866 02/22/2018 02/13/2019
29867 02/22/2018 02/13/2019
29868 02/22/2018 02/13/2019 04/26/2019 Stopped providing GPS fixes
29869 02/22/2018 02/13/2019
29871 02/22/2018 02/13/2019
29872 02/22/2018 02/13/2019
1
Collars were collected following mortality events or after programmed drop-off on 15–January, 2019.
2
Mortality due to meningeal worm was confirmed by Michigan Department of Natural Resources (MDNR) wildlife biologist
specialist and pathologist Thomas M. Cooley.
3
Illegal kills were confirmed by MDNR biologists and conservation officers.
4
Cow elk was determined to have died of capture-related causes the day following capture and collaring. A field examination by Chad
Williamson revealed internal bleeding.
167
APPENDIX C
Experimental Recreation Use Events
168
Table C1. Experimental equestrian use and mountain biking events monitored with handheld GPS receivers in the Pigeon River
Country (PRC) and Atlanta State Forests (ASF) from 31–August to 30–September, 2018.
Group Start End Distance Speed Elk
Date Event type Region size time time Duration (km) (kph) sighting Volunteer/notes
8/31/2018 Equestrian PRC 3 10:07 13:37 3:30 12.71 3.63 Yes
8/31/2018 Equestrian PRC 2 10:08 14:37 4:29 12.07 2.69
8/31/2018 Equestrian PRC 3 10:09 12:41 2:32 6.92 2.73
8/31/2018 Equestrian PRC 3 11:30 14:20 2:50 11.75 4.15
8/31/2018 Equestrian PRC 3 14:50 16:46 1:56 6.92 3.58
8/31/2018 Equestrian PRC 3 18:08 20:03 1:55 8.37 4.37
9/1/2018 Equestrian PRC 1 8:32 13:57 5:25 21.24 3.92 Yes
9/1/2018 Equestrian PRC 2 10:16 13:16 3:00 12.39 4.13
9/1/2018 Equestrian PRC 5 10:30 12:35 2:05 7.72 3.71 Yes
9/1/2018 Mtn. Biking ASF 3 10:42 11:37 0:55 6.92 7.55 Jeffrey a
9/1/2018 Mtn. Biking ASF 3 12:00 12:34 0:34 4.35 7.67 Jeffrey a
9/1/2018 Equestrian PRC 5 12:05 15:15 3:10 12.71 4.01
9/2/2018 Equestrian PRC 2 8:18 13:54 5:36 20.60 3.68
9/2/2018 Equestrian PRC 2 8:47 11:25 2:38 12.23 4.64 Yes
9/2/2018 Equestrian PRC 2 9:07 11:45 2:38 11.43 4.34
9/2/2018 Equestrian PRC 4 12:20 13:52 1:32 5.79 3.78
9/2/2018 Equestrian PRC 3 12:29 15:12 2:43 9.50 3.50
9/3/2018 Equestrian PRC 2 7:49 10:37 2:48 11.75 4.20 Yes
9/3/2018 Equestrian PRC 2 10:13 16:03 5:50 24.14 4.14
9/3/2018 Equestrian PRC 3 10:23 11:45 1:22 6.76 4.95
9/17/2018 Equestrian PRC 2 11:13 13:00 1:47 8.53 4.78
9/18/2018 Equestrian PRC 2 11:48 14:11 2:23 11.43 4.79 Yes
9/19/2018 Equestrian PRC 2 9:52 12:18 2:26 10.62 4.37
9/19/2018 Equestrian PRC 2 18:38 19:29 0:51 3.38 3.98
9/21/2018 Equestrian ASF 9 18:35 20:05 1:30 6.28 4.18 Chuck and family
9/22/2018 Equestrian ASF 12 18:40 20:21 1:41 5.95 3.54 Yes Chuck and family
9/25/2018 Equestrian PRC 5 9:31 13:07 3:36 15.61 4.34
9/25/2018 Equestrian PRC 2 10:08 13:44 3:36 18.02 5.01
169
Table C1. (cont’d)
9/25/2018 Equestrian PRC 2 10:20 13:28 3:08 17.54 5.60
9/25/2018 Equestrian PRC 1 14:56 15:41 0:45 3.86 5.15 + 2 dogs
9/25/2018 Equestrian PRC 5 18:40 19:54 1:14 6.44 5.22
9/26/2018 Equestrian PRC 3 9:54 11:32 1:38 8.85 5.42 Yes
9/26/2018 Equestrian PRC 2 11:51 13:55 2:04 9.33 4.52
9/26/2018 Equestrian PRC 2 12:26 16:46 4:20 15.61 3.60
9/26/2018 Equestrian PRC 2 13:26 16:01 2:35 10.14 3.92 Jeff and Christine
9/26/2018 Equestrian PRC 3 14:18 17:16 2:58 19.15 6.46
9/26/2018 Equestrian PRC 3 14:57 16:17 1:20 7.56 5.67 Jane
9/26/2018 Equestrian PRC 2 15:08 16:17 1:09 4.83 4.20
9/26/2018 Equestrian PRC 3 18:55 19:39 0:44 3.70 5.05
9/27/2018 Equestrian PRC 2 7:43 9:14 1:31 6.44 4.24 Jeff and Christine
9/27/2018 Equestrian PRC 6 10:42 14:20 3:38 18.51 5.09
9/27/2018 Equestrian PRC 6 11:05 14:23 3:18 17.54 5.32 Yes
9/27/2018 Equestrian PRC 3 15:11 15:55 0:44 3.54 4.83
9/27/2018 Equestrian PRC 2 16:59 19:47 2:48 10.94 3.91
9/27/2018 Equestrian PRC 2 17:48 19:45 1:57 9.50 4.87 Jeff and Christine
9/27/2018 Equestrian PRC 3 18:11 19:41 1:30 8.69 5.79
9/27/2018 Equestrian PRC 4 18:38 19:11 0:33 2.90 5.27
9/28/2018 Equestrian PRC 2 8:14 11:07 2:53 11.10 3.85 Jane
9/28/2018 Equestrian PRC 3 14:40 17:33 2:53 14.64 5.08
9/28/2018 Equestrian PRC 4 15:23 17:15 1:52 9.50 5.09
9/28/2018 Equestrian PRC 6 15:32 17:05 1:33 6.44 4.15 Yes Darlene + 2 dogs a
9/28/2018 Equestrian PRC 1 16:10 18:31 2:21 9.33 3.97 + 2 dogs
9/28/2018 Equestrian PRC 2 16:24 19:01 2:37 11.27 4.31 Jane
9/28/2018 Equestrian PRC 2 17:31 19:18 1:47 6.60 3.70
9/28/2018 Equestrian PRC 2 17:34 19:48 2:14 9.98 4.47 Yes Jeff and Christine
9/28/2018 Equestrian PRC 3 17:45 19:48 2:03 9.98 4.87
9/29/2018 Equestrian PRC 2 9:26 15:41 6:15 24.94 3.99 Jane
9/29/2018 Equestrian PRC 2 9:43 13:21 3:38 18.02 4.96 Yes
9/29/2018 Equestrian PRC 2 9:49 13:34 3:45 17.70 4.72
9/29/2018 Equestrian PRC 5 10:25 13:26 3:01 14.81 4.91
170
Table C1. (cont’d)
9/29/2018 Equestrian PRC 6 11:10 12:19 1:09 5.15 4.48 Darlene + 2 dogs
9/29/2018 Equestrian PRC 3 11:17 14:07 2:50 13.04 4.60
9/29/2018 Equestrian PRC 6 12:32 15:37 3:05 14.32 4.65 Darlene + 2 dogs
9/29/2018 Equestrian PRC 1 14:36 15:37 1:01 4.67 4.59
9/29/2018 Equestrian ASF 2 15:41 19:09 3:28 16.25 4.69 Yes Marv and Donna
9/29/2018 Equestrian PRC 3 16:50 18:45 1:55 10.30 5.37
9/29/2018 Mtn. Biking PRC 2 16:55 18:52 1:57 13.04 6.68 Jeffrey
9/29/2018 Equestrian PRC 2 17:03 20:07 3:04 13.20 4.30 Yes
9/30/2018 Equestrian ASF 2 8:23 12:07 3:44 10.30 2.76 Yes Marv and Donna
9/30/2018 Equestrian PRC 2 8:31 10:45 2:14 8.69 3.89 Jane
9/30/2018 Equestrian PRC 2 9:00 10:21 1:21 7.24 5.36
9/29/2019 Equestrian PRC 2 11:15 13:37 2:22 10.30 4.35 Jeff and Christine
a
Chad Williamson participated
171
APPENDIX D
Metadata
172
Data for this project was stored on 2 external hard drives and a desktop computer. In each
location, a master folder titled “MSU_Elk_ Rec_Project” contains 6 subfolders containing raw
data files (e.g., .csv, .docx, .pdf ) and data analysis files (e.g., .R) . Those folders are:
1) “1–Elk Habitat Suitability and Potential”
This folder contains a subfolder titled “Habitat_Analysis” which contains subfolders titled
“HABITAT_SUITABILITY” and “HABITAT_POTENTIAL” that contain all raw data files
and ArcGIS files used to quantify habitat suitability and potential for Chapter 1.
2) “2–Trail-Based Recreation in the Elk Range”
This folder contains a subfolders titled “Recreation_Data” and “Human_Rec_Analyses”. The
“Recreation_Data” folder contains 3 subfolders (ie.., “2016”, “2017”, “2018”) that each
contain trail camera images that were used to quantify recreational intensity described in
Chapter 2, METHODS and RESULTS. The “Human_Rec_Analyses” folder contains raw
data files and R script files used to create the generalized linear model and perform the
statistical analyses described in METHODS for Chapter 2.
3) “Elk Space-Use and Resource Selection”
This folder contains subfolders titled “Movement_Analyses” and “Resource_Selection”. The
“Movement_Analyses” folder contains subfolders (e.g., “dBBMM”, “Distance_roads_trails”,
“Home_Range_Analysis”) containing raw data files and R script files for the elk space-use
analyses described in Chapter 3, METHODS and RESULTS. The “Resource_Selection”
folder contains subfolders titled “dBBMM_habitat_selection” and
“Landscape_scale_resource_selection” that contain raw data files and R script files for the
elk resource selection analyses described in Chapter 3, METHODS and RESULTS.
173
4) “Elk Behavior in Response to Recreation Events”
This folder contains subfolders titled “Field_Trial_Data” and “Field_Trial_Anlaysis”. The
“Field_Trial_Data” and “Field_Trial_Analysis” folder contains raw data files and R script
files for the recreation events and analyses described in Chapter 4, METHODS and
RESULTS.
5) “ArcGIS”
This folder contains the primary .mxd file titled “Elk_Project_MSU_10.3” and subfolders
containing raw data and .shp files used for all analyses in ArcGIS.
6) “Elk_Collar_Data”
This folder contains the primary database file “Elk Collar Database with Capture Data”
containing the elk capture data, and subfolders containing raw data files for elk locations and
mortality events. The master data file for all locations is located in the “Location_Data”
subfolder and is titled “MASTER_LOCATION_DATA”.
174
APPENDIX E
Outreach and Presentation Experience
175
I attended 17 professional conferences, meetings, and events to promote awareness and
community engagement with my project focusing on elk responses to habitat potential and
human recreation use in the Michigan elk range (Table E1). During these meetings and events, I
presented project objectives, methods and updates regarding findings, plans, project timelines,
and potential management implications and answered questions related to research project
activities and use of information to guide elk management. In 2017, I conducted a children’s
interactive outreach presentation (i.e., facilitated by Alpena Community College and US Fish
and Wildlife Service Outdoor Education Program) at Clear Lake State Park, MI. In 2018, I led an
interactive hike and presentation for Michigan Governor Rick Snyder and family (i.e., facilitated
by MDNR) at the Pigeon River Country State Forest, MI. Additionally, I participated in
recreational activities (e.g., horseback riding, mountain biking) during August of 2018 with user
groups to promote study activities (e.g., monitoring elk responses to human recreation use along
trails and forest roads) and better understand the types of recreation data (e.g., duration and
distance of rides, group size, time of day) being collected during our study. I also developed a
research project Facebook page (i.e., no longer active), which was transitioned to a MSU project
website (https://www.canr.msu.edu/msuelk/) to provide access to project updates, results, and
conclusions.
176
Table E1. Professional conferences, meetings, and events attended to promote awareness and community engagement with my project
focusing on elk responses to habitat potential and human recreation use in the Michigan elk range, from 2016–2020.
Year Presentation/Seminar title Conference/Meeting/Event, location
2020 Collared elk locations during 2016–2018 MDNR1 elk working group meeting, Gaylord, MI
2019 Elk responses to recreational use and habitat potential in Michigan Tri-County Horse Association meeting, Freeland, MI
2019 Elk habitat suitability and potential of public and private lands in MI Michigan Fish and Wildlife Conference, Gaylord, MI
2019 Elk responses to recreational use and habitat potential in Michigan PRCEC2 quarterly meeting, Roscommon, MI
2019 Elk habitat suitability and potential of public and private lands in MI Midwest Fish and Wildlife Conference, Cleveland, OH
2018 Elk responses to recreational use and habitat potential in Michigan PRCEC2 quarterly meeting, Roscommon, MI
2018 Current elk habitat suitability of public and private lands in Michigan Eastern Elk Management Workshop, Lewiston, MI
2017 Quantifying elk habitat suitability and potential in the MI elk range PRCEC2 quarterly meeting, Roscommon, MI
2017 Elk responses to recreational use and habitat potential in Michigan Montmorency County Conservation Club meeting,
Atlanta, MI
2017 Elk responses to recreational use and habitat potential in Michigan PRCEC2 quarterly meeting, Roscommon, MI
2017 Elk responses to recreational use and habitat potential in Michigan MSU FW GSO Student Symposium, East Lansing, MI
2016 Elk responses to recreational use and habitat potential in Michigan MDNR1 Wildlife Division annual meeting,
Roscommon, MI
2016 Elk responses to recreational use and habitat potential in Michigan PRCEC2 quarterly meeting, Roscommon, MI
1
Michigan Department of Natural Resources
2
Pigeon River Country Equestrian Committee
177
LITERATURE CITED
178
LITERATURE CITED
Adams, J. C., and S. F. McCool. 2009. Finite recreation opportunities: The Forest Service, the
Bureau of Land Management, and off-road vehicle management. Natural Areas Journal
49:45–116.
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.
Anderson, D. P., J. D. Forester, M. Turner, J. L Frair, E. H Merrill, D. Fortin, J. S. Mao and M.
S. Boyce. 2005. Factors influencing female home ranges in elk (Cervus elaphus) in North
American landscapes. Landscape Ecology 20:257–271.
Arlettaz, R., S. Nusslé, M. Baltic, P. Vogel, R. Palme, S. Jenni-Eiermann, P. Patthey, and M.
Genoud. 2015. Disturbance of wildlife by outdoor recreation: allostatic stress response
and altered activity-energy budgets. Ecological Applications 25:1197–1212.
Armstrong, D. P., and M. H. Reynolds. 2012. Modelling reintroduced populations: the state of
the art and future directions. Pages 165–222 in J. G. Ewen, D. P. Armstrong, K. A.
Parker, and P. J. Seddon, editors. Reintroduction biology: integrating science and
management. Wiley Blackwell, Oxford, United Kingdom.
Baasch, D. M., J. W. Fischer, S. E. Hygnstrom, K. C. Vercauteren, A. J. Tyre, J. J. Millspaugh, J.
W. Merchant, and J. D. Volesky. 2010. Resource selection by elk in an agro-forested
landscape of northwestern Nebraska. Environmental Management 46:725–737.
Balmford, A., J. M. H. Green, M. Anderson, J. Beresford, C. Huang, R. Naidoo, M. Walpole,
and A. Manica. 2015. Walk on the wild side: estimating the global magnitude of visits to
protected areas. PLoS Biology 13: e1002074.
Balmford, A., J. Beresford, J. Green, R. Naidoo, M. Wapole, and A. Manica. 2009. A global
perspective on the trends in nature-based tourism. PLoS Biology 7: e1000144.
Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. Fitting Linear Mixed-Effects Models
Using lme4. Journal of Statistical Software 67:1–48.
Beall, R. C. 1974. Winter habitat selection and use by a Western Montana elk herd. Thesis,
University of Montana, Missoula, Montana, USA.
Bender, L. C., Beyer, D. E., Jr., Haufler, J. B. 1999. Effects of short-duration, high-intensity
hunting on elk wariness in Michigan. Wildlife Society Bulletin 27:441–445.
Beyer, D. E., Jr. 1987. Population and habitat management of elk in Michigan. Dissertation,
Michigan State University, East Lansing, Michigan, USA.
179
Beyer, D. E., Jr., and J. B. Haufler. 1994. Diurnal versus 24–hour sampling of habitat use.
Journal of Wildlife Management 58:178–180.
Black River Ranch [BRR]. 2019. BRR homepage. . Accessed
30 Jun 2019.
Blackwell, M., A. Pagoulatos, W. Hu, and K. Auchter. 2009. Recreational demand for equestrian
trail-riding. Agricultural and Resource Economics Review 38:229–239.
Boyle, S. A., and F. B. Samson. 1985. Effects of nonconsumptive recreation on wildlife: a
review. Wildlife Society Bulletin 13:110–116.
Breslow, N. E., and D. G. Clayton. 1993. Approximate inference in generalized linear mixed
models. Journal of the American statistical Association 88:9–25.
Brooks, R. P. 1997. Improving habitat suitability index models. Wildlife Society Bulletin
25:163–167.
Buchanan, C. B., J. L. Beck, T. E. Bills, and S. N. Miller. 2014. Seasonal resource selection and
distributional response by elk to development of a natural gas field. Rangeland Ecology
and Management 67:369–379.
Burger T. L., and J. Kotar. 1999. Field guide to forest habitat types of northern lower Michigan.
Department of Forest Ecology and Management, University of Wisconsin, Madison,
USA.
Byrne, M. E., J. C. McCoy, J. W. Hinton, M. J. Chamberlain, and B. A. Collier. 2014. Using
dynamic Brownian bridge movement modelling to measure temporal patterns of habitat
selection. Journal of Animal Ecology 83:1234–1243.
Campa, H., III. 1989. Effects of deer and elk browsing on aspen regeneration and nutritional
qualities in Michigan. Dissertation. Michigan State University, East Lansing, Michigan,
USA.
Campa, H., J. B. Haufler, and S. R. Winterstein. 1993. Effects of white-tailed deer and elk
browsing on regenerating aspen: A ten-year evaluation. Proceedings from the 25th
International Congress of the International Union of Game Biologists and the 6th
International Symposium Perdix, European Journal of Wildlife Research 48 (1,
Supplement): 304–311.
Canada Creek Ranch (CCR). 2019. CCR homepage.
. Accessed 30 Jun 2019.
Carmean W. H., J. T. Hahn, and R. D. Jacobs. 1989. Site Index Curves for Forest Tree Species in
the Eastern United States. United States Forest Service, North Central Forest Experiment
Station General Technical Report NC–128, St. Paul, Minnesota, USA.
180
Chan, N. W., and C. J. Wichman. 2020. Climate change and recreation: evidence from North
America cycling. Environmental and Resource Economics 76:119–151.
Chandra, A., and E. Thompson. 2000. Does public infrastructure affect economic activity?
Evidence from the rural interstate highway system. Regional Science and Urban
Economics 30:457–490.
Cleland, D. T., J. B. Hart, G. E. Host, K. S. Pregitzer, and C. W. Ramm. 1993. Field guide:
ecological classification and inventory system of the Huron-Manistee National Forests.
United States Forest Service, North Central Experimental Station, St. Paul, Minnesota,
USA.
Coffman, M. S., E. Alyanak, and J. Ferris. 1980. Field guide to habitat type classification system
for Upper Peninsula of Michigan. School of Forestry and Wood Products, Michigan
Technological University, Houghton, USA.
Coltrane, J. A., and R. Sinnott. 2015. Brown bear and human recreational use of trails in
Anchorage, Alaska. Human-Wildlife Interactions 9:132–147.
Coppes, J., F. Burghardt, R. Hagen, R. Suchant, and V. Braunisch. 2017a. Human recreation
affects spatio-temporal habitat use patterns in red deer (Cervus elaphus). PLoS ONE
12:e0175134.
Coppes, J., J. Ehrlacher, R. Suchant, and V. Braunisch. 2017b. Outdoor recreation causes
effective habitat reduction in Capercaillie Tetrao urogallus: a major threat for
geographically restricted populations. Journal of Avian Biology 48:1–12.
Coppes, J., U. Nopp-Mayr, V. Grunschachner-Berger, I. Storch, R. Suchant, and V. Braunisch.
2018. Habitat suitability modulates the response of wildlife to human recreation.
Biological Conservation 227:56–64.
Cordell, H. K. 2012. Outdoor recreation trends and futures: a technical document supporting the
Forest Service 2010 RPA Assessment. General Technical Report SRS–150. U.S.
Department of Agriculture Forest Service, Southern Research Station, Asheville, North
Carolina, USA.
Corner, R. A., D. A. Albert, C. J. DeLain, and M. B. Austin. 1999. Landtype associations of
northern lower Michigan. Michigan Natural Features Inventory. Northern Lower
Michigan Ecosystem Management Project Report 1999–02–1999–07, Lansing, Michigan,
USA.
Cushman, S. A., and K. McGarigal. 2007. Multivariate landscape trajectory analysis: an example
using simulation modeling of American marten habitat change under four timber harvest
scenarios. Pages 119–140 in J. A. Bissonette, and I. Storch, editors. Temporal
Dimensions of Landscape Ecology: Wildlife responses to variable resources. Springer,
New York, New York, USA.
181
Dahl, L. M. 2008. Using forward-looking infrared radiography to estimate elk density and
distribution in eastern Kentucky. Thesis, University of Kentucky Lexington, USA.
Daubenmire, R. 1966. Vegetation: Identification of typal communities. Science 151:291–298.
DeBerti, J. M. 2006. Management plan for elk in Pennsylvania. Pennsylvania Game Commision,
Jersey Shore, PA, USA.
Devlin, D. and W. M. Tzilkowski. 1986. Grass use by elk and white-tailed deer in Pennsylvania:
habitat management. Proc. of the Pennsylvania Academy of Science, 60:51–54.
DeVore, R. M., M. J. Butler, M. C. Wallace, S. L. Liley, A. A. Mertz, S. E. Sesnie, and P. S.
Gipson. 2016. Elk resource selection patterns in a semiarid riparian corridor. Journal of
Wildlife Management 80:479–489.
Duke, D., and M. Quinn. 2008. Methodological considerations for using remote cameras to
monitor the ecological effects of trail users: Lessons from research in Western Canada.,
Pages 441–445 in A. Raschi, and S. Trampetti, editors. Management for Protection and
Sustainable Development: The Fourth International Conference on Monitoring and
Management of Visitor Flows in Recreational and Protected Areas. Florence, Italy:
National Research Council, Institute for Biometeorology.
Dwyer, J. F. 1988. Predicting daily use of urban forest recreation sites. Landscape and Urban
Planning 15:127–138.
Edge, W. D., and C. L. Marcum. 1985. Effects of logging activities on home-range fidelity of
elk. Journal of Wildlife Management 49:741–744.
Felix, A. B., D. P. Walsh, B. D. Hughey, H. Campa, III, and S. R. Winterstein. 2007a. Applying
landscape-scale habitat-potential models to understand deer spatial structure and
movement patterns. Journal of Wildlife Management 71:804–810.
Felix, A. B., D. W. Linden, and H. Campa, III. 2007b. Building and Using Habitat Models for
Assessing Temporal Changes in Forest Ecosystems. Pages 195–217 in J. A. Bissonette,
and I. Storch, editors. Temporal Dimensions of Landscape Ecology: Wildlife responses to
variable resources. Springer, New York, New York, USA.
Felix, A. B., H. Campa, III, K. F. Millenbah, S. R. Winterstein, and W. E. Mortiz. 2004.
Development of landscape-scale habitat potential models for forest wildlife planning and
management. Wildlife Society Bulletin 32:795–806.
Felix-Locher, A., and H. Campa, III. 2010. Importance of habitat type classifications for
predicting ruffed grouse use of areas for drumming. Forest Ecology and Management
259:1464–1471.
Fortin, J. K., K. D. Rode, G. V. Hilderbrand, J. Wilder, S. Farley, C. Jorgensen, and B. G.
Marcot. 2016. Impacts of Human Recreation on Brown Bears (Ursus arctos): A Review
and New Management Tool. PLoS ONE 11:e0141983.
182
Frid, A., and L. Dill. 2002. Human-caused disturbance stimuli as a form of predation risk.
Conservation Ecology 6:11.
Gilbert, J., J. Sausen, and B. Dhuey. 2010. Wisconsin elk habitat suitability analysis. Wisconsin
Department of Natural Resources, Research Report 191. Madison, WI, USA.
Giraudoux, P. 2018. pgirmess: Spatial analysis and data mining for field ecologists. R package
version 1.6.9. . Accessed May 2021.
Godbout, G., and J.-P. Ouellet. 2010. Fine-scale habitat selection of American marten at the
southern fringe of the boreal forest. Écoscience 17:175–185.
Gompper, M. E., R. W. Kays, J. C. Ray, S. D. Lapoint, D. A. Bogan, and J. R. Cryan. 2006. A
comparison of noninvasive techniques to survey carnivore communities in northeastern
North America. Wildlife Society Bulletin 34:1142–1151.
Haines, A. M., M. E. Tewes, L. L. Laack, J. S. Horne, and J. H. Young. 2006. A habitat-based
population viability analysis for ocelots (Leopardus pardalis) in the United States.
Biological Conservation 132:424–436.
Hallmann, K., S. Müller, and S. Feiler. 2014. Destination competitiveness of winter sport resorts
in the Alps: how sport tourists perceive destinations? Current Issues in Tourism 17:327–
349.
Hamr, J. 1988. Disturbance behaviour of chamois in an Alpine tourist area of Austria. Mountain
Research and Development 8:65–73.
Hargis, C. D., and D. R. McCullough. 1984. Winter Diet and Habitat Selection of Marten in
Yosemite National Park. Journal of Wildlife Management 48:140–146.
Harris, G., R. M. Nielson, T. Rinaldi, and T. Lohuis. 2014. Effects of winter recreation on
northern ungulates with focus on moose (Alces alces) and snowmobiles. European
Journal of Wildlife Research 60:45–58.
Heinemeyer, K., J. Squires, M. Hebblewhite, J. J. O’Keefe, J. D. Holbrook, and J. Copeland.
2019. Wolverines in winter: indirect habitat loss and functional responses to backcountry
recreation. Ecosphere 10:e02611.
Hirzel, A. H., and G. Le Lay. 2008. Habitat suitability and niche theory. Journal of Applied
Ecology 45:1372–1381.
Horne, J. S., E. O. Garton, S. M. Krone, and J. S. Lewis. 2007. Analyzing animal movements
using Brownian bridges. Ecology 89:2354–2363.
Hughson, D. L., N. W. Darby, and J. D. Dungan. 2010. Comparison of motion-activated cameras
for wildlife investigations. California Fish and Game 96:101–109.
183
Hunt, A. 2019. Two essays on the economic value of elk viewing in Michigan. Thesis, Michigan
State University, East Lansing, Michigan, USA.
Jayakody, S., A. M. Sibbald, I. J. Gordon, and X. Lambin. 2008. Red deer Cervus elaphus
vigilance behaviour differs with habitat and type of human disturbance. Wildlife Biology
14:81–91.
Johnston, W. F. 1977. Manager’s handbook for northern white cedar in the north central states.
United States Forest Service, North Central Forest Experiment Station General Technical
Report NC–35, St. Paul, Minnesota, USA.
Keller, B., R.A. Montgomery, H. Campa, III, D.E. Beyer, Jr., S.R. Winterstein, L.P. Hansen, and
J.J. Millspaugh. 2015. A review of vital rates and cause-specific mortality of elk (Cervus
elaphus) in eastern North America. Mammal Review 45:146–159.
Kelly, M. J., and E. L. Holub. 2008. Camera trapping of carnivores: trap success among camera
types and across species, and habitat selection by species, on Salt Pond Mountain, Giles
County, Virginia. Northeastern Naturalist 15:249–262.
Kerr, J. T., and J. Cihlar. 2004. Patterns and Causes of Species Endangerment in Canada.
Ecological Applications 14:743–753.
Kindall, J. L., L. I. Muller, J. D. Clark, J. L. Lupardus, and J. L. Murrow. 2011. Population
viability analysis to identify management priorities for reintroduced Elk in the
Cumberland Mountains, Tennessee. Journal of Wildlife Management 75:1745–1752.
Kotar, J. 1986. Soil-habitat type relationships in Michigan and Wisconsin. Journal of Soil and
Water Conservation 41:348–350.
Kotar, J., and T. L. Burger. 2000. Field guide to forest habitat type classification for north central
Minnesota. Terra Silva Consultants, Madison, Wisconsin, USA.
Kranstauber, B., M. Smolla, and A. K. Scharf. 2020. move: Visualizing and analyzing animal
track data. R package version 4.0.2. .
Accessed May 2021.
Kranstauber, B., R. Kays, S. D. LaPoint, M. Wikelski, and K. Safi. 2012. A dynamic Brownian
bridge movement model to estimate utilization distributions for heterogeneous animal
movement. Journal of Animal Ecology 81:738–746.
Ladle, A., T. Avgar, M. Wheatley, and M. S. Boyce. 2017. Predictive modelling of ecological
patterns along linear-feature networks. Methods in Ecology and Evolution 8:329–338.
Larson, C. L., S. E. Reed, A. M. Merenlender, and K. R. Crooks. 2016. Effects of recreation on
animals revealed as widespread through a global systematic review. PLoS ONE
11:e0167259.
184
Larson, M. A., F. R. Thompson, J. J. Millspaugh, W. D. Dijak, and S. R. Shifley. 2004. Linking
population viability, habitat suitability, and landscape simulation models for conservation
planning. Ecological Modelling 180:103–118.
Latif, Q. S., V. A. Saab, K. Mellen-Mclean, and J. G. Dudley. 2015. Evaluating habitat
suitability models for nesting white-headed woodpeckers in unburned forest. Journal of
Wildlife Management 79:263–273.
Leblond, M., C. Dussault, and M. H. St-Laurent. 2014. Development and validation of an expert-
based habitat suitability model to support boreal caribou conservation. Biological
Conservation 177:100–108.
Leege, T. A., and W. O. Hickey. 1977. Elk-snow-habitat relationships in the Pete King Drainage,
Idaho. Idaho Department of Fish and Game, Wildlife Bulletin No. 6, Boise, ID, USA.
Lehman, C. P., M. A. Rumble, C. T. Rota, B. J. Bird, D. T. Fogarty, and J. J. Millspaugh. 2016.
Elk resource selection at parturition sites, Black Hills, South Dakota. Journal of Wildlife
Management 80:465–478.
Lenth, R. 2020. emmeans: Estimated Marginal Means, aka Least-Squares Means. R package
version 1.4.5. . Accessed February 2020.
Lesmerises, F., F. Déry, C. J. Johnson, and M.-H. St- Laurent. 2018. Spatiotemporal response of
mountain caribou to the intensity of backcountry skiing. Biological Conservation
217:149–156.
Li, C., and S. Lin. 2012. Influence of weather conditions on hiking behavior. International
Journal of Biometeorology 56:777–781.
Linden, D. W., H. Campa, G. J. Roloff, D. E. Beyer, and K. F. Millenbah. 2011. Modeling
habitat potential for Canada lynx in Michigan. Wildlife Society Bulletin 35:20–26.
Longshore, K., C. Lowrey, and D. B. Thompson. 2013. Detecting short-term responses to
weekend recreation activity: desert bighorn sheep avoidance of hiking trails. Wildlife
Society Bulletin 37:698–706.
Lyon, L. J., and A. L. Ward. 1982. Elk and land management. Pages 443–477 in J. W. Thomas
and D. E. Toweill, editors. Elk of North America: ecology and management. Stackpole
Books, Harrisburg, PA, USA.
Lyon, L. J. 1983. Road density models describing habitat effectiveness for elk. Journal of
Forestry 81:592–613.
Mallord, J. W., P. M. Dolman, A. F. Brown, W. J. Sutherland. 2007. Linking recreational
disturbance to population size in a ground-nesting passerine. Journal of Applied Ecology
44:185–195.
185
Marcum, C. L. 1975. Summer–fall habitat selection and use by a western Montana elk herd.
Dissertation, University of Montana, Missoula, USA.
Martinka, C. J. 1976. Fire and elk in Glacier National Park. Proceedings of Tall Timbers Fire
Ecological Conference 14:377–389.
Martinson, K., M. Hathaway, C. Ward, and R. Johnson. 2018. Caring for horses during hot
weather. . Accessed 07 July 2020.
McCorquodale, S. M. 2003. Sex-specific movements and habitat use by elk in the Cascade
Range of Washington. Journal of Wildlife Management 67:729–741.
Michigan Department of Natural Resources [MDNR]. 1975. Michigan elk management plan.
Michigan Department of Natural Resources, Lansing, Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 1984. Elk management plan. Michigan
Department of Natural Resources, Lansing, Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 2003. IFMAP/GAP Lower Peninsula land
cover [map]. 1997 – 2001. 1:64,000; created by Forest, Mineral, and Fire Management
Division, MDNR using ERDAS Imagine 8.5.
. Accessed 16 August 2012.
Michigan Department of Natural Resources [MDNR]. 2007. A concept of management for the
Pigeon River country. Michigan Department of Natural Resources, Lansing, Michigan,
USA.
Michigan Department of Natural Resources [MDNR]. 2008. Michigan state forest management
plan. Michigan Department of Natural Resources, Forest, Mineral, and Fire Management
Division and Wildlife Division, Lansing, Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 2012. Michigan elk management plan.
Michigan Department of Natural Resources, Lansing, Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 2016. Land use orders of the director.
. Accessed 06 Jun 2021.
Michigan Department of Natural Resources [MDNR]. 2018a. Michigan Statewide Outdoor
Recreation Plan 2018–2022. Michigan Department of Natural Resources, Lansing,
Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 2018b. The HANDBOOK! of Michigan
Off-Road Vehicle Laws and Riding Basics. Michigan Department of Natural Resources,
Lansing, Michigan, USA.
186
Michigan Department of Natural Resources [MDNR]. 2019. Michigan Elk Digest. Michigan
Department of Natural Resources, Lansing, Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 2020a. MNRTF Grant Total by County as
of 09/20/2019.
Accessed 17 August 2020.
Michigan Department of Natural Resources [MDNR]. 2020b. Pigeon River Country State Forest.
.
Accessed 14 June 2020.
Michigan Department of Natural Resources [MDNR]. 2020c. State Forests.
. Accessed 28
April 2020.
Michigan Weather Service. 1974. Climate of Michigan by stations. Michigan Department of
Agriculture cooperating with NOAA – National Weather Service US Department of
Commerce. East Lansing, MI.
Miller, S. G., R. L. Knight, and C. K. Miller. 1998. Influence of recreational trails on breeding
bird communities. Ecological Applications 8:162–169.
Millspaugh, J. J., K. J. Raedeke, G. C. Brundige, and C. C. Willmott. 1998. Summer bed sites of
elk (Cervus elaphus) in the Black Hills, South Dakota: considerations for thermal cover
management. American Midland Naturalist 139:133–140.
Moran, R. J. 1973. The Rocky Mountain elk in Michigan. Michigan Department of Natural
Resources, Wildlife Division Report 267, Lansing, MI, USA.
National Agricultural Imagery Program [NAIP]. 2012. United States Department of Agriculture
digital ortho quarter quad tile (DOQQs) geotiff data.
. Accessed 10 Feb 2017.
National Oceanic and Atmospheric Administration [NOAA]. 2016. National Centers for
Environmental Information Climate Data Online. Climate records for Gaylord, MI.
. Accessed 28 April 2016.
National Oceanic and Atmospheric Administration [NOAA]. 2018. National Centers for
Environmental Information Climate Data Online. Snow cover records for Gaylord, MI.
. Accessed 10 Feb
2018.
National Oceanic and Atmospheric Administration [NOAA]. 2020. Record of Climatological
Observations. . Accessed 26 February
2020.
187
Natural Resources Conservation Service, United States Department of Agriculture. Web Soil
Survey. . Accessed [10 September 2010].
Naylor, L. M., M. J. Wisdom, and R. G. Anthony. 2009. Behavioral responses of North
American elk to recreational activity. Journal of Wildlife Management 73:328–338.
Neumann, W., G. Ericsson, and H. Dettki. 2009. Does off–trail backcountry skiing disturb
moose? European Journal of Wildlife Research 56:513–518.
Neumann, W., G. Ericsson, and H. Dettki. 2011. The impact of human recreational activities:
moose as a case study. Alces 47:17–25.
Neuvonen, M., E. Pouta, J. Puustinen, and T. Sievänen. 2010. Visits to national parks: Effects of
park characteristics and spatial demand. Journal for Nature Conservation 18:224–229.
Nickelson, T. E., and P. W. Lawson. 1998. Population viability of coho salmon, Oncorhynchus
kisutch, in Oregon coastal basins: Application of a habitat-based life cycle model.
Canadian Journal of Fisheries and Aquatic Sciences 55:2383–2392.
Nielson, R. M., and H. Sawyer. 2013. Estimating resource selection with count data. Ecology
and Evolution 3:2233–2240.
O’Gara, B. W., and R. G. Dundas. 2002. Distribution: past and present. Pages 67–119 in D. E.
Toweill and J. W. Thomas, editors. North American elk, ecology and management.
Smithsonian Institution Press, Washington, D.C., USA.
Outdoor Foundation. 2017. Outdoor Recreation Participation: Topline Report 2017. .
Accessed 10 February 2020.
Outdoor Industry Association. 2017. The Outdoor Recreation Economy: Michigan.
.
Accessed 13 February 2020.
Papouchis, C. M., F. J. Singer, and W. B. Sloan. 2001. Responses of desert bighorn sheep to
increased human recreation. Journal of Wildlife Management 65:573–582.
Patthey, P., S. Wirthner, N. Signorell, and R. Arlettaz. 2008. Impact of outdoor winter sports on
the abundance of a key indicator species of alpine ecosystems. Journal of Applied
Ecology 45:1704–1711.
Pigeon River Country Advisory Council [PRCAC]. 2021. Meeting Minutes (Thursday, 22-April,
2021).
. Accessed 30 Sep 2021.
Pigeon River Country Association. 2020. Discovery Center.
Accessed 22 July 2020.
188
Popp, J. N., T. Toman, F. F. Mallory, and J. Hamr. 2014. A century of elk restoration efforts in
eastern North America. Restoration Ecology 22:723–730.
Preisler, H. K., A. A. Ager, and M. J. Wisdom. 2006. Statistical methods for analysing responses
of wildlife to human disturbance. Journal of Applied Ecology 43:164–172.
R Core Team. 2018. R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.
Raymer, D. F. N. 2000. Effects of elk and white-tailed deer browsing on aspen communities and
wildlife habitat quality in northern lower Michigan: and 18-year evaluation. Dissertation,
Michigan State University, East Lansing, Michigan, USA.
Reilly, M. L., M. W. Tobler, D. L. Sonderegger, and P. Beier. 2017. Spatial and temporal
response of wildlife to recreational activities in the San Francisco Bay ecoregion.
Biological Conservation 207:117–126.
Remacha, C., J. A. Delgado, M. Bulaic and J. Pérez-Tris. 2016. Human disturbance during early
life impairs nestling growth in birds inhabiting a nature recreation area. PLoS ONE
11:e0166748.
Remacha, C., J. Pérez-Tris, and J. A. Delgado. 2011. Reducing visitors’ group size increases the
number of birds during educational activities: Implications for management of nature-
based recreation. Journal of Environmental Management 92:1564–1568.
Rogala, J. K., M. Hebblewhite, J. Whittington, C. A. White, J. Coleshill, and M. Musiani. 2011.
Human activity differentially redistributes large mammals in the Canadian Rockies
national parks. Ecology and Society 16:1–16.
Rossi, S. D., J. A. Byrne, and C. M. Pickering. 2015. The role of distance in peri-urban national
park use: Who visits them and how far do they travel? Applied Geography 63:77–88.
Rowland, M. M., M. J. Wisdom, B. K. Johnson, and J. G. Kie. 2000. Elk distribution and
modeling in relation to roads. Journal of Wildlife Management 64:672–684.
Ruhl, J. D. 1984. Elk movements and habitat utilization in Northern Michigan. Thesis, Michigan
State University, East Lansing, Michigan, USA.
Schaffer-Smith, D., J. J. Swenson, and A. J. Bóveda-Penalba. 2016. Rapid conservation
assessment for endangered species using habitat connectivity models. Environmental
Conservation 43:221–230.
Schägner, J. P., L. Brander, J. Maes, M. L. Paracchini, and V. Hartje. 2016. Mapping
recreational visits and values of European National Parks by combining statistical
modelling and unit value transfer. Journal for Nature Conservation 31:71–84.
189
Schemnitz, S. D. 1996. Capturing and handling wild animals. Pages 106–124 in T. A. Bookhout,
editor. Research and management techniques for wildlife and habitats. Allen Press,
Lawrence, Kansas, USA.
Schirpke, U., C. Meisch, T. Marsoner, and U. Tappeiner. 2018. Revealing spatial and temporal
patterns of outdoor recreation in the European Alps and their surroundings. Ecosystem
Services 31:336–350.
Shilling, F., J. Boggs, and S. Reed. 2012. Recreational system optimization to reduce conflict on
public lands. Environmental Management 50:381–395.
Shively, K. J., A. W. Alldredge, and G. E. Phillips. 2005. Elk reproductive response to removal
of calving season disturbance by humans. Journal of Wildlife Management 69:1073–
1080.
Sibbald, A. M., R. J. Hooper, J. E. McLeod, and I. J. Gordon. Responses of red deer (Cervus
elaphus) to regular disturbance by hill walkers. European Journal of Wildlife Research
57:817–825.
Siegel, S., and N. J. Castellan, Jr. 1988. Nonparametric statistics for the behavioral sciences.
New York: McGraw-Hill, Inc.
Silva, I., M. Crane, P. Suwanwaree, C. Strine, and M. Goode. 2018. Using dynamic Brownian
bridge movement models to identify home range size and movement patterns in king
cobras. PLoS ONE 13:e0203449.
Skovlin, J. M., P. Zager, and B. K. Johnson. 2002. Elk habitat selection and evaluation. Pages
530–555 in D. E. Toweill, and J. W. Thomas, editors. North American elk: ecology and
management. Smithsonian Institution Press, Washington, D.C., USA.
Speigel, L. E., C. H. Huntley, and G. R. Gerber. 1963. A study of the effects of elk browsing on
woody plant succession in northern Michigan. The Jack Pine Warbler 41:68–72.
Spitz, D. B., M. M. Rowland, D. A. Clark, M. J. Wisdom, J. B. Smith, C. L. Brown, and T. Levi.
2019. Behavior changes and nutritional consequences to elk (Cervus canadensis)
avoiding perceived risk from human hunters. Ecosphere 10:e02864.
Stankowich, T. 2008. Ungulate flight responses to human disturbance: a review and meta-
analysis. Biological Conservation 141:2159–2173.
Steven, R., C. J. Pickering, and G. Castley. 2011. A review of the impacts of nature based
recreation on birds. Journal of Environmental Management 92:2287–2294.
Sweeney, J. M., H. M. Steinhoff. 1976. Elk movements and calving as related to snow cover.
Pages 415–436 in HW Steinhoff and J. D. Ives, eds., Ecological impacts of snowpack
augmentation in the San Juan Mountains, Colorado. Colorado State University, Fort
Collins, USA.
190
Taylor, A. R., and R. L. Knight. 2003. Wildlife responses to recreation and associated visitor
perceptions. Ecological Applications 13:951–963.
Telesco, R. L., F. T. Van Manen, J. D. Clark, and M. E. Cartwright. 2007. Identifying sites for
elk restoration in Arkansas. Journal of Wildlife Management 71:1393–1403.
Thiel, D., E. Menoni, J. F. Brenot, L. Jenni. 2007. Effects of recreation and hunting on flushing
distance of Capercaillie. Journal of Wildlife Management 71:1784–1792.
Thiel, D., S. Jenni-Eiermann, R. Palme, and L. Jenni. 2011. Winter tourism increases stress
hormone levels in the Capercaillie Tetrao urogallus. Ibis 153:122–133.
Thomas, J.W., H. Black, Jr., R.J. Scherzinger, and R.J. Pedersen. 1979. Deer and elk. Pages 96–
103 in J.W. Thomas, ed. Wildlife habitats of managed forests, the Blue Mountains of
Oregon and Washington. U.S. Department of Agriculture, Forest Service Handbook No.
553, Washington, D.C., USA.
Thompson, I. D., and P. W. Colgan. 1994. Marten Activity in Uncut and Logged Boreal Forests
in Ontario. Journal of Wildlife Management 58:280–288.
Thompson, M. J., and R. E. Henderson. 1998. Elk habituation as a credibility challenge for
wildlife professionals. Wildlife Society Bulletin 26:477–483.
Thuiller, W., and T. Münkemüller. 2010. Habitat suitability modelling. Pages 77–85 in A. P.
Moller, W. Fiedler, and P. Berthold, ed. Effects of climate change on birds. Oxford
University Press, New York, USA.
United States Department of Agriculture [USDA]. 2019. Soil Survey Staff, Natural Resources
Conservation Service. Web Soil Survey for northern Michigan.
. Accessed 30 Jun 2019.
Van Deelen, T. R., L. B. McKinney, M. G. Joselyn, and J. E. Buhnerkempe. 1997. Can we
restore elk to southern Illinois? The use of existing digital land-cover data to evaluate
potential habitat. Wildlife Society Bulletin 25:886–894.
van der Zande, A. N., J. C. Berkhuizen, H. C. van Latesteijn, W. J. ter Keurs, and A. J.
Poppelaars. 1984. Impact of outdoor recreation on the density of a number of breeding
bird species in woods adjacent to urban residential areas. Biological Conservation 30:1–
39.
Van Horne, B., and J. A. Wiens. 1991. Forest Bird Habitat Suitability Models and the
Development of General Habitat Models. U.S. Fish and Wildlife Service Fish and
Wildlife Research Publication 8, Washington, D.C., USA.
Van Horne, B., and J. A. Wiens. 2015. Managing Habitats in a Changing World. Pages 34–43 in
M. L. Morrison and H. A. Mathewson, ed. Wildlife Habitat Conservation: Concepts,
Challenges, and Solutions. Johns Hopkins University Press, Baltimore, Maryland, USA.
191
Vaske, J. J., M. D. Needham, and R. C. Cline. 2007. Clarifying interpersonal and social values
conflict among recreationists. Journal of Leisure Research 39:182–195.
Verme, L.J. 1965. Swamp conifer deeryards in northern Michigan, their ecology and
management. Journal of Forestry 63:523–529.
Walsh, D. P. 2007. Population estimation and fixed kernel analyses of elk in Michigan.
Dissertation, Michigan State University, East Lansing, Michigan, USA.
Walsh, D. P., C. F. Page, H. Campa III, S. R. Winterstein, and D. E. Beyer Jr. 2009.
Incorporating estimates of group size in sightability models for wildlife. Journal of
Wildlife Management 73: 136–143.
Weitowitz, D. C., C. Panter, R. Hoskin, and D. Liley. 2019. The effect of urban development on
visitor numbers to nearby protected nature conservation sites. Journal of Urban Ecology
5:1–12.
Whittle, J. 2017. Poll recap: favorite riding season. Accessed 07 July 2020.
Wickham, H. 2016. ggplot2: Elegant graphics for data analysis. Springer-Verlag, New York,
USA. . Accessed May 2021.
Wilcove, D. S., D. Rothstein, J. Dubow, A. Phillips, and E. Losos. 1998. Quantifying Threats to
Imperiled Species in the United States. BioScience 48:607–615.
Williamson, C. R., H. Campa III, A. B. Locher, S. R. Winterstein, and D. E. Beyer. 2021.
Applications of integrating wildlife habitat suitability and habitat potential models.
Wildlife Society Bulletin 45:70–84.
Windmuller-Campione, M. A., J. Kotar, and N. M. Nagel. 2015. Habitat Types – What they can
tell us now and in the future. Applied Ecology and Environmental Research 13:893–913.
Winter, P. L., S. Selin, L. Cerveny, and K. Bricker. 2020. Outdoor recreation, nature-based
tourism, and sustainability. Sustainability 81:1–12.
Wisdom, M. J., H. K. Preisler, L. M. Naylor, R. G. Anthony, B. K. Johnson, and M. M.
Rowland. 2018. Elk responses to trail-based recreation on public forests. Forest Ecology
and Management 411:223–233.
Witmer, G. 1990. Re-introduction of elk in the United States. Journal of the Pennsylvania
Academy of Science 64:131–135.
Wolf, I. D., G. Hagenloh, and D. B. Croft. 2013. Vegetation moderates impacts of tourism usage
on bird communities along roads and hiking trails. Journal of Environmental
Management 129:224–234.
192