The Greater Yellowstone Area Mountain Ungulate Project

Home , Abiathar Peak, Absaroka Range, Mount Everts, Mount Norris, Temple Peak, The Thunderer (Wyoming)

2011 ANNUAL REPORT www.gyamountainungulateproject.com

May 2012

Authors: Montana State University – Bozeman Robert Garrott, Professor, Ecology Department, Principal Investigator Jay Rotella, Professor, Ecology Department, Principal Investigator Megan O’Reilly, M.S. Graduate Student Jesse DeVoe, M.S. Graduate Student Carson Butler, Research Technician Elizabeth Flesch, Research Technician Mike Sawaya, Database Contractor

Collaborators: Idaho Fish and Game Hollie Miyasaki, Wildlife Biologist, Idaho Falls Dale Toweill, Wildlife Program Coordinator, Boise

Montana Fish Wildlife and Parks Julie Cunningham, Wildlife Biologist, Bozeman Tom Lemke, Wildlife Biologist, Livingston (retired) Karen Loveless, Wildlife Biologist, Livingston Justin Paugh, Wildlife Biologist, Big Timber Shawn Stewart, Wildlife Biologist, Red Lodge

National Park Service P.J. White, Supervisory Wildlife Biologist, Yellowstone NP, Mammoth WY Sarah Dewey, Wildlife Biologist, Grand Teton NP, Moose, WY

United States Forest Service Jody Canfield, Wildlife Biologist, Gallatin National Forest, Bozeman, MT Rachel Feigley, Wildlife Biologist, Gallatin National Forest, Livingston, MT Andrew Pils, Wildlife Biologist, Shoshone National Forest, Cody, WY Dan Tyers, Wildlife Biologist, Bozeman, MT

Wyoming Game and Fish Department Doug Brimeyer, Wildlife Biologist, Jackson Kevin Hurley, Bighorn Sheep Coordinator, Cody (retired) Doug McWhirter, Wildlife Biologist, Cody

Funding Yellowstone National Park Canon U.S.A., Inc. through the Yellowstone Park Foundation Montana State University Wyoming Game and Fish Department Idaho Fish and Game Department United States Forest Service Wyoming Governor’s Big Game License Coalition Wyoming Wild Sheep Foundation Montana Wild Sheep Foundation 2

Preface It has been an interesting road we have all traveled over the past 2.5 years of this project. Starting with a recognition of the need to gain better knowledge of bighorn sheep and mountain goat ecology and the interactions between the two species, and a vision of putting together a collaborative research project among as many natural resource agencies and professionals as possible, together we have built what I think is a strong foundation for a productive research program that can inform and enhance conservation and management of these two mountain ungulates in the region. Our collaboration now includes 18 professionals representing nearly every natural resource agency in the region that manages bighorn sheep and mountain goats and their habitat. Our financing has expanded from the initial funding provided by Yellowstone National Park and Montana State University to additional grants and/or in-kind contributions from Wyoming, Idaho, and Montana state fish and wildlife agencies, the U.S. Forest Service, as well as contributions from corporate and sportsman organizations. These developments are real milestones and it is great to all be pulling together on this effort. The first two years of work has focused on putting together the collaborations and funding, and in aggregating all available data so that we can benefit from the work that management biologists have put into monitoring their populations over the past 2-3 decades. It has been quite a task to understand all the data from the various sources, and the many differing formats, but we think we have captured nearly all the fundamental data on the 26 bighorn sheep populations as well as the 11 mountain goat survey areas in the greater Yellowstone area. Two students have mined these data for biological insight and have provided our first major research products which are included in this report. With these accomplishments behind us we are now moving into establishing new field research initiatives in a number of study sites throughout the region. This is no small task, starting with securing all the special permits required by state, federal, and university entities! The first graduate student supported by our initiative, Megan O’Reilly, who is responsible for developing and refining occupancy survey methodologies that we hope will lead to enhance habitat models, has completed her field work and is scheduled to defend her thesis within the year. Jesse DeVoe will be building on Megan’s work as the project’s second graduate student, with the goal of building and validating mountain goat habitat models to predict mountain goat range expansion and abundance throughout the GYA. Telemetry studies focused on questions of spatial ecology and population dynamics have been initiated on a number of study areas, with plans in the works to add several more. We are also building our expertise in health, body condition, and disease to develop a solid research program in these topics to help address the ever pressing questions dealing with bighorn sheep die-off’s. Many thanks to all of you who have enthusiastically participated in this venture thus far. I look forward to what we will accomplish together in the year to come. - Bob Garrott

Table of Contents

Project Overview ...... 6 Building Databases for Bighorn Sheep and Mountain Goats in the GYA ...... 7 Introduction ...... 7 Database Design Considerations ...... 7 Point Data Summary ...... 7 Polygon Data Summary ...... 18 Database Summary ...... 20 Mountain Goat Range Expansion ...... 21 Population Trends of Bighorn Sheep and Mountain Goats in the Greater Yellowstone Area ...... 24 Abstract ...... 24 Introduction ...... 25 Background ...... 27 Methods ...... 27 Compilation of Bighorn Sheep and Mountain Goat Data from the GYA ...... 27 Ln-linear Regression Analysis ...... 28 Comparison of Sympatric and Allopatric Bighorn Sheep Population Growth Rates ...... 30 Synthesis of Mountain Ungulate Data from the GYA ...... 31 Results ...... 32 Data Compilation ...... 32 Ln-linear Regression Analysis ...... 35 Comparison of Sympatric and Allopatric Bighorn Sheep Population Growth Rates ...... 35 Synthesis of Mountain Ungulate Data from the GYA ...... 41 Discussion ...... 45 Data Compilation ...... 45 Ln-linear Regression Analysis ...... 46 Comparison of Sympatric and Allopatric Bighorn Sheep Population Growth Rates ...... 48 Synthesis of Mountain Ungulate Data from the GYA ...... 49 Acknowledgements ...... 50 Literature Cited ...... 50 Climatic Variation and Age Ratios in Bighorn Sheep and Mountain Goats in the Greater Yellowstone Area 53 Abstract ...... 53 Introduction ...... 54 Study Area and Methods ...... 56 GYA Study Area Description ...... 56 Collecting and Censoring Data ...... 56 Development of A Priori Climate Covariates ...... 59 Sympatry ...... 63 Region ...... 63 A Priori and Exploratory Models ...... 64 Statistical Analyses ...... 66 Results ...... 67 Response Variable Dataset ...... 67 Winter Bighorn Sheep Dataset ...... 71 Spring Bighorn Sheep Dataset ...... 76 Mountain Goat Dataset ...... 79

Comparison of models using lamb:adult ratio as response variable to models using lamb:ewe ratio as response variable ...... 82 Discussion ...... 83 Winter Bighorn Sheep Analysis ...... 83 Spring Bighorn Sheep Analysis ...... 85 Mountain Goat Analysis ...... 86 Use of management classification data for demography studies...... 86 Climate Change Predictions ...... 87 Further Research ...... 87 Scope of Inference ...... 88 Acknowledgements ...... 88 Literature Cited ...... 88 Development and Testing of Occupancy Surveys for Sympatric Bighorn Sheep and Mountain Goats in Northern Yellowstone ...... 100 Abstract ...... 100 Introduction ...... 101 Methods ...... 104 Study Area ...... 104 Presence-Only Modeling Effort ...... 105 Occupancy Modeling Efforts ...... 110 Results ...... 116 Presence-Only Modeling Effort ...... 116 Formal Occupancy Field Surveys ...... 119 Future Efforts ...... 123 Acknowledgements ...... 123 Literature Cited ...... 123 GPS-VHF Telemetry Studies of Mountain Ungulates in the Greater Yellowstone Area ...... 129 Abstract ...... 129 Introduction ...... 130 Capture and instrumentation Methodology ...... 131 Dual Collar Strategy ...... 132 Multiple Deployment Strategy ...... 135 Study sites ...... 136 Capture Techniques ...... 145 Chemical Immobilization ...... 150 Collaring Technique...... 152 Health and Disease Sampling ...... 154 Field Studies for population dynamics ...... 156 Database Overview ...... 157 Capture and Monitoring Database ...... 158 Spatiotemporal GPS Database ...... 159 Acknowledgements ...... 160 Literature Cited ...... 161

Project Overview

The greater Yellowstone area (GYA) of Montana, Wyoming, and Idaho comprise an important region for bighorn sheep (Ovis canadensis) in North America. However, there has been a gradual increase in the abundance and distribution of non-native mountain goats (Oreamnos americanus) in this region over the past six decades. High densities of mountain goats could negatively affect bighorn sheep because their ranges now overlap substantially. Also, the limited information available indicates potential for dietary overlap in some seasons and dominance of mountain goats over bighorn sheep when foraging in the same areas, which suggests that bighorn sheep may be sensitive to inter-specific competition. In addition, bighorn sheep are well known for their sensitivity to a variety of diseases that can cause episodic die-offs that result in substantial population reductions. Though mountain goat populations do not appear to be susceptible to disease die-offs to any appreciable extent, mountain goats are effective hosts for a variety of parasites and pathogens that also infect bighorn sheep. Thus, information regarding potential competition, disease transfer, and/or displacement of bighorn sheep by mountain goats is a key issue for natural resource managers in this region.

Temperatures across western North America have shown a pronounced warming over the past 50 years and contribute to decreased snow levels and increased drought. Mountain ungulates may be sensitive to these changes in climate through influences on forage availability and quality in alpine and subalpine areas. For example, warmer temperatures could speed up the rate of snow melt and cause the wave of growing vegetation to occur at higher elevations than it would under cool conditions for corresponding times of the year. Also, the peak of forage quality (~30 days after snow melt) could occur over a narrower elevation band under a warmer climate than it would under otherwise similar but cooler conditions. In turn, these changes could influence the abundance and distribution of mountain ungulates, their migration patterns, the degree to which they transmit diseases, and the extent and outcome of competitive interactions.

In the fall of 2009, Yellowstone National Park provided funding to Drs. Robert Garrott and Jay Rotella to initiate comparative studies of sympatric mountain goats and bighorn sheep in the greater Yellowstone ecosystem to gain insight into the ecological interactions between these two species of mountain ungulates and to advance our basic ecological knowledge of these species in the region to inform policy and enhance conservation and management. The vision is to develop a long-term research program that can be sustained through collaboration and partnerships with the natural resource agencies that share responsibility for managing wildlife populations and their habitats in the region.

Building Databases for Bighorn Sheep and Mountain Goats in the GYA Mike Sawaya and Robert A. Garrott

INTRODUCTION

The goals of the database effort are numerous and include the development of a series of compatible databases for storing point, polygon and demography data for bighorn sheep and mountain goats. The 3 main objectives for creating an aggregate point database were to: 1) consolidate all bighorn sheep and mountain goat observations in the GYA, 2) document the expansion of mountain goat range from initial release sites, and 3) model habitat suitability. The 3 main objectives for creating a seasonal herd polygon database were to: 1) document expert knowledge of area biologists, 2) examine seasonal herd ranges in GYA, and 3) compare point and polygon data. The main objective for creating a demography database is to describe and evaluate population trends and recruitment rates (as index by young-adult ratios) of bighorn sheep and mountain goat herds in the GYA.

DATABASE DESIGN CONSIDERATIONS

A detailed description of database fields, structure and relationships can be found in the 2010 Greater Yellowstone Area Mountain Ungulate Collaborator’s Report (Garrott et al. 2010).

POINT DATA SUMMARY

We captured data provided by Yellowstone National Park, Grand Teton National Park, Montana Fish Wildlife & Parks, Wyoming Game & Fish, Idaho Fish & Game, Montana State University 7

and Wild Things Unlimited. Data was also gleaned from Master of Science theses from Colorado State University, Montana State University and University of Montana. We successfully captured a diverse assortment of data types such as aerial and ground locations, GPS and radio-collar data, harvest data, and even observations made opportunistically by hunters and park visitors.

Between 15 January 2010 and 31 December 2011, we captured 27,826 observations of bighorn sheep and mountain goats in the Greater Yellowstone Area Mountain Ungulate Project point database. During this time we captured 21,838 bighorn sheep and 5988 mountain goat point locations. The earliest bighorn sheep observation in the database was made on September 15, 1937 in Grand Teton National Park and the most recent observations were made by Montana State University on September 17, 2011 in Montana. The earliest mountain goat observation was made on June 15, 1947 near the South Fork of Rock Creek in the Beartooth Mountains of Montana and the most recent observations were made by Montana State University in Montana on September 20, 2011.

We captured data from a diverse collection of digital and hard copy files. A total of 5884 observations captured in the database were made from the ground while 6925 were made from the air; the observer’s position could not be determined for 14,924 observations. Radio-collars were used to obtain 3541 records. The only harvest records that have been captured in the database to date are 81 bighorn sheep records provided by WYGF.

Figure 1 – Pie-chart showing the number of bighorn sheep records each agency contributed to the Greater Yellowstone Area Mountain Ungulate Project point database. Observations made spanning the YNP boundary as part of the Northern Range Working Group, funded by YNP, MTFWP, and USFS, were grouped with MTFWP’s data so YNP’s contribution is significantly more than 100 records.

Figure 2 – Pie-chart showing the number of mountain goat records each agency contributed to the Greater Yellowstone Area Mountain Ungulate Project point database.

Figure 3 – Histogram showing the spatial resolution of bighorn sheep records in the Greater Yellowstone Area Mountain Ungulate Project point database. Spatial resolution increases along the X-axis from left to right. Geographic Area= general descriptions, Section=Township/ Range/ Section to nearest Section, 15- Minute Quad=1:50,000 scale topographic map, GIS=taken from GIS layer(s), GPS=recorded with a GPS unit, 7.5 Minute Quad=1:25,000 scale topographic map.

Figure 4 – Histogram showing the spatial resolution of mountain goat records in the Greater Yellowstone Area Mountain Ungulate Project point database. Spatial resolution increases along the X-axis from left to right. Unknown= unknown spatial resolution, Geographic Area= general descriptions, Section=Township/ Range/ Section to nearest Section, Quarter Section=Township/ Range/ Section to nearest Quarter Section, 15- Minute Quad=1:50,000 scale topographic map, GPS=recorded with a GPS unit, 7.5 Minute Quad=1:25,000 scale topographic map.

Figure 5. – Histogram illustrating the seasonal distribution of bighorn sheep records in the Greater Yellowstone Area Mountain Ungulate Project point database.

Figure 6 – Histogram showing the seasonal distribution of mountain goat records in the Greater Yellowstone Area Mountain Ungulate Project point database.

Figure 7 – Temporal patterns in the number of bighorn sheep records in the Greater Yellowstone Area Mountain Ungulate Project point database. The peaks in 1982-83 and 1997-98 are due to graduate research projects on specific herds, Keating and Legg respectively.

Figure 8 – Temporal trends in the number of mountain goat records by year in the Greater Yellowstone Area Mountain Ungulate Project point database. The peak in 1982-83 is due to James Hayden’s graduate research project on the Palisade Herd in Idaho.

The bighorn sheep and mountain goat locations that we captured in our point databse were widely distributed throughout the GYA. We obtained data on both mountain ungulate species from 3 state wildlife management agencies (Idaho, Montana and Wyoming) and 2 federal agencies (Yellowstone NP and Grand Teton NP). We captured locations of bighorn sheep and mountain goats from the Absaroka Range, Beartooth Range, Gallatin Range, Madison Range, Palisades Range and Teton Range. Many of our bighorn sheep observations were obtained in the Southern Absarokas of Wyoming where it is believed there are 4000-5000 bighorns. Many of our mountain goat observations were made during fixed-wing flight surveys in the Northern Absarokas conducted by Montana Fish, Wildlife and Parks and Yellowstone National Park.

Figure 9 – Locations of 21,838 bighorn sheep observations in the Greater Yellowstone Mountain Ungulate point database. Earliest bighorn sheep observation in database was made on September 15, 1937 in Grand Teton National Park, WY and most recent observation was made on September 17, 2011 by Montana State University in Southeast Montana. 13

Figure 10 – Locations of 5988 mountain goat observations in the Greater Yellowstone Mountain Ungulate point database. Earliest mountain goat observation in database was made on June 15, 1947 in South Fork of Rock Creek in the Beartooth Mountains of Montana and most recent was made on September 20, 2011 by Montana State University in Southeast Montana.

Figure 11 – Locations of bighorn sheep (blue) and mountain goat (red) observations in the Greater Yellowstone Mountain Ungulate Project point database.

Figure 12 – Locations of mountain goat observations from May-October (summer n=10,278) and November-April (winter n=11,560) captured in the Greater Yellowstone Mountain Ungulate point database.

Figure 13 – Locations of mountain goat observations from May-October (summer n=4969) and November-April (winter n=1019) captured in the Greater Yellowstone Mountain Ungulate point database.

POLYGON DATA SUMMARY

We captured 56 of Shawn Stewart’s and 76 of Tom Lemke’s (MTFWP) seasonal herd polygons in the Greater Yellowstone Area polygon database. Sixty-seven bighorn sheep and 65 mountain goat polygons were drawn on 7.5 minute USGS topographic maps and converted to shapefiles using ARCGIS (Fig. 13-15).

Figure 14 – Shawn Stewart and Tom Lemke’s bighorn sheep herd polygons that have been captured in the Greater Yellowstone Mountain Ungulate polygon database. Lambing (n=11) and rutting areas (n=1) are coded as light blue, summer range is medium blue (n=17) and winter range is dark blue (n=38).

Figure 15 – Shawn Stewart and Tom Lemke’s mountain goat herd polygons that have been captured in the Greater Yellowstone Mountain Ungulate polygon database. Spring (n=1) and rutting areas (n=1) are coded as light red, summer range is medium red (n=46) and winter range is dark red (n=17).

Figure 16 – Shawn Stewart’s mountain ungulate herd polygons that have been captured in the Greater Yellowstone Mountain Ungulate polygon database. Bighorn sheep are coded as blue and mountain goats are coded as red. Bighorn sheep lambing areas are coded as light blue, summer range is medium blue and winter range is dark blue. Mountain goat rutting areas are coded as light red, summer range is medium red and winter range is dark red. 19

DATABASE SUMMARY

Our aggregation of point and polygon data for bighorn sheep and mountain goats in the GYA is unique in breadth, depth, and scope. The multi-species nature of the research initiative allows for a broader ecological understanding of mountain ungulates than single-species efforts. The inclusion of records both past and present provides a deeper understanding of mountain ungulate ecology than short-term studies. The distribution of data we’ve collected throughout the GYA gives us a much wider ecosystem perspective than smaller-scale studies.

We effectively built a database to store point data for bighorn sheep and mountain goats that was functional, flexible, adaptable, and quality-controllable. In just four months, we captured over 24,000 mountain ungulate observations from 1 federal agency, 3 state agencies, and 3 universities (Figures 1 and 2). Our flexible database design allowed us to capture any type of data that we desired, including locations with extremely low spatial and temporal accuracy. We captured spatial data ranging from points with just geographic names and no geographic coordinates to points with UTMs that had been marked as waypoints on a GPS (Figures 3 and 4). While we knew the date for most observations, we were even able to capture locations with just a year or season associated with them. We captured over 30 years of bighorn sheep data (Figure 5) and 60 years of mountain goat data (Figure 6). To summarize, we were successful in meeting the research initiative’s objectives to consolidate all sheep and goat observations in the GYA, compile data for habitat suitability modeling, and document the expansion of mountain goat range from initial release sites, but we still have much to do.

Even though we were successful at building and populating the point database, a number of data gaps exist which should be filled. We still need to capture more data from the Gallatin Range, the Madison Range and the Southern GYA. Bighorn sheep observations were distributed evenly between summer and winter (Figures 7 and 12). However, the data for mountain goats is strongly skewed towards late-summer and early fall which limits the seasonality of the habitat models that we can build (Figures 8 and 13). Therefore, it would be good to capture mountain goat observations from other months of the year if they exist so that we have a more even distribution of observations throughout the year. Additional historical data is needed to fully document the range expansion of mountain goats in the GYA. Unfortunately, much of this data resides in hard copy documents which are often hard to find and have low spatial and temporal resolution. Fortunately, our database can accommodate data with almost any degree of spatial or temporal resolution so the onus really falls on the biologists to dig out these old records if we are to piece together an accurate account of mountain goat range expansion in the GYA.

Since our efforts have been focused on building the point database, we have only been able to capture seasonal herd polygon data for two biologist’s districts. Shawn Stewart and Tom Lemke were generous enough to spend some days tracing polygons onto 7.5 minute topographic maps. 20

We were able to create a permanent record of Shawn and Tom’s 30+ years of expert knowledge by creating shapefiles of his drawings in ARCVIEW. Developing a full-set of seasonal herd polygons for the GYA is important for a number of reasons: distilling years of expert knowledge, highlighting core-use areas, and comparing polygons to point data. In the future, we will transfer the remaining polygon drawings to GIS, standardize existing polygon layers and/or create new polygons with additional biologists.

MOUNTAIN GOAT RANGE EXPANSION

Mountain goats were released at 9 different locations in the GYA (Figure 16). The first releases took place in Montana in the 1940s (Rock Creek, Stillwater Canyon, W. Fork of the Gallatin River). Releases continued in Montana in the 1950s (Bear Trap, E. Rosebud River, Pine Creek, W. Fork of the Gallatin River and Wolf Creek). Releases were conducted in Idaho in the 1960s (Palisades Creek) and 1970s (Black Canyon). A total of 170 goats were released at the nine different sites: Black Canyon-7, Bear Trap-24, E. Rosebud River-27, Palisades Creek-5, Pine Creek-22, Rock Creek-14, Stillwater Canyon-33, Wolf Creek-27, W. Fork of the Gallatin River- 11.

Figure 15 – Locations of the 9 mountain goat release sites that have been identified by the Greater Yellowstone Mountain Ungulate Project. Sites labeled with year of first release.

We are still trying to uncover early observations of mountain goats from the first several decades after the introductions in Montana, however, records from the most recent 3-4 decades provide a general picture of mountain goat range expansion in the greater Yellowstone area from the initial 9 release sites (Figures 16, 17).

Figure 16 – Locations of 7 sites in Montana where mountain goats were released in the late 1940s and early 1950s (green) and mountain goat locations (red) which illustrate mountain goat range in the Greater Yellowstone Area through 1979.

Figure 17 – Locations of 7 sites in Montana where mountain goats were released in the late 1940s and early 1950s (green) and 2 sites in Idaho where mountain goats were released in the late 1960s and early 1970s and mountain goat locations (red) which illustrate mountain goat range in the Greater Yellowstone Area through 2011.

Population Trends of Bighorn Sheep and Mountain Goats in the Greater Yellowstone Area Elizabeth P. Flesch and Robert A. Garrott

ABSTRACT

Bighorn sheep (Ovis canadensis) and mountain goats (Oreamnos americanus) are important components of the large mammal community in the Greater Yellowstone Area (GYA) and are of considerable public interest. However, foundational ecological research concerning these species is limited. We analyzed historic bighorn sheep and mountain goat population counts collected by management biologists using ln-linear regression to estimate herd growth rates (). The analyzed dataset consisted of 538 bighorn sheep counts since 1971 and 120 mountain goat counts since 1966. Most mountain goat count units experienced a positive growth rate and increased their distributions over recent decades. Bighorn sheep growth rates were more variable among the 26 recognized herd units in the GYA. We used the historic count data to evaluate the hypothesis that sympatry of non-native mountain goats with bighorn sheep adversely affected bighorn sheep populations. This was accomplished by comparing the growth rates of sympatric herds with that of allopatric herds. There was no evidence that sympatric herd growth rates were significantly lower than allopatric herd growth rates. We caution, however, that many counts in consecutive years suggested larger changes in abundance than what would be reasonable to expect from biological processes. We suspect that variability in counts likely reflects varying detection probability and the overall difficulty of counting mountain ungulates. Therefore, conclusions derived from these data should be further evaluated with more detailed demographic studies in the future.

INTRODUCTION

Bighorn sheep (Ovis canadensis) and mountain goats (Oreamnos americanus) are important wildlife species in the Greater Yellowstone Area (GYA) that are of considerable public interest. Over the past 60 years, non-native mountain goat population numbers have increased in areas of the GYA used by native bighorn sheep. Mountain goats have been hypothesized to negatively affect the survival and health of bighorn sheep because they may compete for similar forage, introduce foreign diseases or displace bighorn sheep from an area (Laundré 1994; Simberloff 2000). The GYA is an excellent place to evaluate these predictions because it contains both sympatric (living in the same geographic area) and allopatric (separate ranges) bighorn sheep and mountain goat herds (Figure 1). However, foundational ecological research concerning these herds is limited and in need of future development.

Counts and demographic information regarding these mountain ungulates have been collected over the past several decades by management biologists in the GYA. Historically, state wildlife management agencies and the National Parks have maintained separate databases to manage these data. Therefore, it would be advantageous to aggregate the data across agencies into one standardized relational database. Analysis of these data collectively could potentially provide new biological insights regarding bighorn sheep and mountain goat population dynamics across the GYA. Basic understanding of historical population trends can also provide an important context for future mountain ungulate studies. Therefore, the objectives of this study were to:

1. Aggregate demographic information regarding bighorn sheep and mountain goats in the GYA into one standardized Microsoft Access database for effective data management and analysis;

2. Use ln-linear regression models to describe the overall population trend for each bighorn sheep herd and mountain goat population segment in the GYA;

3. Compare growth rates among herds to evaluate whether the presence of mountain goats adversely affects bighorn sheep populations; and

4. Synthesize count data across the GYA to describe the overall abundance and trend in growth rate of bighorn sheep and mountain goats.

Figure 1 – Locations of 26 bighorn sheep herds and regions containing mountain goats in the GYA. Allopatric bighorn sheep herds are symbolized by light blue circles and sympatric herds by dark blue circles. Mountain goat regions are symbolized in red and labeled by the predominant mountain range in the area.

BACKGROUND

When analyzing data to derive a population trend, growth rate is one of the most important variables to estimate (Sibly and Hone 2002). Rate of change in abundance over time can be used to predict future trends and evaluate environmental stressors and resource competition (Sibly and Hone 2002). The Greek letter lambda, , usually designates the population multiplication rate and is known as the “finite rate of change” or “finite population multiplier” (Sibly and Hone 2002). A lambda less than one describes a population whose size is decreasing, a lambda equal to one describes a population with a constant size over time, and a lambda greater than one describes a population whose size is increasing (Williams et al. 2002).

To estimate the finite rate of change, the natural logarithm of counts can be fitted with a linear regression model that assumes constant growth rate over time (Eberhardt 1987). Linear regression is a simple and convenient model to derive a population trend with data that may include wide variability frequently inherent to direct population counts (Eberhardt 1987). This model has been effective in improving population trend analyses when there is significant variation in counts between consecutive years (Eberhardt 1987) and is the preferable method when variation in a time series is dominated by visibility bias (Humbert et al. 2009). These characteristics make linear regression a suitable tool to estimate the population growth rates of mountain ungulates in the GYA from historical data.

Linear regression has been used successfully to estimate ungulate population trends in previous studies (Eberhardt 1987). Rubin et al. 1998 used linear regression to estimate bighorn sheep population trends from 25 years of annual waterhole count data in the Peninsular Ranges of California. Linear regression was also used to estimate a population trend from 18 years of aerial count data for mountain goats in the Absaroka Mountains of Montana (Swenson 1985).

METHODS

COMPILATION OF BIGHORN SHEEP AND MOUNTAIN GOAT DATA FROM THE GYA

Herd count and demography information were compiled and entered into the GYA Mountain Ungulate Research Initiative Access database. The data were obtained from reports provided by collaborators from the Montana Department of Fish, Wildlife, and Parks (MFWP), Wyoming Game and Fish Department, Idaho Department of Fish and Game, Yellowstone National Park, and Grand Teton National Park. The MFWP data included the Montana Bighorn Sheep Conservation Strategy of 2009 and reports of aerial surveys from individual biologists. The Wyoming Game and Fish Department provided job completion reports, trend count summaries, and classification survey tables for ten Wyoming bighorn sheep herds and two mountain goat count units. The Idaho Department of Fish and Game provided information regarding the Palisades mountain goat count unit. Yellowstone National Park provided aerial survey data of

bighorn sheep and mountain goats, and Grand Teton National Park provided bighorn sheep herd count information.

All of these data were entered manually into the demography table fields in Microsoft Access. Information was standardized by inserting it into the predefined columns of the data table. Bighorn sheep and mountain goat count units that were used as the basis for data management are fundamentally different due to differing data availability regarding the movement dynamics of the two species. Bighorn sheep count units represent relatively distinct population units that were defined by regional biologists based on their understanding of population distributions and movement patterns. However, mountain goat count units are aggregated by hunting districts that are defined primarily along recognized geographic boundaries and the boundary of Yellowstone National Park. It is uncertain how these units relate to the underlying movement dynamics of mountain goat populations.

To effectively analyze mountain ungulate population trends, it was necessary to aggregate demographic information between population segments that were likely indistinct before conducting ln-linear regression analysis. The data from Franc’s Peak and the Dubois-Badlands were combined due to documented movement by animals among those areas. In addition, we aggregated the Wyoming and Idaho data for mountain goats in the Palisades mountain range into one herd unit. We also combined four Montana mountain goat hunting districts (519, 518, 517 and 514) in the Beartooth mountain range.

LN-LINEAR REGRESSION ANALYSIS

After all available demographic data were compiled and standardized, the total count for each year from each count unit was exported into Excel for ln-linear regression analysis. Survey platform (e.g. aerial, ground) can significantly affect herd count data (Gilbert and Grieb 1957), so this variable was considered when analyzing count data for each herd. Some herd data contained a mix of survey platform types over time. When available, aerial surveys were used in the analysis instead of ground counts, because aerial counts generally provide better detection of animals, cover more area, and are considered effective at monitoring population trends (Bender et al. 2003; Festa-Bianchet and Côté 2008). However, ground counts were used in the analyses when aerial data were not available or when management biologists recommended the use of a ground count over an aerial count. Early (pre-1990) ground counts were excluded from the ln- linear regression analysis if the counts appeared to be radically different than the later (post- 1990) aerial counts. If two herd counts from the same survey platform were available for the same year, the higher estimate was used due to imperfect sighting probability (Neal et al. 1993; Gonzalez-Voyer et al. 2001).

Data from the Wyoming Game and Fish Department included classification counts to estimate young-adult ratios and trend counts to estimate relative changes in abundance over time. 28

Whenever possible, we used trend counts instead of classification counts in regression analyses. However, many Wyoming herd count data contained extended periods of time informed by only classification count data. In this situation, classification count data were used in the ln-linear regression analyses, unless the counts were radically lower than the trend count data.

The natural logarithm of each total herd count for each year was calculated using Excel, and a linear regression was completed on each herd time series. The slope of the resulting regression line estimated the instantaneous growth rate. This value was exponentiated to estimate lambda, the finite rate of change. Confidence intervals were calculated by raising “e” (the base of natural logarithms) to the power of the 95% upper and lower limits of “r” (instantaneous growth rate) from the ln-linear regression. The ln-linear regression results for each herd were then displayed in a graph on the natural logarithmic scale. Summary tables and graphs of ln-linear regression details for bighorn sheep and mountain goat data were created. All herd count values used for analysis and completed ln-linear regressions were sent to regional biologists for comment to ensure accurate representation of herd information.

The count data for eight of the analyzed bighorn sheep herds appeared to contain more than one herd growth rate trend during the time series for which data were available. This is not unexpected because bighorn sheep herds seem to trend toward regular population fluctuations of decline and growth (British Columbia Ministry of Water, Land and Air Protection 2004). When this situation took place, the ln-linear regression analysis was completed separately on each time series of data with a different trend. Management biologists also provided insight as to how time series should be analyzed to provide a biological basis for this variation in analysis. This approach was employed so that the ln-linear regression results did not inaccurately represent the herd growth rate with a constant value, when the growth rate had actually varied over time. This methodology produced two or three lambda values for the herds that required more than one ln- linear regression. However, one ln-linear regression was also completed on the entire time series of data for herds with multiple herd growth rates to provide a reference point.

We also evaluated other methods that are used to estimate population trends. The method previously described, ln-linear regression, is also called exponential growth observation error (EGOE) in the literature (Humbert et al. 2009). Ln-linear regression is the traditional method used to estimate population growth rates, and it is most effective when variability in counts is caused by varying detection probability (Humbert et al. 2009). The exponential growth process noise model (EGPN) is an alternative that assumes variability in counts is completely due to process noise, and this method is typically used for population viability analyses (Humbert et al. 2009). Humbert et al. (2009) recommended the use of the exponential growth state-space model (EGSS), which accounts for both varying detection probability and environmental variability. In order to determine if there were significant differences between ln-linear regression and the state-space model for this study, we completed a preliminary analysis of four bighorn sheep herd

time series and two mountain goat count unit time series, using both ln-linear regression and the state-space model (Table 1).

Table 1 – Comparison between ln-linear regression and state-space model (REML) results for two sympatric and two allopatric bighorn sheep herds, as well as two mountain goat count units, found in the Greater Yellowstone Area.

Ln-linear State-space (REML) Count Unit   lower  upper   lower  upper Process Observation CI (95%) CI (95%) CI (95%) CI (95%) Variance Error Bighorn sheep Franc's Peak 1.05 1.02 1.07 1.05 1.00 1.11 0.02 0.13 Complex Darby Mountain 0.92 0.85 0.99 0.95 0.79 1.13 0.01 0.41 (1990 Norris-2008) Peak to 1.14 1.08 1.20 1.12 1.01 1.24 0.04 0.07 Tower Junction Stillwater 0.96 0.93 0.98 0.96 0.93 0.99 0.00 0.02 (1983-1993) Mountain goats Palisades MG 1.06 1.03 1.10 1.06 1.02 1.11 0.01 0.12 MT_323_MG 1.03 1.02 1.03 1.03 0.99 1.08 0.02 0.01

Ln-linear regression and the state-space model produced very similar lambda estimates (Table 1). In addition, the 95% confidence interval estimated through ln-linear regression was included within or was similar to the 95% confidence interval of the state-space model in every count unit analysis (Table 1). The state-space model also estimated that observation error was higher than process variance in every analyzed time series except for one. Therefore, it can be inferred that ln-linear regression, which accounts for observation error only, would likely provide an effective estimate for these data. Due to these preliminary results and the inherent difficulty of counting mountain ungulates in mountainous terrain that causes observation error, we decided to use ln- linear regression for the remaining count unit analyses.

COMPARISON OF SYMPATRIC AND ALLOPATRIC BIGHORN SHEEP POPULATION GROWTH RATES

To compare the population growth rates of bighorn sheep herds that were sympatric or allopatric with mountain goats, a mean growth rate was calculated for each bighorn sheep herd type (sympatric, allopatric) from individual herd lambda estimates. When herd count data were analyzed to have more than one population growth rate, the lambda of the most recent time series was used in the calculation of the overall mean.

In addition, herd counts for bighorn sheep herds that were the sympatric or allopatric with mountain goats were aggregated to analyze synthesized count data, which could provide an additional comparison between the sympatric and allopatric herds that was unrelated to individual herd ln-linear regression results. To achieve an aggregated count, the time series were

combined for bighorn sheep herds within both sympatric and allopatric categories. The criteria for selection of data for the aggregated analysis were slightly different than that used for the individual herd ln-linear regression analysis. This is because every herd required a count of some type for every year in the aggregated time series to accurately represent the overall trend using ln-linear regression.

We felt that actual counts were more ideal than count interpolations (described below) for aggregated analysis because the aggregation of herd data would likely limit the effect of minor variation in counts on the overall population trend. Therefore, years with counts not suitable for the individual herd ln-linear regression were filled with counts that may have been excluded due to poor survey conditions, varying survey platform, and the survey being considered a classification count. However, when a count excluded from the ln-linear regression was over 200 animals lower in number than the count the year before or after, the unsuitable, low count was replaced with an interpolation to avoid unreasonable yearly changes in the data used to estimate overall allopatric and sympatric population trends.

When a survey was not completed for a herd or all counts were unsuitable for analysis in a particular year, the value was interpolated by calculating the mean of the counts in the previous and following year. When there was a series of multiple years without count data for a particular herd, the difference between the counts before and after the missing time series was calculated. This difference was then divided by a value one greater than the number of years with missing values. The quotient was then added to the count before the missing time series to produce the first interpolation. The same quotient was next added to this interpolation to produce the following year’s interpolated value. This method was continued until the missing time series was complete. If herd count data lacked a value at the end of the time series, the most recent count was copied to fill that year. These methods allowed for effective ln-linear regression analysis on aggregated allopatric and sympatric herd data, which were compared with the means of the individual herd population growth rates.

SYNTHESIS OF MOUNTAIN UNGULATE DATA FROM THE GYA To characterize the overall trend of each mountain ungulate species throughout the GYA, the time series for all count units were combined, with missing values interpolated for years when surveys of a unit were not performed. The data were selected and interpolated in the same method as in the comparison of sympatric and allopatric bighorn sheep population growth rates. However, this synthesis aggregated counts and interpolations by region and for the entire GYA.

In addition to analyzing population trends, we estimated the total number of bighorn sheep and mountain goats in the GYA through the synthesis of count data. To provide a consistent basis for this estimation, the most recent year for which data were available for most herd units was selected. Therefore, the counts and interpolations from 2009 were summed by region and across the GYA. We also adjusted the estimate to account for detection probability or visibility bias, 31

which is a common problem for surveys where it is difficult to observe all animals present (Cook and Jacobson 1979).

Detection probability represents the likelihood that an observer will spot an animal, which can vary by species, group size, terrain, survey conditions, and other factors. Mean detection probability for bighorn sheep ewes was estimated to be 0.57 during a study in sagebrush areas in Idaho and 0.58 in a study in Colorado (Bodie et al. 1995; Neal et al. 1993). Udevitz et al. (2006) found an overall detection rate of 0.88 for Dall’s sheep in Alaska. Mean detection probability for mountain goats was estimated to be 0.70 in Alberta and 0.85 in Washington (Gonzalez-Voyer et al. 2001; Rice et al. 2009). Gonzalez-Voyer et al. (2001) found a sightability range of 0.55-0.85 for mountain goats with aerial surveys. Rice et al. (2009) found that detection probability for single mountain goats had a range of 0.75-0.91, and the range of values was caused by variation in terrain. Therefore, we used potential detection probabilities of low (0.55), moderate (0.70), and high (0.90) to estimate a viable range of bighorn sheep and mountain goat population sizes across the GYA.

RESULTS

DATA COMPILATION With the assistance and collaboration of regional biologists, 801 counts of 26 bighorn sheep herds (Table 2) and 11 mountain goat count units (Table 3) were aggregated into the standardized Microsoft Access database. A total of 680 counts described bighorn sheep herds, and 541 bighorn sheep records were used in the individual herd ln-linear regression analyses. Bighorn sheep data contained counts from 1971 to 2011. The average length of time that data were available for a bighorn sheep herd was 26 years. The most recent count for any bighorn sheep herd ranged from 5 to 1054 animals, and the average was 190. In addition, management biologists provided insights regarding bighorn sheep migration patterns and the number of mountain goats found in the same geographic region as sympatric bighorn sheep herds. Ten bighorn sheep herds were classified as allopatric and were mainly located in the southern and eastern GYA. Sixteen herds were classified as sympatric with mountain goats and were predominately located in the northern and western GYA (Figure 1).

A total of 121 records regarding mountain goat count units were entered into the database, and 115 counts were analyzed using ln-linear regression. Mountain goat data contained counts from 1966 to 2011. The average length of time for which data were available regarding a mountain goat count unit was 19 years. The most recent survey count for any mountain goat unit ranged from 29 to 252 animals, and the average was 123 mountain goats. We did not compile information regarding migration patterns of mountain goats because the movement dynamics of this species are not well understood throughout the GYA. Overall, mountain goats seem to predominately occupy the Madison, Gallatin, Absaroka, Beartooth, and Palisades mountain ranges (Figure 1). 32

Table 2 – Summary of geographic location, survey history, recent population counts, migration patterns, and sympatry information for 26 bighorn sheep herds in the GYA. Herds are listed in order of increasing sympatry strength with mountain goats and are grouped by mountain range within sympatry strength categories. Average Most Most Approximate No. of Mountain Herd Recent Recent Sympatry Sympatry Mtn. Herd ID Range Counta Count Year Count Migration Patterns Duration (yr) Strength Goats 1. Temple Peak BS610 Wind River 8 2008 5 Partially migratory 1 Weak 1c 2. Whiskey Mountain BS609 Wind River 576 2010 489 Partially migratory 1 Weak 1c 3. Yount's Peak BS204 Absaroka 500 2009 530 Partially migratory 10 Weak <10b 4. Franc's Peak and Dubois-Badlands Absaroka 819 2008 1054 Partially migratory 20 Weak <5b 5. Wind River Indian Reservation Absaroka 116 2006 130 Partially migratory 20 Weak <5b 6. Trout Peak BS202 Absaroka 267 2009 290 Partially migratory 20 Weak <50b 7. Wapiti Ridge BS203 Absaroka 741 2009 806 Partially migratory 20 Weak <10b 8. Jackson BS107 Gros Ventre 285 2009 309 Partially migratory 10 Weak <5b 9. Darby Mountain BS121 Wyoming 20 2008 28 Non-migratory 20 Weak <10b 10. Targhee BS106 Teton 65 2010 81 Non-migratory 20 Weak <25b 11. Beattie Gulch, Devil’s Slide, and Gallatin 49 2011 54 Seasonally migratory 10 Moderate 7c Cinnabar Mountain 12. Mount Everts and Mammoth Absaroka 78 2011 78 Partially migratory 20 Moderate 6c 13. Travertine and Deckard Flats Absaroka 3 2011 16 Seasonally migratory 20 Moderate 113c 14. Norris Peak and Tower Junction Absaroka 49 2011 70 Partially migratory 20 Moderate <50b 15. Mill Creek Segment Absaroka 20 2010 14 Seasonally migratory 30 Moderate 16c 16. Yankee Jim Canyon, Corwin, and Absaroka 22 2011 37 Seasonally migratory 60 Moderate 20c LaDuke Springs 17. Clark's Fork BS201 Beartooth 318 2009 409 Partially migratory 50 Moderate 130a 18. Black Canyon and Barronette Absaroka 27 2011 34 Partially migratory 40 Strong 50b 19. Monument Peak Absaroka 26 2011 35 Unknown 40 Strong 36a 20. Stillwater Absaroka 50 2011 57 Seasonally migratory 60 Strong 125a 21. Hellroaring Beartooth 12 2011 14 Seasonally migratory 50 Strong 80-100b 22. West Rosebud Beartooth 55 2011 85 Partially migratory 70 Strong 40-50b 23. Tom Miner Basin Gallatin 35 2011 51 Seasonally migratory 50 Strong 37c 24. Point of Rocks, Rock Creek, and Gallatin 37 2011 47 Seasonally migratory 60 Strong 144c Lower Big Creek 25. Hilgards Madison 69 2010 86 Seasonally migratory 60 Strong 25-30b 26. Spanish Peaks Madison 162 2011 140 Seasonally migratory 60 Strong 71c Total 4409 4949 a Average of the five most recent counts. b Management biologist’s estimation. c Most recent count by biologist. 33

Table 3 – Summary of the geographic location, survey history, recent population counts, and sympatry information for 11 mountain goat count units in the GYA. Herds are in order of increasing sympatry strength with bighorn sheep and are grouped by mountain range within sympatry strength categories.

Most Most Mountain Sympatric/ Average Migration Duration in Sympatry Herd ID State Recent Recent Range Allopatric Herd Sizea Patterns Area (yr) Strength Count Year Count 1. Palisades Palisades WY & ID Allopatric 194 2006 252 ? 40 None 2. MT_330_MG Absaroka MT Allopatric 31 2011 29 ? 60 Weak 3. MT_323_MG Absaroka MT Sympatric 146 2011 203 ? 30 Moderate 4. Beartooth MG201 Beartooth WY Sympatric 130 2006 157 ? 60 Moderate 5. MT_329_MG Absaroka MT Sympatric 120 2011 125 ? 40 Strong Absaroka 6. Yellowstone National Park MT & WY Sympatric 150 2010 117 ? 40 Strong and Gallatin 7. MT_314_MG Gallatin MT Sympatric 86 2011 176 ? 50 Strong 8. MT_316_MG Beartooth MT Sympatric 90 2011 109 ? 70 Strong 9. Montana Beartooth Complex (MT 519, 518, Beartooth MT Sympatric 82 2010 52 ? 60 Strong 517, 514) 10. Madison-Taylor Hilgards Madison MT Sympatric 86 2009 57 ? 60 Strong (MT 325, 326, 327, 362) Spanish 11. Spanish Peaks (MT 324) MT Sympatric 64 2010 71 ? 60 Strong Peaks Total 1179 1348 a Average of the five most recent counts.

LN-LINEAR REGRESSION ANALYSIS

Variability of herd-specific growth rates calculated through ln-linear regression analysis was substantial and similar for both species. Population growth rate estimates for bighorn sheep herds averaged 1.02 and ranged from 0.86 to 1.23 (Figures 2 and 3). The number of bighorn sheep counts that contributed to a ln-linear regression analysis for a herd averaged 21 and ranged from 6 to 44 counts. Bighorn sheep regressions produced a mean p-value of 0.22. Widths of 95% confidence intervals ranged from 0.01 to 0.33 (Tables 3 and 4). The confidence intervals for 11 herds spanned  = 1.00. Figure 4 depicts the relationship between average bighorn sheep herd size and population growth rate.

Population growth estimates for nine mountain goat count units averaged 1.08 and ranged from 0.99 to 1.23 (Figure 5). The number of counts that contributed to a ln-linear regression analysis for a count unit averaged 13 and ranged from 6 to 25. Two mountain goat count units (Madison- Taylor Hilgards, Spanish Peaks) could not be analyzed with ln-linear regression due to limited count data. Mountain goat ln-linear regressions produced a mean p-value of 0.15. Widths of 95% confidence intervals ranged from 0.01 to 0.31 (Table 6). Confidence intervals in four count units spanned  = 1.00.

COMPARISON OF SYMPATRIC AND ALLOPATRIC BIGHORN SHEEP POPULATION GROWTH RATES

Bighorn sheep herds that did not share ranges with mountain goats (allopatric) had finite growth rates that averaged 0.99 and ranged from 0.86 to 1.05 (Table 4). Sympatric bighorn sheep herds had a range of finite population growth rates that averaged 1.05 and ranged from 0.92 to 1.23 (Table 5).

Relatively consistent count data were available regarding allopatric bighorn sheep herds for a 22- year time period from 1988 to 2009. Ln-linear regression of this time series required 220 counts and interpolations, and 21.8% of the values used in the aggregated analysis were interpolated. Aggregated allopatric bighorn sheep herd count data had a growth rate of 1.01 (95% C.I. = 1.00, 1.02; P<0.05) (Figure 6). Since allopatric herd data were available for aggregation across all herds for a slightly longer time period than sympatric herds, the data were also analyzed in two time series (Figure 7) to enable a comparison of lambda values with sympatric herds from similar time series. Allopatric bighorn sheep had growth rates of 1.05 (95% C.I. = 1.01, 1.10; P<0.05) during 1988 to 1994 and 1.02 (95% C.I. = 1.00, 1.03; P<0.05) during 1995 to 2009.

Figure 2 – Summary graph of GYA allopatric bighorn sheep herd ln-linear regressions to illustrate each herd’s population growth trend. Herds are grouped by mountain range.

Figure 3 – Summary graph of GYA sympatric bighorn sheep ln-linear regressions to illustrate each herd’s population growth trend. Herds are listed in order of increasing magnitude of sympatry with mountain goats and are grouped by mountain range.

Table 4 – Summary of database information and ln-linear regression results for allopatric bighorn sheep herd units in the GYA. The confidence intervals that are highlighted in gray encompass  = 1.

Number of Length of Ln- Mountain Start End  lower  upper Herd ID State Counts Used Time Series linear Range Year Year CI (95%) CI (95%) in Analysis (Years)  1. Temple Peak BS610 Wind River WY 14 1994 2008 15 0.86 0.79 0.94 2. Whiskey Mountain BS609 Wind River WY 20 1986 2010 25 0.99 0.98 1.00 3. Yount's Peak BS204 Absaroka WY 15 1989 2009 21 1.01 0.99 1.02 4. Franc's Peak BS205 and Absaroka WY 44 1981 2008 28 1.05 1.02 1.07 Dubois-Badlands BS622 5. Wind River Indian Reservation Absaroka WY 11 1992 2006 15 1.01 0.91 1.12 6. Trout Peak BS202 Absaroka WY 23 1986 2009 24 1.03 0.99 1.06 7. Wapiti Ridge BS203 Absaroka WY 26 1982 2009 28 1.04 1.02 1.06 8. Jackson BS107 Gros Ventre WY 25 1975 2009 35 1.00 0.99 1.01 9. Darby Mountain BS121 Wyoming WY 9 1981 1989 9 1.11 1.03 1.18 (1981-1989) Darby Mountain BS121 Wyoming WY 16 1990 2008 19 0.92 0.85 0.99 (1990-2008) 10. Targhee BS106 Teton WY 20 1976 2010 35 1.00 0.99 1.02

Table 5 – Summary of database information and ln-linear regression results for sympatric bighorn sheep herd units in the GYA. The confidence intervals that are highlighted in gray encompass  = 1.

Number of Length of Ln-  lower Mountain Start End  upper Herd ID State Counts Used Time Series linear CI Range Year Year CI (95%) in Analysis (Years)  (95%) 11. Beattie Gulch, Devil's Slide, and Gallatin MT 20 1989 2011 23 1.01 0.98 1.03 Cinnabar Mountain 12. Mount Everts and Mammoth Absaroka MT 26 1982 2011 30 1.03 1.01 1.05 13. Travertine and Deckard Flats Absaroka MT 17 1988 2011 24 0.92 0.85 0.99 14. Norris Peak and Tower Junction Absaroka WY 15 1995 2011 17 1.14 1.08 1.20 15. Mill Creek Segment Absaroka MT 6 2002 2010 9 0.94 0.87 1.02 16. Yankee Jim Canyon, Corwin, and Absaroka MT 16 1992 2011 20 1.07 1.00 1.14 LaDuke Springs 17. Clark's Fork BS201 Beartooth WY 11 1989 2009 21 1.04 1.00 1.07 18. Black Canyon and Barronette Absaroka MT 5 1995 2000 6 0.78 0.66 0.92 (1995-2000) Black Canyon and Barronette Absaroka MT 8 2002 2011 10 1.13 0.98 1.29 (2002-2011) 19. Monument Peak (1973-1993) Absaroka MT 14 1973 1993 21 0.95 0.89 1.01 Monument Peak (1994-2011) Absaroka MT 16 1994 2011 18 1.06 1.02 1.10 20. Stillwater (1972-1982) Absaroka MT 11 1972 1982 11 1.04 1.01 1.06 Stillwater (1983-1993) Absaroka MT 11 1983 1993 11 0.96 0.93 0.98 Stillwater (1994-2011) Absaroka MT 18 1994 2011 18 1.05 1.04 1.07 21. Hellroaring (1971-1990) Beartooth MT 14 1971 1990 20 0.98 0.94 1.03 Hellroaring (1991-2011) Beartooth MT 14 1991 2011 21 0.99 0.94 1.03 22. West Rosebud Beartooth MT 27 1972 2011 40 1.02 1.00 1.03 23. Tom Miner Basin Gallatin MT 18 1989 2011 23 1.04 1.02 1.05 24. Point of Rocks, Rock Creek, and Gallatin MT 6 1991 1997 7 0.98 0.81 1.18 Lower Big Creek (1991-1997) Point of Rocks, Rock Creek, and Gallatin MT 6 1998 2003 6 0.98 0.87 1.11 Lower Big Creek (1998-2003) Point of Rocks, Rock Creek, and Gallatin MT 6 2005 2011 7 1.08 0.97 1.21 Lower Big Creek (2005-2011) 25. Hilgards (1978-1994) Madison MT 11 1978 1994 17 1.03 0.95 1.11 Hilgards (2003-2010) Madison MT 7 2003 2010 8 1.23 1.07 1.40 26. Spanish Peaks (1980-1996) Madison MT 11 1980 1996 17 1.02 0.97 1.07 Spanish Peaks (1998-2011) Madison MT 12 1998 2011 14 1.10 1.06 1.15 38

Figure 4 – Summary graph of the relationship between GYA bighorn sheep average herd size (a mean of the five most recent herd counts) and the herd’s population growth rate () estimated through ln-linear regression. Bighorn sheep counts that are sympatric with mountain goats are displayed in red and allopatric bighorn sheep counts are labeled in blue.

Figure 5 – Summary graph of GYA mountain goat count unit ln-linear regressions to illustrate each unit’s population growth trend. Herds are listed in order of increasing magnitude of sympatry with bighorn sheep and are grouped by mountain range.

Table 6 – Summary of database information and ln-linear regression results for 9 mountain goat count units in the GYA. The confidence intervals that are highlighted in gray encompass  = 1.

Number of Length of Mountain Sympatric/ Start End Ln-linear  lower  upper Herd ID State Counts Used Time Series Range Allopatric Year Year  CI (95%) CI (95%) in Analysis (Years) 1. Palisades Palisades WY & ID Allopatric 19 1982 2006 25 1.06 1.03 1.10 2. MT_330_MG Absaroka MT Allopatric 15 1989 2011 23 1.05 0.99 1.10 3. MT_323_MG Absaroka MT Sympatric 25 1966 2011 46 1.03 1.02 1.03 4. Beartooth MG201 Beartooth WY Sympatric 15 1986 2006 21 1.02 1.01 1.04 5. MT_329_MG Absaroka MT Sympatric 12 1993 2011 19 1.01 0.98 1.04 6. Yellowstone Absaroka WY & MT Sympatric 9 1998 2010 13 1.23 1.08 1.39 National Park and 7. MT_314_MG GallatinGallatin MT Sympatric 6 1997 2011 15 1.12 1.02 1.24 8. MT_316_MG Beartooth MT Sympatric 6 1997 2011 15 1.07 0.97 1.19 9. Montana Beartooth MT Sympatric 8 1995 2010 16 0.99 0.94 1.04 Beartooth Complex (MT 519, 518, 517, 514)

Consistent count data for sympatric bighorn sheep herds were available for a 17 year time period from 1995 to 2011. Ln-linear regression of this time series required 272 counts and interpolations, and 22.4% of these values were interpolated. Sympatric bighorn sheep herds had an overall growth rate of 1.05 (95% C.I. = 1.03, 1.06; P<0.01) (Figure 8). Aggregated sympatric data were also analyzed in two time series, due to a possible change in growth rate during the time series (Figure 9). Sympatric bighorn sheep had growth rates of 1.01 (95% C.I. = 0.99, 1.03; P<0.05) during 1995 to 2004 and 1.05 (95% C.I. = 1.01, 1.08; P<0.05) during 2004 to 2011.

SYNTHESIS OF GYA MOUNTAIN UNGULATE DATA

To estimate bighorn sheep population growth rates across the GYA, herd count data and interpolations were also aggregated and analyzed from 1995 to 2009 across four regions defined by geographic location (Figure 10). The western portion of the GYA included two herds and 36.7% of the values used in the region’s analysis were interpolated. Bighorn sheep herds in the western GYA had an overall growth rate of 1.03 (95% C.I. = 0.98, 1.08; P<0.20). The northern portion of the GYA included 13 herds and 21.5% of the values in the analysis were interpolated. Bighorn sheep herds in the northern GYA had an overall growth rate of 1.03 (95% C.I. = 1.01, 1.04; P<0.01). The eastern portion of the GYA included six bighorn sheep herds and 22.2% of the values used in the region’s analysis were interpolations. Bighorn sheep herds in the eastern GYA had an overall growth rate of 1.05 (95% C.I. = 1.03, 1.06; P<0.01). Finally, the southern portion of the GYA included five bighorn sheep herds and 20.0% of the values included in the analysis were interpolations. Bighorn sheep herds in the southern portion of the GYA had an overall growth rate of 0.97 (95% C.I. = 0.95, 0.98; P<0.01). The data for each of these regions were also aggregated to analyze the population trend of bighorn sheep across the entire GYA (Figure 11). The overall GYA ln-linear regression analysis produced a growth rate of 1.02 (95% C.I. = 1.01, 1.03; P<0.01). We attempted to complete the same type of analysis for mountain goat populations, but due to limited mountain goat count data, the percent of values (60%) interpolated to estimate a GYA-wide trend for the same time series was not acceptable for calculating a reasonable estimate.

The data from 2009 used for the aggregated population trend analysis were also used to estimate the total number of bighorn sheep and mountain goats in the GYA. Using these data and potential detection probabilities of mountain ungulates in the GYA, a table was created describing potential estimates for both species throughout the area (Table 7). The aggregated count and interpolation data were used to estimate a minimum count of 4998 bighorn sheep and 1349 mountain goats in 2009. Using a moderate estimate of detection probability (0.70) derived from literature on aerial surveys of ungulates in heterogeneous landscapes, the total number of bighorn sheep and mountain goats in the GYA can be approximated as 7140 and 1927, respectively.

Figure 6 – Graph of aggregated GYA allopatric bighorn sheep count data and ln-linear regression results. Aggregated counts for allopatric bighorn sheep herds resulted in  = 1.01.

Figure 7 – Graph of aggregated GYA allopatric bighorn sheep count data and ln-linear regression results divided into two time series. The years included in the second time series with a  = 1.02 were analyzed to provide a comparable time series to the aggregated data available for the sympatric bighorn sheep herds in the GYA.

Figure 8 – Graph of aggregated GYA sympatric bighorn sheep count data and ln-linear regression results. Aggregated counts for sympatric bighorn sheep herds resulted in a  = 1.05.

Figure 9 – Graph of aggregated GYA sympatric bighorn sheep count data and ln-linear regression results divided into two time series. This is an alternative method to derive a population trend for these herds instead of analyzing one time series, as in Figure 8.

Figure 10 – Map of 26 bighorn sheep herds in the GYA, Figure 11 – Summary graph of bighorn sheep regional growth separated into regions that were aggregated for ln-linear rates across the GYA, estimated through synthesis of herd regression analysis. Allopatric bighorn sheep herds are count data and ln-linear regression analysis. symbolized by light blue circles and sympatric herds by dark blue circles.

Table 7 – Estimates of bighorn sheep and mountain goats in the GYA, based on counts from management data and detection probabilities for mountain ungulates in rugged terrain.

Minimum Detection Probability Species Region 2009 Counts & Low Moderate High Interpolations 0.55 0.70 0.90 Bighorn Sheep West GYA 279 507 399 310 North GYA 478 869 683 531 East GYA 3219 5853 4599 3577 South GYA 1022 1858 1460 1136 GYA Total 4998 9087 7140 5553 Mountain Goat West GYA 125 227 179 139 North GYA 793 1442 1133 881 Yellowstone 178 324 254 198 South GYA 253 460 361 281 GYA Total 1349 2453 1927 1499

DISCUSSION

DATA COMPILATION

While the demography database can be helpful in drawing conclusions regarding mountain goat and bighorn sheep population dynamics, it should be noted that some important attributes of the data varied from survey to survey. Since survey data were generously provided by regional biologists, the records reflect the reality and challenges of wildlife management. Therefore, survey intensity varied over time for many of the mountain ungulate herds. In addition, many variables vary from region to region, including the time of the survey, survey method, survey conditions, observer, and the amount of detail recorded. These variables were recorded in the database when available and should be acknowledged, but the data shown and analyzed represent the best efforts by biologists to collect information about these herds. Many historical counts also experienced large changes from year to year, to the extent that variability in population size could not be explained through biological processes. We believe that variability in counts reflects varying detection probability from one survey to the next and the overall difficulty of counting mountain ungulates.

Finally, a major limitation of the data is the definition of the herd unit or herd ID. Except for those bighorn sheep herd units that were aggregated, most herds were defined in the database as named by management biologists historically during data collection. Therefore, bighorn sheep herd ranges and names have been allocated based on predicted geographic limits on herd movements. However, mountain goat count unit ranges and names have been allocated by hunting district and the Yellowstone National Park boundary, and there is potential for

movement of individuals among these specifically outlined areas. We were not able to account for this potential issue with limited herd movement data, and future efforts are necessary to observe herd distribution and migration patterns in more detail.

LN-LINEAR REGRESSION ANALYSIS

Despite variation in herd count data availability in each time series, ln-linear regression analysis provided foundational biological insights for both mountain goats and bighorn sheep in the GYA. Mountain goat populations experienced a positive growth rate across most count units. Bighorn sheep growth rates were more variable and generally remained stable or suggested modest growth.

Population dynamics of mountain goats are generally not well understood and can differ considerably by herd (Hamel et al. 2006). However, the range of finite rate of change values (0.99 to 1.23) estimated for mountain goat count areas in the GYA are reasonable in comparison with lambda estimates for mountain goat herds in other locations. A mountain goat population inhabiting alpine tundra and subalpine forest in the Sheep Mountain-Gladstone Ridge area of Colorado had a finite growth rate that varied from 0.93 to 1.16, and the difference in growth rate was closely linked with differences in harvest quotas (Adams and Bailey 1982). The population trend of mountain goats occupying similar habitats on the front range of the Rocky Mountains in Alberta was calculated using aerial counts in two different models, and lambdas of 1.02 and 1.03 were estimated (Hamel et al. 2004). However, higher growth rates have been recorded, and an introduced mountain goat herd in central Montana was estimated to have a  = 1.30 (Williams 1999). Long term population trends of mountain goats seem to be largely affected by the amount of annual harvest and mountain goat density over an area (Adams and Bailey 1982; Williams 1999).

Bighorn sheep population growth rates are often positively correlated with herd range size, due to the opportunity for farther migration for resources and predator avoidance (Singer et al. 2001). The absence of domestic sheep from a bighorn sheep herd’s range is also important to prevent disease outbreak (Singer et al. 2001). These factors may be more important to a herd’s finite rate of change than the herd’s total size (Caughley 1994; Singer et al. 2001), despite the conservation paradigm that small animal groups will ultimately result in local extirpation (Berger 1990). For example, management biologists hypothesize that the decline of the small, allopatric Temple Peak BS610 herd in the southern GYA with  = 0.86 (95% C.I. 0.79, 0.94 P<0.01) is due to the presence of domestic sheep in the herd’s range. It is also believed that the decline ( = 0.94, 95% C.I. 0.89, 1.01 P<0.15) of the sympatric Monument Peak herd in the northern GYA into the mid-1990s was due to the presence of domestic sheep, even though these bighorn sheep share range with a substantial number of mountain goats. The domestic sheep allotment in this area was retired around 1995, and the bighorn sheep population growth rate from 1994-2010

significantly increased afterward ( = 1.06, 95% C.I. 1.02, 1.10 P<0.01). In regard to the population dynamics of small herds in comparison to large herds of the GYA, Figure 4 shows that several of the smaller, sympatric herds actually have higher estimated growth rates than the larger, allopatric herds. These results were not unexpected because there are many processes that influence population dynamics. Large bighorn sheep populations in the GYA may have lower growth rates due to density-dependent resource limitation. The highly variable growth rates of smaller bighorn sheep populations (<100) somewhat conflict with the idea that small populations are not viable, but these results should be further evaluated due to uncertainty around population growth rate estimates.

The lambda range of 0.86 to 1.23 (mean of 1.02) for bighorn sheep in the GYA is comparable to lambda values found in other areas. Growth rates () of 24 translocated bighorn sheep herds in the western U.S. ranged from 0.85 to 1.26, excluding a population crash by a herd in North Dakota with a  = 0.27 (Singer et al. 2001). The 23 herds included in this range had a mean  of 1.00 (Singer et al. 2001), similar to mean lambda for bighorn sheep herds in the GYA. A bighorn sheep herd in southwestern Alberta of about 200-250 sheep had an average  = 1.05 (Jokinen et al. 2008).

There are limitations to our ln-linear regression results. One assumption of linear regression is that underlying variables remain constant over the time series of the data (Eberhardt 1987). It is unreasonable to expect that all external variables remained constant from year to year during the mountain ungulate surveys. Annual variation in weather may potentially affect population growth rates through influencing reproduction and survival (Butler and Garrott 2012). The proportion of the population counted can also change from year to year due to differences in the area covered on each survey and varying detection probability. Detection probability can change for each survey due to many different external factors, such as wind, snow cover, survey platform, and distribution of animals. Therefore, we expect that the variability of points along the linear regression and consequent wide confidence intervals for many lambda estimates were caused by changing underlying variables. The ln-linear regression is still likely to be a good representation of the population trend, despite variation among individual counts (Festa-Bianchet and Côté 2008). An additional limitation of the analysis is that the 95% confidence intervals for 21 out of 55 total ln-linear regressions encompassed  = 1 (Tables 3, 4 and 5). This means that there is a possibility that the herd’s actual growth rate could be in the opposite direction than the plotted linear regressions shown in Figures 2, 3 and 4. Due to these limitations, more detailed studies regarding mountain ungulate population dynamics need to be pursued in the future.

COMPARISON OF SYMPATRIC AND ALLOPATRIC BIGHORN SHEEP POPULATION GROWTH RATES

We used the historical data to determine if ln-linear regression results supported the hypothesis that sympatry of mountain goats with bighorn sheep adversely affects bighorn sheep herd growth rates. Since the mean of sympatric bighorn sheep herd growth rates ( = 1.05, stdev = 0.08, n = 16) was similar to and higher than that of allopatric herds ( = 0.99, stdev = 0.06, n = 10), individual herd ln-linear regression results did not support this hypothesis. The aggregation of sympatric and allopatric counts to estimate an overall growth rate for each herd type also did not provide conclusive evidence for lower sympatric growth rates. The aggregated growth rate of sympatric herds, 1.05 (95% C.I. = 1.03, 1.06; P<0.01), was higher than the growth rate of allopatric herds during a similar time period,  = 1.02 (95% C.I. = 1.00, 1.03; P<0.05). The difference in growth rates is in the opposite direction than predicted by the competition hypothesis. These results are likely not a result of sympatry, but are due to other factors that influence large herbivore ecology and are confounded by geographic clustering of allopatric and sympatric herds. These factors might include regional differences in weather/climate conditions, geology, and soil fertility. The necessity to incorporate interpolations in growth rate estimations is a limitation to this analysis, but similarity between the mean individual herd growth rates and the aggregated growth rate for each herd type suggests this method produced reasonable results.

We recognize that mountain ungulate populations exist in different ecological settings across the GYA, which can significantly affect growth rates, population sizes, and potential competition between bighorn sheep and mountain goats. Timing of sympatry due to resource availability and migration could potentially change how or if mountain goats affect bighorn sheep herd population dynamics. For example, bighorn sheep herds in the West Rosebud and Hellroaring areas of the northern GYA, are both sympatric with mountain goats on their winter range, but their population dynamics differ (Table 5). West Rosebud bighorn sheep have expanded their winter range and increased in number, which has also coincided with the decline of mountain goat numbers within their range. Hellroaring bighorn sheep are experiencing an increase in mountain goat numbers on their winter range, but their numbers are believed to be low because they have been unable to recover from a population crash due to a severe snowstorm event in 1991. The allopatric Targhee BS106 herd (Table 4) in the southern GYA winters at high elevations in the Teton mountain range, because of loss of low elevation wintering habitat as a result of human development (Whitfield 1983). The shift of the Targhee BS106 herd to non- migratory, high elevation areas may be an important factor to consider as small numbers of mountain goats begin to appear in the Teton mountain range. Examples of bighorn sheep herds that are sympatric with mountain goats on their summer ranges include the herd whose range encompasses Point of Rocks, Rock Creek, and Lower Big Creek, as well as the herd that inhabits Mount Everts and Mammoth in Yellowstone. Both herds are currently increasing in number (Table 5). The Point of Rocks, Rock Creek, and Lower Big Creek population trend is split into

three time series due to a winter die-off in 1997, followed by a period of stability and eventual increase in population size. Therefore, it is unclear how potential competition with mountain goats may affect the population growth rate of these herds and others in the GYA.

SYNTHESIS OF GYA MOUNTAIN UNGULATE DATA

Analysis of bighorn sheep population trends by region and across the GYA also yielded reasonable results. The northern and western portions of the GYA, which are composed of completely sympatric bighorn sheep herds, both appear to have experienced modest growth. The eastern GYA, with the largest number of bighorn sheep in the GYA that are mostly allopatric, also appears to have experienced modest growth across the region. The slight overall decline of the southern GYA appears to be largely driven by the slight decline of a large herd, Whiskey Mountain BS60, but is also caused by the decline of both Temple Peak BS610 and Darby Mountain BS121 herds (Table 4). The overall GYA trend depicted modest growth. These results were not unexpected, as one might predict native bighorn sheep population trends to be relatively stable. This is because bighorn sheep have existed in the region for thousands of years and were successfully restored to the region after the period of over-exploitation in the late 1800s and early 1900s. Regarding non-native mountain goats, limited count data across the GYA resulted in limited ability to analyze historical overall and regional population growth trends. However, most populations within count units with sufficient data for ln-linear regression analysis appear to have increased in number (Figure 5). In contrast to the extensive history of bighorn sheep in the region, mountain goat population growth seems to be influenced by the relatively recent introduction of the species to the region. Mountain goat populations are continuing to expand their distribution in the GYA, with extensive mountainous habitat remaining unoccupied and potentially capable of supporting robust mountain goat populations. Therefore, one might expect regional growth of mountain goat abundance for at least several more decades into the future.

Estimation of bighorn sheep and mountain goat numbers across the GYA was useful in gaining insight on population size and distribution of both species in the region. It is likely that detection probability varies across the GYA, but a series of potential detection probabilities provides a reasonable range of the number of animals potentially in the area. Based on 2009 counts, the mountain goat population size appears to be approximately 27% the size of the overall bighorn sheep population. The 2009 counts also provided insights regarding the heterogeneous distribution of bighorn sheep across the GYA, with the western GYA comprising approximately 6%, the northern GYA comprising 10%, the eastern GYA comprising 64%, and the southern GYA comprising 20% of the total regional population.

Varying habitat and herd attributes across the GYA, such as ecological settings, migration patterns, and distribution, emphasize the complexity of the system and the multitude of factors that may affect population dynamics of mountain ungulates. The overall bighorn sheep 49

population in the GYA appears to be experiencing modest growth, and it is important that their numbers continue to be monitored as the number of mountain goats increases across the region. The growth of non-native mountain goat population sizes and distribution across the GYA should also continue to be studied as we work to understand their population ecology and potential effects on bighorn sheep herds. However, we feel that the ln-linear regression analysis using existing data is an informative preliminary effort toward understanding overall GYA mountain ungulate trends.

ACKNOWLEDGEMENTS

Data and biological insights were generously provided by agency personnel from Idaho Department of Fish and Game, including Hollie Miyasaki, Montana Department of Fish, Wildlife and Parks, including Julie Cunningham, Tom Lemke, Karen Loveless, Justin Paugh, and Shawn Stewart, Wyoming Game and Fish Department, including Doug Brimeyer, Stan Harter, Kevin Hurley, and Doug McWhirter, Yellowstone National Park, including P.J. White, and Grand Teton National Park, including Sarah Dewey. Thanks also to Carson Butler, Jesse DeVoe, Megan O’Reilly, and Mike Sawaya. Financial support for this project was sponsored by the National Park Service, Canon USA Inc., Yellowstone Park Foundation, MSU Undergraduate Scholars Program, MSU Office of the Provost, MSU Office of the Vice President for Research, Wyoming Game and Fish Department, Idaho Department of Fish and Game, and U.S. Forest Service.

LITERATURE CITED

Adams, L.G. and J.A. Bailey. 1982. Population dynamics of mountain goats in the Sawatch Range, Colorado. The Journal of Wildlife Management 46:1003-1009. Bender, L.C., W.L. Meyers, and W.R. Gould. 2003. Comparison of helicopter and ground surveys for North American elk Cervus elaphus and mule deer Odocoileus hemionus population composition. Wildlife Biology 9:199-205. Berger, J. 1990. Persistence of different-sized populations: an empirical assessment of rapid extinctions in bighorn sheep. Conservation Biology 4:91-98. Bodie, W.L., E.O. Garton, E.R. Taylor, and M. McCoy. 1995. A sightability model for bighorn sheep in canyon habitats. Journal of Wildlife Management 59:832-840. British Columbia Ministry of Water, Land and Air Protection. 2004. Accounts and Measures for Managing Identified Wildlife: Northern Interior Forest Region. National Library of Canada Cataloguing in Publication Data, British Columbia. Butler, C.J. and R.A. Garrott. 2012. Climate variation and age ratios in bighorn sheep and mountain goats in the greater Yellowstone area. http://www.gyamountainungulateproject.com/science.html

Caughley, G. 1994. Directions in conservation biology. Journal of Animal Ecology 63:215- 244. Cook, R.D. and J.O. Jacobson. 1979. A design for estimating visibility bias in aerial surveys. Biometrics 35:735-742. Eberhardt, L.L. 1987. Population projections from simple models. Journal of Applied Ecology 24:103-118. Easterbrook R. 2008. Yellowstone National Park Tract and Boundary Data. Department of the Interior (DOI), National Park Service (NPS), Land Resources Division (LRD), Intermountain Land Resources Program Center. Geospatial Dataset-1048199. Festa-Bianchet, M. and S.D. Côté. 2008. Mountain Goats: Ecology, Behavior, and Conservation of an Alpine Ungulate. Island Press, Washington D.C., USA. Gilbert, P.F. and J.R. Grieb. 1957. Comparison of air and ground deer counts in Colorado. The Journal of Wildlife Management 21:33-37. Gonzalez-Voyer, A., M. Festa-Bianchet, and K.G. Smith. 2001. Efficiency of aerial surveys of mountain goats. The Wildlife Society Bulletin 29:140-144. Hamel, S., S.D. Côté, K.G. Smith, and M. Festa-Bianchet. 2006. Population dynamics and harvest potential of mountain goat herds in Alberta. The Journal of Wildlife Management 70:1044-1053. Humbert, J., L.S. Mills, J.S. Horne, and B. Dennis. 2009. A better way to estimate population trends. Oikos 118:1940-1946. Jokinen, M.E., P.F. Jones, and D. Dorge. 2008. Evaluating survival and demography of a bighorn sheep (Ovis canadensis) population. Bienn. Symp. North. Wild Sheep and Goat Counc. 16:138-159. Laundré, J.W. 1990. Resource overlap between mountain goats and bighorn sheep. Great Basin Naturalist 54:114-121. Neal, A.K., G.C. White, R.B. Gill, D.F. Reed, and J.H. Olterman. 1993. Evaluation of mark- resight model assumptions for estimating mountain sheep numbers. Journal of Wildlife Management 57:436-450. Rice, C.G., K.J. Jenkins, and Wan-Ying Chang. 2009. A sightability model for mountain goats. Journal of Wildlife Management 73:468-478. Rubin, E.S., W.M. Boyce, M.C. Jorgensen, S.G. Torres, C.L. Hayes, C.S. O’Brien, and D.A. Jessup. 1998. Distribution and abundance of bighorn sheep in the Peninsular Ranges, California. Wildlife Society Bulletin 26:539-551. Sibly, R.M., and J. Hone. 2002. Population growth rate and its determinants: an overview. Philosophical Transactions: Biological Sciences 357:1153-1170.

Simberloff, D. 2000. Nonindigenous species: a global threat to biodiversity and stability. Pages 325-334 in P. Raven and T. Williams, editors. Nature and human society: the quest for a sustainable world. National Academy Press, Washington, D.C., USA. Singer, F.J., L.C. Zeigenfuss, and L. Spicer. 2001. Role of patch size, disease, and movement in rapid extinction of bighorn sheep. Conservation Biology 15:1347-1354. Swenson, J.E. 1985. Compensatory reproduction in an introduced mountain goat population in the Absaroka Mountains, Montana. Journal of Wildlife Management 49:837-843. Udevitz, M.S., B.S. Shults, L.G. Adams, C. Kleckner. 2006. Evaluation of aerial survey methods for dall’s sheep. Wildlife Society Bulletin 34:732-740. Whitfield, M.B. 1983. Bighorn sheep history, distributions, and habitat relationships in the Teton Mountain Range, Wyoming. M.S Thesis, Idaho State University, Pocatello. 244 pp. Williams, B.K., J.D. Nichols, and M.J. Conroy. 2002. Analysis and Management of Animal Populations. Academic Press, San Diego, California. Williams, J.S. 1999. Compensatory reproduction and dispersal in an introduced mountain goat population in central Montana. Wildlife Society Bulletin 27:1019-1024.

Climatic Variation and Age Ratios in Bighorn Sheep and Mountain Goats in the Greater Yellowstone Area Carson J. Butler and Robert A. Garrott

ABSTRACT

Using management data regularly collected by state and federal agencies, we indexed recruitment rates of bighorn sheep and mountain goats in the Greater Yellowstone Area (GYA) by calculating young:adult ratios. Annual and long term regional climatic conditions were indexed using data from Natural Resource Conservation Service Snotel sensors across the GYA. Linear regression models were used to assess hypotheses that recruitment rates in bighorn sheep and mountain goats in the GYA were associated with annual and regional variation in climatic conditions. The initial dataset consisted of 685 bighorn sheep lamb:ewe ratios from 21 herds since 1960 and 184 mountain goat kid:adult ratios from 18 herds since 1966. After censoring data, 369 bighorn sheep records remained, which were split into three seasonal subsets, and 123 mountain goat records remained in a single dataset. Findings suggest that recruitment rates in bighorn sheep and mountain goats were associated with annual variation in both pre-birth and post-birth climatic conditions, interacting with long term regional climate conditions. Additionally, strong interactions were found between precipitation during the birthing season and winter severity. Collectively, these findings suggest that recruitment in bighorn sheep and mountain goat populations in the GYA may be sensitive to changes in future climate conditions and that the response may vary regionally across the GYA. 53

INTRODUCTION

Bighorn sheep (Ovis canadensis) populations have been drastically reduced since the settlement of western North America. Historically, bighorn sheep were widely distributed across western North America, but by the 19th century bighorn populations were limited to a fraction of their historical range (Montana Bighorn Sheep Conservation Strategy 2010). Bighorn sheep populations have increased in the last 50 years, but many populations are still small and isolated, making conservation of these populations a difficult task facing wildlife managers. Although most western states and provinces allow limited harvest, bighorn sheep are currently classified as a “sensitive species” by the U.S. Forest Service.

Bighorn sheep are a charismatic species appreciated by a wide variety of people and the species is part of a community of North American species that have coevolved for millennia. It is of importance for wildlife managers to conserve this species in its native range so people can experience bighorn sheep on the landscape. More importantly, there is an ethical responsibility to conserve this species as part of the landscape for the simple reason that it exists.

The Greater Yellowstone Area (GYA) is a stronghold for bighorn sheep in North America, as a minimum of 5,000 bighorn sheep inhabit this area of Wyoming and Montana (Montana Bighorn Sheep Conservation Strategy, Wyoming Job Completion Reports). Although the region is a stronghold for bighorn sheep relative to many other areas of the continent, some herds in the GYA have declined in recent decades (Montana Bighorn Sheep Conservation Strategy, Wyoming Game and Fish Department Job Completion Reports). Furthermore, bighorn populations in the GYA face challenges of future climate change, habitat loss, and disease. Climate change is expected to have large impacts on vegetative communities, as well as plant phenology, across the continent and the GYA in the next century, potentially impacting bighorn sheep populations (Bartlein, et al. 1997, Hansen et al. 2001, Romme and Turner 1991). Human encroachment continually reduces habitat available for bighorn sheep, and various human impacts on the landscape may change disease vectors in devastating ways. In face of the multiple challenges facing bighorn sheep conservation, there is a need to assess potential factors that drive sheep population dynamics.

Of further interest, non-native mountain goats introduced in the mid-20th century have expanded their range in the GYA (Lemke 2004) and may impact native sheep populations. In Olympic National Park, introduced mountain goats were found to have a negative impact on plant communities (Lemke 2004). In the GYA, effects of mountain goats on bighorn sheep or vegetation communities are largely unknown, although limited data suggests there is dietary overlap between mountain goats and bighorn sheep (Laundre 1994), that mountain goats may affect bighorn sheep through interference competition, and there is also the possibility that mountain goats will affect bighorn sheep and other ungulate populations through disease transmission (Lemke 2004). A comparison of bighorn sheep population dynamics between 54

allopatric herds and herds sympatric with non-native mountain goats has never been done. The findings of such an analysis may play a role in the management of introduced mountain goats by state and federal natural resource management agencies. For example, the National Park Service is mandated to have clear evidence that an introduced species negatively affects the natural resources, such as native species, of the park prior to initiating any control actions.

It is also of interest to understand drivers of mountain goat population dynamics in the GYA. In the event that negative impacts of introduced mountain goats are discovered, an understanding of mountain goat population dynamics may provide insight to how the species may be expected to further expand in the GYA under various future climatic conditions. Population dynamics of introduced mountain goat herds appear to behave slightly differently than those of native herds (Houston and Stevens 1988, Bailey 1991, Festa-Bianchet et al.1994, Hamel et al. 2006). However, these studies do not show evidence that introduced populations respond fundamentally different to changes in drivers of population dynamics, rather they appear to be differentially sensitive to deviance from “normal” conditions. Thus, studies of population dynamics of introduced mountain goat herds may be beneficial to conservation of native herds by providing a better understanding as to how changing environmental conditions are associated with population dynamics.

The young:adult female ratio, is a metric commonly used in ecological studies as an index of recruitment in ungulate populations (Picton 1984, Douglas and Leslie Jr. 1986, Post and Klein 1999, Garrott et al. 2003, Gilbert and Raedeke 2004, Grøtan et al. 2009a) and has been measured annually in bighorn sheep populations of the GYA for over 20 years by state and federal wildlife management agencies. A similar index of recruitment has also been recorded for mountain goats in the GYA, but due to the difficulty of classifying the sex of mountain goats during aerial surveys, the analogous index for mountain goats is the young:adult ratio.

Several authors have challenged the validity of using age ratios to assess population dynamics stating that changes in the ratio can be driven by adult fecundity, survival of young, or survival of adult female and thus do not elucidate the actual vital rates of a population (Caughley 1974, McCullough 1994). These authors demonstrated that the same young:adult female ratio can be observed regardless of whether a population is actually growing or declining and that a high young:adult female ratio can be the result of high fecundity and young survival (the intuitive interpretation) or the result of poor adult female survival. However, multiple studies of age structured survival in ungulate populations have found that adult survival, in light of density dependent and independent factors, is consistently high (>90%) with very little variation (coefficients of variation 2-15%) and that juvenile survival is always the most variable demographic metric (CV 12-88%) followed by age of first reproduction (Gaillard et al. 1998). Harris et al. (2008) found, using demographic data from elk (Cervus canadensis) in Yellowstone National Park, that young:adult female ratios are primarily explained by survival of young and by age of first reproduction. The model presented by Harris et al. (2008) also found that 55

young:adult female ratios explained most variation in lambda (R2=0.89), indicating utility in predicting overall population trends, if young:adult ratios are accurate. The validity of the young:adult female ratio has also been questioned due to sightability bias of different seasons (Bonenfant et al. 2005), different survey platforms (Bender et al. 2003) and different age and sex classes (Samuel et al. 1992). However, these concerns of bias are more pertinent when attempting to estimate true demographic parameters. An analysis, such as this, that focuses on trends of population dynamics rather than parameters is more robust to bias, so long as any bias is consistent.

Ungulate population dynamics are driven by multiple stochastic and density dependent factors. Studies of ungulate populations have found dynamics to be driven by climatic variables (Douglas and Leslie Jr. 1986, Portier et al. 1998, Garrott et al. 2003, Gilbert and Raedeke 2004, Grøtan 2009a, Grøtan 2009b), disease (Jorgenson et al. 1997, Monello et al. 2001, Cassirer and Sinclair 2007), predation (Festa Bianchet 1994, Garrott et al. 2003, White and Garrott 2005), anthropogenic disturbances (Murphy 1990, Hamel et al. 2006), density dependence (Skogland 1985, Douglas and Leslie Jr. 1986, Portier et al. 1998, Singer et al. 1997, Monello 2001, Festa- Bianchet et al. 2003) and interspecific competition (Picton 1984). Studies have also reported findings that suggest the effects of apparent density independent variables on population dynamics, such as severe weather, are actually contingent on the population density (Portier et al. 1998, Enk et al. 2001). The objective of this analysis is to use long term demographic data of bighorn sheep and mountain goat herds in the GYA to determine meaningful climatic covariates that are associated with changes in recruitment as indexed by young:adult ratios.

STUDY AREA AND METHODS

GYA STUDY AREA DESCRIPTION

The GYA is a mountainous region encompassing over 4,000,000 ha in parts of Montana, Idaho and Wyoming, the majority of which is ecologically intact. Generally speaking, the GYA as defined here extends as far north as Bozeman, Montana, as far south as the southern end of the Wind River Range of Wyoming, west to the Madison Valley of Montana and east to the eastern front of the Absaroka and Wind River Ranges of Wyoming. Elevations range from approximately 1,500 m to over 4,200 m. There are diverse eco-regions ranging from shrub- steppe to temperate forest to alpine communities. Within the region there are two national parks encompassing over 1,000,000 ha and 11 designated wilderness areas encompassing over 1,600,000 ha.

COLLECTING AND CENSORING DATA

Demographic data used in this analysis were provided by Montana Department of Fish, Wildlife and Parks (MFWP), Yellowstone National Park (YNP), Wyoming Department of Game and Fish (WGF), Grand Teton National Park (GTNP), and Idaho Department of Fish and Game (IFG). 56

The demographic data were collected through annual bighorn sheep and mountain goat surveys by these management agencies. Survey methods included ground counts, aerial counts by planes and helicopters, and combinations of these methods. Bighorn sheep surveys were primarily conducted in winter and spring months when bighorns were on winter range, however some surveys were conducted in summer and early fall. Mountain goat surveys were primarily conducted from July to October. The definition of a herd unit for bighorn sheep varied with the regions. For many areas outside of Yellowstone National Park, the bighorns observed in surveys across an entire hunting district are summarized as a single herd. For other areas, however, bighorn herds were defined by specific survey regions and not by the hunting district. Mountain goat populations are defined by hunting district. Two distinct mountain goat populations occupying Yellowstone National Park are surveyed as part of the populations in the adjacent Montana hunting districts. For herds defined by large geographic areas surveys were often conducted over multiple days to obtain a sufficient sample size but summarized as a single survey. Combining multiple surveys results in the possibility of double counting groups of animals in a single survey, however, incidental double counting was deemed acceptable, assuming that no significant changes to population structure occurred between surveys.

To ensure accuracy and consistency, all data underwent rigorous quality assurance/quality control measures. Hardcopy records directly from management agencies were used to ensure that the survey records in our database were accurate. Of particular concern were the number of animals of different age and sex classes counted in a survey as well as the date(s) a survey was conducted. When available, notes for individual surveys were checked for signs that the findings may be biased in any way. For example, biologists sometimes noted that only small proportions of herd ranges were surveyed or that animals were simply difficult to locate and classify. In such scenarios, the data were not included in the dataset. The initial dataset was further refined to fit the needs of the analysis. The criteria for data to be included in the analyses were decided by balancing the benefits of having the highest quality data with the benefits of a sufficient sample size. A basic inspection of the initial dataset and the resulting dataset with varying levels of criteria was used to choose a set of criteria that would best balance the trade-offs between high quality data and sample size. Four criteria were used to select records for use in the analyses.

Greater than 75% of animals observed in a survey must have been adequately classified for a survey to be included in the analysis. If the majority of an animal group was not classified there is the potential that certain age and sex classes were likely to be unclassified, leading to biased ratios (Samuel et al. 1992). For bighorn sheep, adequate classification required indication of age and sex so that the lamb:ewe ratio could be derived. For mountain goats adequate classification only required indication of age because identifying gender is extremely difficult during aerial surveys. Similarly, a minimum number of animals must have been counted in a survey to be included in the analysis. The criterion for inclusion of bighorn sheep surveys was that at least 10 adult females must have been observed and for mountain goats at least 15 adults must have been

observed. The difference in minimum counts required between the species is due to the number of records with low counts, as there were many bighorn sheep records that had 15 or less adult females classified, but relatively few mountain goat records. When small numbers of animals are observed several problems occur. When a small proportion of a population is classified the resulting young:adult ratio is more likely to be biased (Samuel et al. 1992). Also, when small numbers of animals are observed, the resulting age ratios have limited potential values and may be determined largely by chance. For example, if only 5 adult animals are observed, missing a single young reduces the young:adult ratio by 20%. A large proportion of the bighorn sheep records had low numbers of animals observed. To address this issue, data from five adjacent bighorn sub-populations surveyed at the same time each year were combined into two larger herds for analysis. Bighorn sheep herds along the Yellowstone River west of Gardiner, MT originally named “Point of Rocks, Rock Creek, and Lower Big Creek Segments”, “Tom Miner Basin Segment”, and “Upper Yellowstone Complex 2”, were combined into a single herd called “Tom Miner to Beatty Gulch”. Bighorn herds along the Yellowstone River and tributaries east of Gardiner, MT originally named “Black Canyon and Barronette Segments” and “Norris Peak and Tower Junction Segments” were combined into a single herd called “Black Canyon to Soda Butte”.

We also required a level of temporal accuracy of a survey for inclusion in the analysis so that at least the season when a survey was conducted was known. Rocky Mountain bighorn sheep and mountain goats have synchronous birthing, with most births occurring over a short window in May (Geist 1971, Côté and Fiesta-Bianchet 2001). Throughout a given year, the observed age ratio is expected to change as mortality occurs in the absence of births and changes in visibility of different age and sex classes occurs (Bonenfant et al. 2005). Therefore, the results of a classification survey are partially contingent on the date it is performed, making knowledge of an accurate date necessary for interpretation of age ratios. The surveys were temporally distributed throughout the year, so that the majority of data were collected in three time periods for bighorn sheep (December-February, April-May, July-September) and one period (July-October) for mountain goats. Data from each of these periods were analyzed separately because they correspond to different biological periods for bighorn sheep and mountain goats. Surveys conducted outside these time periods were not analyzed in order to tighten the temporal window of each analysis around meaningful biological periods of the year. The resulting dataset from summer bighorn sheep surveys conducted from July to September included only 28 records, which we felt was insufficient to provide meaningful results.

Climate covariates used in the analyses were derived from Natural Resource Conservation Service (NRCS) Snotel sensors (http://www.wcc.nrcs.usda.gov/snow/). Most Snotel sensors used in these analyses began collecting data in the 1981 water year, although one sensor started collecting data in 1967 and several did not begin collecting data until as late as 1991. To provide consistency in the climate covariates, the only source used for these data was from Snotel

Stations. Surveys that were conducted in years when the Snotel sensor assigned to that herd did not collect data could not be used in the analysis. The majority of Snotel sensors did not begin collecting temperature data until several years after SWE and precipitation were first recorded, and as such, fewer records had temperature data available, reducing the sample size of each dataset if temperature was assessed. In order to increase sample size when assessing the other climatic covariates, a subset of each of the seasonal datasets was created for assessment of models including the temperature covariate, while the full dataset was utilized for assessment of models excluding the temperature covariate.

DEVELOPMENT OF A PRIORI CLIMATE COVARIATES

Each bighorn sheep and mountain goat herd was assigned a Snotel sensor for winter and summer season that was thought to most closely estimate the climatic conditions experienced by the herd. However, Snotel sensors could not always be exclusively assigned to a single herd and some Snotel sites were used for multiple herds. Even Snotel sensors that are located very near winter and summer range of bighorn sheep and mountain goats may not accurately portray the onsite climatic condition experienced by the animals themselves because of micro climates that they seek out. Furthermore, directly comparing data from different Snotel sensors in a given year does not suffice as a comparison of conditions experienced by the corresponding herds because sensors do not necessarily provide a representative proximate of their surrounding area. To address this problem, all data from Snotel sensors were centered and standardized to index the departure, in standard deviations, from the mean value since 1981 (30 year average) or since a sensor began collecting data in cases where this was more recent than 1981. The climate covariates then became indices of the extremeness of a season’s weather compared to normal, which was expected to be robust to variation in site-specific weather conditions.

Mountain goats and bighorn sheep occupy similar habitats so the same climate covariates were used for analyses of both lamb:ewe ratios in bighorn sheep and kid:adult ratios in mountain goats. A priori annual climate covariates chosen for analysis were cumulative snow water equivalent (SWE), summer precipitation during July, August, and September (SuP), May 15th – June 15th average temperature (SpT), and May 15th – June 15th precipitation (SpP). SWE and SuP measurements from the biological year an age ratio was obtained as well as the year previous were used as separate covariates because these indices are expected to affect young:adult ratios at different stages in the reproductive cycle. See Figure 1 and Table 1 for additional information regarding the annual climate covariates. Additionally, there were 2 a priori regional climate covariates, AvgSWE and AvgSuP.

Cumulative Snow Water Equivalent (SWE)

Winter severity, relevant to ungulate survival and body condition, can be influenced by multiple parameters such as snow depth, length of time snow cover is present, temperature, as well as

physical characteristics of the snowpack itself. SWE, as referred to in this analysis, is the sum of daily snow water equivalent values from the first day snow accumulates at the Snotel sensor until the snow melts or, in the case of the spring dataset, until the date that a classification survey was conducted. SWE indexes snowfall, snow depth and length of time snow cover is present in a single covariate. Recruitment rates can be affected by winter severity from multiple years. Recruitment of a cohort may be directly impacted by winter severity, indexed by SWE, through mortality of young induced by more difficult locomotion, more difficult access to forage, a longer period when forage is difficult to access, as well as a delay in the onset of the growing season. Recruitment of the same cohort can also be influenced by the severity of the winter experienced by pregnant mothers, indexed by SWEt-1, by reducing maternal condition, either causing pregnant females to abort pregnancies or by reducing nutrition delivered to young before and after birth so that they are less likely to survive their first year (Gaillard et al. 2000). Studies of northern ungulates have varied in their ability to detect correlations of winter severity with recruitment. A study of bighorn sheep lamb survival in southwestern Alberta found no correlation between winter precipitation or temperature and winter lamb survival (Portier et al. 1998), while studies of bighorn sheep in northern Montana (Picton 1984) and elk in Yellowstone National Park (Garrott et al. 2003) detected negative correlations between indices of SWE and young:adult female ratios. Garrott et al. (2003) also found evidence that recruitment of a cohort is correlated to winter severity when the cohort is in utero. Given the findings of these studies, lamb:ewe and kid:adult ratios ratios were hypothesized to be negatively correlated to SWE and

SWEt-1.

Summer Precipitation

Accumulated precipitation during July, August, & September (SuP) can potentially influence recruitment of bighorn sheep through multiple pathways. Total forage production is thought to increase as precipitation during the growing season increases (Lauenroth et al. 1992, Nippert et al. 2006), leading to the expectation that summer forage increases with SuP. Furthermore, summer precipitation increases the length of time that forage remains green, and thus more nutritious. Additionally, summer precipitation may affect the amount of forage present on winter range. In ungulates, summer forage availability during a cohort’s first year is thought to affect survival by influencing maternal condition during lactation and condition of young entering the winter (Verme 1969, Geist 1971, Post and Klein 1999, Cook et al. 2004). Portier et al. (1998) found a positive association between precipitation in the growing season and bighorn sheep lamb survival. Additionally Singer et al. (1997) and Garrott et al. (2003) found evidence of an association between recruitment and growing season precipitation in elk. Forage production the year previous to a cohort’s first year may also affect the cohort by affecting the condition of potential mothers entering the breeding season and winter. The condition of adult females can affect conception rate as well as maternal condition while lambs are in utero, which in turn is associated with lamb survival the following year (Verme 1969, Geist 1971, Gaillard et al. 2000,

Cook et al. 2004). Enk et al. (2001) found evidence that production of bighorn lambs in central Montana was correlated to summer precipitation in the year previous their birth. Other studies have found delayed environmental effects on northern ungulate population dynamics even greater than one year (Saether 1997, Douglas 2001). The multitude of studies supporting the link between summer precipitation, forage production, and herd productivity led to the hypothesis that young:adult ratios are positively correlated to both precipitation during a cohort’s first summer of life (SuP) and precipitation the summer before a cohort is born (SuPt-1).

Neonate Weather

Most pre-weaning deaths occur in the first month of life for newborn ungulates, and can make up a significant proportion of first year mortality (Portier et al 1998, Gaillard et al 2000). Weather conditions during this month may be an important factor in neonate mortality (Geist 1971, Portier et al. 1998). Newborn ungulates are likely susceptible to unfavorable temperatures as their small body size increases the surface: volume ratio (Geist 1971). Portier et al. (1998) found that at high densities, neonatal survival was positively correlated to temperature during the bighorn lambing season, which peaks in late May in the Rocky Mountains (Geist 1971). Mountain goat kidding also peaks during late May (Côté and Festa-Bianchet 2001). A study of mule deer recruitment in the western Cascade Range similarly found a positive association between recruitment and birthing season temperature (Gilbert and Raedeke 2004). May 15th – June 15th average temperature (SpT) is hypothesized to be positively correlated with lamb:ewe ratios and kid:adult ratios, because cold temperatures may increase neonatal mortality. Gilbert and Raedeke (2004) found a negative association between recruitment and birthing season (May 15th-June 15th) precipitation, which we abbreviate as SpP. However, precipitation during this period may also increase forage production and, as stated previously, increased forage production is hypothesized to increase survival of young through the winter. Therefore, the direct effects of May 15th – June 15th precipitation on neonatal mortality were only evaluated a priori using young:adult ratios from summer and fall months, before the benefits of the precipitation on winter survival are expressed. Thus, young:adult ratios derived from summer classification surveys were hypothesized to be negatively correlated to May 15th-June 15th precipitation, and we had no hypotheses regarding the association between winter or spring young:adult ratios and SpP.

Figure 1 – Timeline displaying the temporal relationship between the time periods that the 4 data sets present young: adult ratios from, the climate covariates and the reproductive stages pertinent to the cohort classified as young in surveys.

Regional Climate

Transforming the climate covariates limited our analyses by losing any index of actual climatic conditions across the GYA. However, the actual climatic conditions are also of interest because long-term climatic conditions experienced by different bighorn sheep and mountain goat herds may also affect how vital rates respond to annual variation in climate. For example, herds occupying arid climates may be more sensitive to summer precipitation and less sensitive to winter severity than herds occupying a temperate climate. To explore this hypothesis, bighorn sheep and mountain goat herds were assigned to five climatic regions (Madison Range, Absaroka-Beartooth, Absaroka Range (Wyoming), Wind River Range, and Southern GYA) and data from Snotel stations already used to create the other climate covariates were averaged within regions. Two numeric covariates were created: average annual cumulative SWE measured in inches (AvgSWE) and average July, August, and September precipitation measured in inches (AvgSuP), indexing 30 year averages of these measurements for each of the five regions. We hypothesized that in addition to fundamental differences in recruitment due to regional climate, interactions exist between these indices of long-term climatic conditions and the annual climate covariates such that the response in recruitment of a bighorn sheep or mountain goat population to annual climatic variation is contingent on the regional climate as indexed by AvgSWE and AvgSuP. We expected sensitivity to variation in annual SWE to increase with greater regional average annual SWE and we expected sensitivity to variation in annual summer precipitation to increase as average regional precipitation decreases.

SYMPATRY

In addition to potential associations between climate covariates and young:adult ratios in bighorn sheep and mountain goats, there is particular interest in whether there is association between lamb:ewe ratios in bighorn sheep and sympatry with mountain goats. Conversely, there may be an association between kid:adult ratios in mountain goats and sympatry with bighorn sheep. There is evidence of dietary competition between mountain goats and bighorn sheep (Laundre 1994) and competition may affect recruitment of both species.

Generally, GYA bighorn sheep and mountain goat herds in Montana are sympatric with established mountain goat herds and most bighorn sheep herds in Wyoming are allopatric with mountain goats. Although small numbers of mountain goats may occupy ranges of several bighorn sheep herds classified as allopatric, the herds were still considered allopatric due to the minimal effect a small number of colonizing mountain goats would have on a bighorn sheep herd. Bighorn sheep and mountain goat herds were assigned to three levels of sympatry with mountain goats: allopatric, sympatric during the warm season, or sympatric year-round. The level of sympatry was determined using locations of wintering bighorn sheep and mountain goat observations throughout the GYA. Evidence of competition between bighorn sheep and mountain goats led to the hypothesis that sympatric herds of both species have lower lamb:ewe or kid:adult ratios than allopatric herds and that herds that are sympatric year-round have lower lamb:ewe ratios than herds that are only sympatric during the warm season.

REGION

The GYA was divided into six geographic regions to which bighorn sheep and mountain goat herds were assigned. The six regions were “Madison Range”, “Upper Yellowstone”, “Absaroka- Beartooth”, “Absaroka Range (Wyoming)”, “Wind River Range”, and “Southern GYA”. The geographic regions were primarily defined by mountain ranges. However, for the Upper Yellowstone region the region was defined by the Yellowstone River drainage upstream of Paradise Valley, Montana, and for the Southern GYA region several mountain ranges were combined because in this region each mountain range was only occupied by a single herd, making it impossible to distinguish whether differences in lamb:ewe or kid:adult ratios were associated with differences in mountain ranges or the individual herds themselves. See Appendix A for map of regions.

Table 1 – Descriptions of each of the covariates used in analyses. * Reg and Sym were only tested in exploratory models.

Covariate Abbreviation Description

Standardized cumulative snow water equivalent from the winter experienced by

SWEt-1 pregnant mothers of the cohort classified as “young” during annual surveys

Standardized cumulative snow water equivalent from the winter experienced by the cohort classified as “young” during annual surveys (only applicable to the SWE spring bighorn sheep dataset)

Standardized cumulative precipitation from the summer (July-September)

SuPt-1 previous the birth of the cohort classified as “young” during annual surveys

Standardized cumulative precipitation from the first summer (July-September) SuP experienced by the cohort classified as “young” during annual surveys

Standardized cumulative spring precipitation (May 15th-June 15th) during the SpP birthing season for the cohort classified as “young” during annual surveys

Standardized average spring temperature (May 15th-June 15th) during the SpT birthing season for the cohort classified as “young” during annual surveys

Regional average cumulative snow water equivalent as determined by the AvgSWE Snotel sensors assigned to the herds within each region

Regional average summer precipitation (July-September) as determined by the AvgSuP Snotel sensors assigned to the herds within each region

Reg* Geographic region within the GYA

The level of sympatry experienced between bighorn sheep and mountain goat populations. Sympatry levels were “allopatric”, “warm-season sympatric”, or Sym* “year-round sympatric”

A PRIORI AND EXPLORATORY MODELS

To assess the relationship between lamb:ewe/kid:adult ratios and the climatic covariates, 6 model suites were created. For each of the 3 datasets (bighorn sheep winter, bighorn sheep spring, mountain goat fall) that were analyzed with regression, a model suite with different combinations of covariates, including SpT, average spring (May 15th-June 15th) temperature, was created and 64

evaluated with the subset of data collected during the temporal period that SpT were available. In addition, model suites excluding SpT were also created and evaluated with the full datasets. The covariates AvgSWE and AvgSuP were highly collinear (r=0.85), so only AvgSWE was used in a priori model suites in order to reduce collinearity in the covariates while maintaining a reasonable number of candidate models. Additionally, SpP was only included in a priori models assigned to the mountain goat dataset. An intercept only model was also included in each model suite. The a priori covariates and interactions that were considered for each dataset are shown in Table 2. The mountain goat model suite including SpT consisted of 63 candidate models and the suite excluding temperature consisted of 28 candidate models. The model suite for the winter bighorn sheep dataset including SpT consisted of 59 candidate models and the suite excluding temperature consisted of 28 candidate models. The model suite for the spring bighorn sheep dataset including SpT consisted of 82 candidate models and the suite excluding SpT consisted of

40 candidate models. For each of the datasets, SuPt-1 and SWEt-1 were not included in any a priori models without SuP and SWE, respectively, and interactions involving the time lagged covariates were not considered without inclusion of equivalent interactions involved the non- time lagged covariates (SuP and SWE).

Results from a priori models and coded scatter plots were used to guide exploratory modeling efforts. The top ranked models for each of the datasets were used as base models and combinations of covariates and interactions that were not included a priori for a given dataset, but were supported by a priori findings of other datasets, were added to the top ranked models. Unless SpT was included in top a priori models, the exploratory analyses were conducted on the full datasets. Sympatry (Sym) and the region encompassing a herd’s range (Reg) were added to the top ranked models from each dataset, including interactions with well supported a priori covariates. Variation in lamb:ewe or kid:adult ratios directly attributed to herd specific effects and regional effects was not explored due to uncertainty as to whether differences in lamb:ewe or kid:adult ratios among herds represent real differences in productivity or if the differences are due to different survey conditions in different areas. Reg was included in exploratory models as part of various interactions in order to determine if there were regional differences in the associations between young:adult ratios and the annual climate covariates.

Table 2 – Covariates and interactions that were used in a priori analyses for each of the datasets that were analyzed with regression. See Table 1 for detailed covariate descriptions

Dataset A priori covariates A priori interactions considered

SuPt-1, SWEt-1, SpT, SpP, SuPt-1* SWEt-1, SpT*SpP, Mountain Goat AvgSWE AvgSWE*SWEt-1

SuPt-1, SWEt-1, SpT, SuP, SuPt-1*SWEt-1, SpT*SuP, Winter Bighorn Sheep AvgSWE AvgSWE*SWEt-1

SuPt-1*SWEt-1, SuP*SWE, SWEt- SuPt-1, SWEt-1, SpT, SuP, 1*SuP, AvgSWE*SWEt-1, Spring Bighorn Sheep SWE, AvgSWE AvgSWE*SWE

STATISTICAL ANALYSES

Each of the 6 model suites were evaluated with the appropriate lamb:ewe or kid:adult ratios using the linear model (lm) function in program R. Akaike’s information criteria adjusted for small sample sizes (AICc) was used to determine the most “properly parsimonious” model for each dataset (Bedrick and Tsai 1994). The top a priori models from the reduced datasets (that allowed SpT to be tested) were only assessed if SpT was included in the top model; otherwise, only the top a priori models from the full datasets were further explored. After initial a priori models were evaluated with each dataset, exploratory models were constructed and ranked in the same fashion.

For the winter bighorn sheep dataset, scatter plots suggested there were regional differences in what climatic covariates lamb:ewe ratios were associated with and the nature of the associations. As such, exploratory analysis was conducted on 124 records from the Absaroka-Beartooth (Wyoming portion only), where we felt there was adequate high quality data. Similar to the winter bighorn sheep dataset, findings from the mountain goat dataset also suggested that there are regional differences in how kid:adult ratios respond to the climate covariates. To investigate, we conducted exploratory analysis on a subset of the mountain goat dataset, consisting of 71 records from the Absaroka-Beartooth region. Other regions in the mountain goat dataset, as well as the spring bighorn sheep datasets, did not have adequate records to be analyzed independently.

To investigate the limitations of using the kid:adult ratio as the response variable for the mountain goat dataset (as opposed to the more precise young:adult female ratio), the a priori model suite, including SpT as a covariate, for the spring bighorn sheep dataset was assessed using the lamb:adult ratio (analogous to the kid:adult ratio in mountain goats) as the response variable and compared to the results of the same model suite using the lamb:ewe ratio as the 66

response variable. When using the kid:adult ratio as the response variable, the adult sex ratio is an additional factor to confound the estimate of recruitment, and as such the kid:adult ratio was suspected to reduce the ability of our models to detect relationships between recruitment rates and variability in climatic variables. The analysis of the spring bighorn sheep dataset, where both response variables were available, allowed us to evaluate how results of the mountain goat analysis may have been affected by using a less precise response variable. It must be noted, however, that these species have different social structures and bighorn rams and mountain goat billies may have different propensities to be detected in classification surveys.

RESULTS

RESPONSE VARIABLE DATASET

The initial dataset consisted of 625 records for bighorn sheep and 184 records for mountain goats. After censoring criteria were applied to the dataset, 369 records for bighorn sheep and 123 records for mountain goats remained. Of the 256 bighorn sheep records excluded from analyses, 54 resulted from combining herds. Mountain goat records span from 1982 to 2009 and bighorn sheep records span from 1980 to 2011 (Figure 2). The bighorn sheep dataset was split into three subsets for analysis based on the time of year the data were collected. The summer subset from July through September consisted of 28 records, 1 from MFWP and 27 from WGF. Due to the small sample size, these data were not formally analyzed. The winter dataset from December through February consisted of 273 records with 79 records from MFWP and 194 from WGF. The spring dataset from April and May consisted of 68 records, with 63 records from MFWP and 5 from WGF. The mean lamb:ewe ratio from surveys conducted in the summer was 48.4 ± 16.5 (1 standard deviation) with the variation in ratios of individual herds shown in Figure 3. For winter surveys the mean ratio drops to 33.1±13.5, and for spring surveys the mean ratio drops further to 26.6 ±12.8 (See Figures 4 & 5 for ratios of individual herds in these datasets). Mean lamb:ewe ratios for individual bighorn sheep herds ranged from 21.4 to 52.2, however the herd with highest mean ratio, the Targhee herd in the Southern GYA region, was only represented by 2 records in the dataset. The next highest mean ratio for a herd was 49.6 for the Darby Mountain herd in the Southern GYA region. The herd with the lowest mean ratio, 21.44, was the Temple Peak herd in the Wind River Range. The mean lamb:ewe ratio for all herds combined was 33.3 ± 8.3. See Appendix B for map of herd locations and herd summaries.

All mountain goat records used for analysis were collected in summer or early fall and were included in a single analysis. Of the 123 mountain goat records used in analyses, 3 were collected by IGF, 88 by MFWP, 22 by WGF, and 10 by YNP. The mean kid:adult ratio was 28.5 ± 10.99 and variability of ratios for individual herds are shown in Figure 6. The highest mean ratio, 36.0, was from the Wyoming Palisades herd in the Southern GYA region and the lowest mean ratio, 15.8, was from the mountain goat herd in hunting district 326 (MT_326_MG) in the

Madison Range. The mean kid:adult ratio for mountain goat herds combined was 27.2 ± 6.2. See Appendix C for map of herd locations and herd summaries.

Figure 2 – Temporal distribution of bighorn sheep and mountain goat records included in the analyses.

Figure 3 – Boxplot illustrating variability of lamb:ewe ratios in the summer bighorn dataset. Herds are arranged by geographic proximity, with northern herds to the left and southern herds to the right. Box outlines represent the inter-quartile range, whiskers represent values within 1.5 times the inter-quartile range from the upper or lower quartiles. Open circles represent outliers, which are observations outside the range of the whiskers. Horizontal center lines represent mean values.

Figure 4 – Boxplot illustrating variability of lamb:ewe ratios in the winter bighorn dataset. Herds are arranged by geographic proximity, with northern herds to the left and southern herds to the right. Box outlines represent the inter-quartile range, whiskers represent values within 1.5 times the inter-quartile range from the upper or lower quartiles. Open circles represent outliers, which are observations outside the range of the whiskers. Horizontal center lines represent mean values. 69

Figure 5 – Boxplot illustrating variability of lamb:ewe ratios in the spring bighorn dataset. Herds are arranged by geographic proximity, with northern herds to the left and southern herds to the right. Box outlines represent the inter-quartile range, whiskers represent values within 1.5 times the inter-quartile range from the upper or lower quartiles. Open circles represent outliers, which are observations outside the range of the whiskers. Horizontal center lines represent mean values.

Figure 6 – Boxplot illustrating variability of kid:adult ratios of GYA mountain goat herds. Herds are arranged by geographic proximity, with northern herds to the left and southern herds to the right. Box outlines represent the inter-quartile range, whiskers represent values within 1.5 times the inter-quartile range from the upper or lower quartiles. Open circles represent outliers, which are observations outside the range of the whiskers. Horizontal center lines show mean values. 70

WINTER BIGHORN SHEEP DATASET

The top ranked a priori model from the full winter dataset included only SuPt-1 and scored 10 2 AICc points lower than the intercept only model (Table 3). The adjusted R of the top model, however, was only 0.039 and the model weight was 0.22. The top ranked model predicted a positive relationship between lamb:ewe ratios and SuPt-1 (βSuPt-1=2.78, 95% CI: 1.22, 4.34). Other top models included a nested covariate along with SuPt-1. The estimate of βSuPt-1 was stable among top models.

Table 3 – Top ranked a priori and intercept only models from the full winter bighorn sheep dataset (n=273). Covariates included in competing models were SuPt-1, SWEt-1, SuP, AvgSWE and Sym. Interactions that were considered included SuPt-1* SWEt-1 and AvgSWE*SWEt-1. See Table 1 for detailed covariate descriptions

Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

SuPt-1 2 0.039 0.22 0.22

SuPt-1 + SuP 3 0.8 0.040 0.15 0.37

SuPt-1 + Avg SWE 3 0.8 0.040 0.15 0.52

Intercept only 1 10.0 0.00 0.52

The top model from the exploratory analysis of 124 winter bighorn sheep records from the eastern GYA in Wyoming fit the data better than did the top a priori model of the full winter bighorn sheep dataset. The adjusted R2 was 0.28 and the model weight was 0.34 (Table 4), greatly outperforming the top a priori model from the full dataset which had an adjusted R2 of 0.039. The top ranked model was rather complex compared to the top a priori model, including

SWEt-1, SpP, SuP, and AvgSWE. The only covariate included in the top a priori model from the full dataset, SuPt-1, was not included in any of the top models from the reduced dataset. All covariates in the top exploratory model were included in interactions, making direct interpretation of the main effect coefficients less meaningful. There was a negative interaction between AvgSWE and SWEt-1, although the 95% CI slightly overlapped 0 (βAvgSWE*SWEt-1= - 0.0019, 95% CI: -0.0039, 0.0001). There was also a negative interaction between AvgSWE and

SuP. (βAvgSWE*SuP= -0.0034, 95% CI: -0.0055, -0.0012). The negative interactions between AvgSWE and SWEt-1 and between AvgSWE and SuP predict positive relationships between SWEt-1/SuP and lamb:ewe ratios in areas that experience low AvgSWE and negative relationships in areas that experience high AvgSWE, when other climate covariates are at average (Figure 7). The AvgSWE*SWEt-1 was complicated by a strong negative interaction 71

between SWEt-1 and SpP (βSpP*SWEt-1= -3.69, 95% CI: -5.65, -1.73). When SWEt-1 is average, SpP has very little association with lamb:ewe ratios, in years when SWEt-1 is below average, SpP has a positive association with lamb:ewe ratios, and in years when SWEt-1 is above average SpP has a negative association with lamb:ewe ratios (Figure 8). Conversely, the association between

SWEt-1 and lamb:ewe ratios becomes increasingly negative as SpP increases such that in years with low SpP, there is a predicted positive association between SWEt-1 and lamb:ewe ratios and in years with high SpP the predicted association becomes negative (Figure 8). Once again, these predictions assume other climate covariates are at average. However, the degree that this pattern is expressed, varies with AvgSWE due to the SWEt-1*AvgSWE interaction (Figure 9). The second ranked model simply substituted AvgSWE with AvgSuP (the two are highly collinear) with similar coefficients for the other covariates. No other models scored within 5

AICc points of the top model (Table 4).

Table 4 – Top ranked exploratory models from 124 winter bighorn sheep records from the eastern GYA in Wyoming. Main changes from a priori models were the use of SpP as a covariate, and in various interactions, and AvgSWE was substituted with AvgSuP or Reg. Also, interactions not considered a priori were considered. See Table 1 for detailed covariate descriptions

Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

SWEt-1+SpP + SuP+AvgSWE+SWEt- 1*SpP+ SWEt-1*AvgSWE+SuP*AvgSWE 8 0.277 0.34 0.34

SWEt-1+SpP +SuP+AvgSuP+SWEt-1*SpP+ SWEt-1*AvgSuP+SuP*AvgSuP 8 0.5 0.274 0.26 0.60

SWEt-1+SpP+ AvgSWE + SWEt-1*SpP+ SWEt-1*AvgSWE 6 5.5 0.229 0.02 0.62

Intercept Only 1 32.1 0.00 0.62

Wind River Absaroka Absaroka- Wind River Absaroka Absaroka- Range Range Beartooth Range Range Beartooth

Figure 7 – Predicted relationships between lamb:ewe ratios and SWEt-1 (left graph) and between lamb:ewe ratios and SuP (right graph) in the 3 regions of the Wyoming winter bighorn sheep dataset. AvgSWE values for the Wind River Range, Absaroka Range, and Beartooth Range respectively are 1030”, 2493”, and 3244”. Gray shading symbolizes 95% confidence intervals. See Table 1 for detailed covariate descriptions. **Please note that the y-axes are not equal and that these predictions are based on holding other climate covariates at average**

Figure 8 – Scatter plots illustrating the SpP*SWEt-1 interaction in the Wyoming winter bighorn sheep dataset. The plots show SWEt-1 vs. lamb:ewe ratios at varying levels of SpP and standardized SpP vs. lamb:ewe ratios at varying levels of SWEt-1. Low SpP and SWEt-1 are any values less than -0.5 standard deviations from average. Average SpP and SWEt-1 are values between –0.5 and 0.5 standard deviations from average. High SpP and SWEt-1 are values greater than 0.5 standard deviations from average. Trend lines were fit directly to the data. See Table 1 for detailed covariate descriptions.

Low SpP Avg SpP High SpP Low SWEt-1 Avg SWEt-1 High SWEt-1

ver Range

AvgSWE: 1030” Wind Ri Wind

Low SWEt-1 Avg SWEt-1 High SWEt-1

Low SpP Avg SpP High SpP

”

(Wyoming)

2493

Range

AvgSWE:

Absaroka

Low SpP Avg SpP High SpP Low SWEt-1 Avg SWEt-1 High SWEt-1

”

3244

Beartooth(Wyoming)

AvgSWE: Absaroka

Figure 9 – Prediction plots showing the SpP*SWEt-1 interaction in each region of the dataset. Similarly, for plots on the left depicting the relationship between SWEt-1 and lamb:ewe ratios, the three panels show predicted relationships when SpP is 1 standard deviation below average (Low SpP), at average (Avg SpP), and 1 standard deviation above average (High SpP), respectively. For plots on the right depicting the relationship between SpP and lamb:ewe ratios, the three panels show predicted relationships when SWEt-1 is 1 standard deviation below average (Low SWEt-1), average (Avg SWEt-1), and 1 standard deviation above average (High SWEt-1), respectively. Gray shading symbolizes 95% confidence intervals. See Table 1 for detailed covariate descriptions **Please note that the y-axes are not equal and that these predictions are based on holding other climate covariates at average**

SPRING BIGHORN SHEEP DATASET

The top a priori model from the spring bighorn sheep dataset included SWEt-1, SpT, SuP, SWE as well as a positive interaction between SWE and SuP (Table 5). The adjusted R2 was 0.22 and the model weight was 0.31 (Table 5). There was a strong negative relationship between lamb:ewe ratios and SWEt-1 (βSWEt-1= -5.38, 95% CI: -7.85, -1.78), supporting the hypothesis that as SWEt-1 increases, bighorn sheep recruitment decreases (Figure 10). There was evidence of a negative relationship between lamb:ewe ratios and SpT, although the 95% confidence interval slightly overlaps zero (βSpT= -3.24, 95% CI: -6.47, 0.001). This finding contradicts the hypothesis of a positive effect of spring temperature on bighorn sheep recruitment. The model detected a positive interaction between SuP and SWE (βSuP*SWE=3.302, 95% CI: 0.10, 6.50), such that as SuP values increase, the relationship between lamb:ewe ratios and SWE becomes less negative and when SuP values decrease, the relationship between lamb:ewe ratios and SWE becomes more negative. Like-wise, this interaction predicted a very similar pattern in the association between SuP and lamb:ewe ratios as SWE values vary (Figure 11). Other top models were less complex, including covariates nested in the top model with similar coefficients.

Table 5 – Top ranked a priori and intercept only models from spring bighorn sheep dataset (n=66). Covariates included in competing models were SuPt-1, SWEt-1, SpT, SuP, SWE, AvgSWE and Sym. Interactions that were considered included SuPt-1* SWEt-1, SpT*SuP, SuP*SWE and AvgSWE*SWEt-1. See Table 1 for detailed covariate descriptions.

Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

SWEt-1 + SpT + SuP + SWE + SuP * SWE 6 0.221 0.31 0.31

SWEt-1 + SpT + SuP + SWE 5 1.8 0.182 0.12 0.43

SWEt-1 + SpT + SWE 4 2.3 0.160 0.10 0.53

Intercept only 1 9.6 0.00 0.53

Figure 10 – Scatter plot from spring bighorn sheep dataset, showing negative relationship between SWEt-1 and lamb:ewe ratios. The trend line was fit from the top exploratory model. See Table 1 for detailed covariate descriptions

Low SWE Avg SWE High SWE Low SuP Avg SuP High SuP

Figure 11 – Prediction plots showing the SuP*SWE interaction found a priori in the spring bighorn sheep dataset. The left plot shows the predicted relationship between SuP and lamb:ewe ratios when SWE is 1 standard deviation below average (Low SWE), average (Avg SWE), and 1 standard deviation above average (High SWEt-1), respectively. The right plot shows the predicted relationship between SWE and lamb:ewe ratios when SuP is 1 standard deviation below average (Low SuP), average (Avg SuP), and 1 standard deviation above average (High SuP), respectively. Gray shading represents 95% confidence intervals. See Table 1 for detailed covariate descriptions. **Please note that these predictions are based on holding other climate covariates at average** 77

The top ranked exploratory model for the spring bighorn dataset was similar to the top ranked a priori model, including SWEt-1, SpT, SpP, SWE, Sym and a positive interaction between SWE and SpP. Compared to the top a priori model, the top exploratory model included Sym and substituted SuP with SpP. The corresponding coefficients between the top exploratory and top a priori models were all similar, and coefficients for SpP and the SpP*SWE interaction in the top exploratory model were similar to the coefficients for SuP and the SuP*SWE in the top a priori model. The adjusted R2 for the top ranked exploratory model was 0.39 and the model weight was 0.76 (Table 6), compared to an adjusted R2 of 0.22 for the top a priori model. There was a very strong negative association between SWEt-1 and lamb:ewe ratios (βSWEt-1= -6.17, 95% CI: - 8.89, -3.45). There was also a strong negative relationship between SpT and lamb:ewe ratios

(βSpT= -4.36, 95% CI: -7.35, -1.37). The top ranked model predicted bighorn herds sympatric with mountain goats to have higher mean lamb:ewe ratios than allopatric herds. Lamb:ewe ratios in herds sympatric with mountain goats during summer months were predicted to have 6.87 (95% CI: 1.42, 12.32) more lambs per 100 ewes in a given year than allopatric herds and herds sympatric with mountain goats year round were predicted to have 13.04 (95% CI: 3.53, 22.55) more lambs per 100 ewes than allopatric herds. The coefficient for the SpP*SWE interaction was 4.96 (95% CI: 1.66, 6.25). This interaction predicted that as SpP values increase, the relationship between lamb:ewe ratios and SWE becomes less negative and when SpP values decrease, the relationship between lamb:ewe ratios and SWE becomes more negative. Like- wise, this interaction predicted a very similar pattern in the association between SpP and lamb:ewe ratios as SWE values vary (Figure 12). The second ranked model included an extra covariate and associated interaction, while the third ranked model was the top a priori model. These models carried substantially less model weight than the top ranked model (Table 6).

Table 6 – Top ranked exploratory models from the spring bighorn sheep dataset (n=66). A priori findings were used to guide exploratory models. Covariates included in competing models were SuPt-1, SWEt-1, SpT, SpP, SuP, SWE, AvgSWE and Sym. Additionally, interactions that were not considered a priori were included in exploratory models. See Table 1 for detailed covariate descriptions. Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

SWEt-1 + SpT + SpP + SWE + Sym + SpP * SWE 7 0.394 0.76 0.76

SWEt-1 + SpT + SpP + SuP + SWE + Sym + SuP *SWE + SpP * SWE 9 2.8 0.398 0.18 0.94

SWEt-1 + SpT + SuP + SWE + SuP * SWE+ Sym 7 6.3 0.333 0.03 0.97

Intercept Only 1 23.1 0.00 0.97

Low SWE Avg SWE High SWE Low SpP Avg SpP High SpP

Figure 12 – Prediction plots showing the SpP*SWE interaction found in exploratory analysis of the spring bighorn sheep dataset. The left plot shows the predicted relationship between SpP and lamb:ewe ratios when SWE is 1 standard deviation below average (Low SWE), average (Avg SWE), and 1 standard deviation above average (High SWE), respectively. The right plot shows the predicted relationship between SWE and lamb:ewe ratios when SpP is 1 standard deviation below average (Low SpP), average (AvgSpP), and 1 standard deviation above average (AvgSpP), respectively. Gray shading represents 95% confidence intervals. See Table 1 for detailed covariate descriptions. **Please note that the y-axes are not equal and that these predictions are based on holding other climate covariates at average**

MOUNTAIN GOAT DATASET

A priori models of the mountain goat dataset failed to find any meaningful associations between kid:adult ratios and the climate covariates, as the intercept only model was the top ranked model.

The top exploratory model for the full mountain goat dataset scored 13.5 AICc points lower than the intercept only model and included Reg, SpP, and an interaction between the two. The adjusted R2 value for this model was 0.18. However, this R2 may be misleading in regards to this analysis because much of the variation was explained by regional intercept adjustments, and not by climatic variation. Regional differences in kid:adult ratios are not presented due to the uncertainty that differences found represent true differences in the ratios or if they are the result of differing survey techniques. The SpP*Reg interaction predicts that kid:adult ratios in some regions have a negative relationship with SpP and in other regions have positive relationships, however 95% confidence intervals for the coefficients span zero in all but one of the regions

(βSpP-Absaroka-Beartooth= -2.01, 95% CI: -4.63, 0.61; βSpP-Upper Yellowstone = -11.46, 95% CI: -18.37, - 4.55; βSpP-Madison Range = 0.31, 95% CI: -3.91, 4.52, βSpP-Southern GYE= 3.66, 95% CI: -1.63, 8.96) . All other models performed very poorly compared to the top model (Table 7).

Table 7 – Top ranked models from the full mountain goat dataset (n=123). Covariates included in competing models were SuPt-1, SWEt-1, SpT, SpP, AvgSWE, Sym and Reg. Any interactions considered reasonable were also tested. See Table 1 for detailed covariate descriptions

Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

Reg + SpP + Reg*SpP 4 0.178 0.92 0.92

Reg 2 6.5 0.089 0.04 0.96

Reg + SWEt-1 + Reg*SWEt-1 4 7.0 0.124 0.03 0.99

Intercept only 1 13.5 0.00 0.99

The top ranked model from the exploratory analysis of the Absaroka-Beartooth mountain goat dataset scored 13.5 AICc points lower than the intercept only model and included SWEt-1, SpP, and an interaction between the two (Table 8). The adjusted R2 value was 0.07. The model provided support for a negative relationship between kid:adult ratios and SpP (βSpP= -4.14, 95% CI: -7.48, -0.80) when SWEt-1 is at average. The model provided weaker support for a negative relationship between kid:adult ratios and SWEt-1 (βSWEt-1= -2.06, 95% CI: -4.41, 0.29). The model also provided moderate evidence for a negative SWEt-1*SpP interaction (βSWEt-1*SpP= - 3.47, 95% CI: -7.41, 0.46). This interaction predicted that as SpP values increase, the relationship between kid:adult ratios and SWEt-1 becomes increasingly negative. Like-wise, this interaction predicted a very similar pattern in the association between SpP and kid:adult ratios ratios as SWEt-1 values vary (Figure 13). The second ranked model included SWEt-1 and SpP, but not the interaction and had a ΔAICc of 0.8. The third ranked model included only SWEt-1 and had a ΔAICc of 1.7 (Table 8).

Table 8 – Top ranked models from the exploratory analysis of the Absaroka-Beartooth mountain goat dataset. The top ranked a priori and the intercept only models were the 2nd and 3rd ranked models. Results from a priori models were used to guide exploratory modeling. Covariates included in competing models were SuPt-1, SWEt-1, SpT, SpP, and Sym. Any interactions considered reasonable were also tested. See Table 1 for detailed covariate descriptions.

Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

SWEt-1 + SpP + SWEt-1*SpP 4 0.073 0.34 0.34

SWEt-1 + SpP 3 0.8 0.046 0.23 0.57

SWEt-1 2 1.7 0.018 0.15 0.72

Intercept only 1 1.8 0.14 0.86

Low SWEt-1 Avg SWEt-1 High SWEt-1 Low SpP Avg SpP High SpP

Figure 13 – Prediction plots showing the negative SpP*SWEt-1 interaction found in the exploratory analysis of the mountain goat dataset. The left plot shows the predicted relationship between SpP and kid:adult ratios when SWEt-1 is 1 standard deviation below average (Low SWEt-1), average (Avg SWEt-1), and 1 standard deviation above average (High SWEt-1), respectively. The right plot shows the predicted relationship between SWEt-1 and kid:adult ratios when SpP is 1 standard deviation below average (Low SpP), average (Avg SpP), and 1 standard deviation above average (High SpP), respectively. Gray shading represents 95% confidence intervals. See Table 1 for detailed covariate descriptions. **Please note that the y-axes are not equal** 81

COMPARISON OF MODELS USING LAMB:ADULT RATIO AS RESPONSE VARIABLE TO MODELS USING LAMB:EWE RATIO AS RESPONSE VARIABLE

The top ranked models for the spring bighorn sheep a priori model suite including temperature differed when using the lamb:adult ratio (analogous to kid:adult ratio in mountain goats) as the response variable instead of the lamb:ewe ratio (analogous to kid:nanny ratio in mountain goats) (Table 9). The top ranked model using the lamb:adult ratio as the response variable included 2 SWEt-1 and SWE and the adjusted R value was 0.14. The analogous model using the lamb:ewe ratio as the response variable had an adjusted R2 value of 0.15 and was the 44th ranked model within its model suite. The top ranked model using the lamb:ewe ratio as the response variable 2 included SWEt-1, SpT, SuP, SWE and SuP. The adjusted R value for this model was 0.22; the adjusted R2 value for the analogous model using the lamb:adult ratio as the response was 0.17 and was the 5th ranked model within its model suite. The covariates included in the top ranked model when using the lamb:adult ratio were also included as part of the more complex top ranked model when using the lamb:ewe ratio and the coefficients were similar. The coefficient for SWEt-1 from the top ranked model using the lamb:adult ratio was -3.15 (95% CI: -5.65, - 0.65) and the coefficient from the top ranked model using the lamb:ewe ratio was -4.82 (95% CI:-7.85, -1.78). The coefficient for SWE from the top ranked model using the lamb:adult ratio was -1.91 (95% CI: -4.10, 0.28) and coefficient from the top ranked model using the lamb:ewe ratio was -1.48 (95% CI: (-4.32, 1.35).

Table 9 – Top performing models from a priori model suite for spring bighorn sheep (does not include Sym), using the lamb:adult ratio as the response variable and using the lamb:ewe ratio as the response variable.

Also included is the intercept only model for reference. Covariates included in competing models were SuPt-1,

SWEt-1, SpT, SuP, SWE, AvgSWE. Interactions that were considered included SuPt-1*SWEt-1, SpT*SuP,

SuP*SWE and AvgSWE*SWEt-1. See Table 1 for detailed covariate descriptions

Adjusted Cumulative 2 Model Structure K ΔAICc R Weightc Weightc

Using lamb:adult ratio as response

SWEt-1 +SWE 3 0.135 0.13 0.13

SWEt-1 +SuP +SWE + SuP*SWE 5 1.1 0.155 0.08 0.21

SWEt-1 +SuP +SWE 4 1.1 0.138 0.08 0.29

Intercept only 1 7.2 0.00 0.29

Using lamb:ewe ratio as response

SWEt-1+ SpT +SuP +SWE + SuP*SWE 6 0.223 0.31 0.31

SWEt-1+ SpT +SuP +SWE 5 1.8 0.182 0.12 0.43

SWEt-1 +SpT+SWE, 4 2.2 0.169 0.10 0.53

Intercept only 1 9.5 0.00 0.53

DISCUSSION

WINTER BIGHORN SHEEP ANALYSIS

A priori analysis yielded only weak results, which may be explained by the exclusion of spring precipitation (SpP) as a covariate in addition to the fact that a priori analysis was conducted on bighorn sheep populations across a very large area. The interactions between annual and regional climate covariates in the exploratory analysis conducted on the reduced winter dataset from the eastern GYA provide evidence that the response in bighorn sheep recruitment to annual climate conditions varies across regions. Similar findings have been reported in moose (Grøtan et al. 2009a), and ibex (Grøtan et al. 2009b) as well as in waterfowl (Sǽther et al. 2008). These interactions (AvgSWE/AvgSuP*SuP and AvgSWE/AvgSuP*SWEt-1) provide evidence that the effects of annual climate conditions on bighorn recruitment depend on regional climate conditions and are as we hypothesized. Since AvgSWE and AvgSuP are highly collinear it is impossible to segregate their associations with recruitment patterns and it is more appropriate to consider both covariates as indices of the annual moisture a region receives. The 83

AvgSWE/AvgSuP*SuP interaction suggests that bighorn recruitment in arid regions benefit more from summer precipitation than in regions that receive more precipitation. This prediction is logical, as precipitation is more of a limiting factor in arid regions than in moist regions. This interaction also predicts that summer precipitation in moist areas can be a detriment to recruitment, though no clear mechanism seems obvious. Similarly the AvgSWE/AvgSuP*SWEt- 1 interaction suggests that bighorn recruitment in moist regions is more adversely affected by winter snow accumulation experienced by pregnant females than in arid regions. In the most arid regions increased snow accumulation experienced by pregnant females is predicted to be beneficial to recruitment. Though unintuitive at first thought, it is not unreasonable that in dry regions increased snow accumulation could benefit recruitment by providing additional moisture for forage production during the growing season. This mechanism would necessitate that in arid regions any negative effects of snow accumulation on pregnant females during the winter would be outweighed by the positive effects of the snowpack on forage production that would be available to lactating mothers and juveniles during the growing season. Beneficial effects of snowfall have been reported in studies of other northern ungulates (Mysterud et al. 2001).

The negative interaction between SWEt-1 and SpP is easily seen in the raw data (Figure 5). We are unaware of other studies reporting a similar interaction between annual climate covariates. It is very interesting that the effects of SpP and SWEt-1 on bighorn recruitment depend on each other. Although the actual predicted relationship between SWEt-1 and lamb:ewe ratios vary with regional climate, either can have a significant positive or a significant negative association with recruitment (Figure 7). SWEt-1 can have a negative association with recruitment (as would be typically expected in a northern environment) when that winter is followed by a wet spring and a positive association when followed by a dry spring. As stated previously, snowfall has been reported to have a beneficial effect on ungulates (Mysterud et al. 2001). Similarly SpP can have a positive effect when preceded by a winter with low snowfall or a negative effect when preceded by a winter with high snowfall. Precipitation during the birthing season has been reported to have both positive and negative associations with ungulate recruitment (Portier et al. 1998, Gilbert and Raedeke 2004). The suspected mechanism of this interaction is that the combination of a snowy winter experienced by pregnant ewes, which results in lowered birth weights of the cohort in utero (Gaillard et al. 2000), and a wet birthing season exposes small, less precocious newborns to unfavorable conditions, leading to increased neonatal mortality. It requires both of these conditions to result in significant neonatal mortality. In years when SWEt- 1 is above average, precipitation during the birthing season may also be in the form of snow, further exposing neonates to winter conditions. Alone, snowy winters or wet birthing seasons may benefit recruitment by providing moisture for favorable summer forage conditions. The combination of a mild winter and dry spring is associated with low recruitment, potentially due to the lack of moisture available for forage production. Without the additional stress from wet conditions experienced by neonates in years with above average SpP, long, snowy winters may not affect development of the cohort in utero sufficiently to impact recruitment while the 84

moisture from the snowpack improves forage production. Lambs born following a winter with average or below average snow accumulation may be vigorous enough immediately after their birth to survive a stressful first month of life and then benefit from the increased forage production as a result of increased precipitation.

SPRING BIGHORN SHEEP ANALYSIS

As in the analysis of the winter bighorn sheep dataset, exploratory models outperformed a priori models due to the exclusion of SpP in a priori models. The strong negative association found between SWEt-1 and lamb:ewe ratios is not surprising and is congruent with the findings of the winter bighorn dataset. Ninety-five percent of the spring bighorn data come from the Upper Yellowstone and Madison Range regions, which are relatively wet areas (AvgSWE=3244 and 3099 respectively). Ignoring the potential differences in observed lamb:ewe ratios between winter and spring surveys, the interaction between AvgSWE and SWEt-1 found in the winter bighorn analysis would predict SWEt-1 to have a strong negative effect on bighorn recruitment in these regions. The positive interaction between SWEt-1 and SpP is less intuitive than the interactions in the winter bighorn dataset, however the extremely low p-value (p=0.001) strongly suggests the finding is not by chance. Considering the confidence intervals (Figure 10), this interaction predicts that precipitation during the birthing period has a negative effect on recruitment, but only when the following winter has below average snow accumulation. The interaction also predicts that snow accumulation during the first winter experienced by juveniles negatively effects recruitment, but only when the birthing season of the cohort has below average precipitation. Collectively, the interaction predicts that recruitment of a cohort is optimized by the combination of a dry birthing season followed by a winter with little snow accumulation, which is reasonable considering the relatively moist climate that bighorn herds in the spring dataset experience.

Several findings from analysis of the spring bighorn dataset are unexpected, however. The strong negative association between SpT (May 15th -June 15th temperature) and lamb:ewe ratios

(βSpT= -4.363, p=0.005) found in our top model is opposite our hypothesis and there was no evidence of this association in other datasets. Pettorelli et al. (2007) found that survival of bighorn sheep and ibex juveniles was negatively correlated to the rate of vegetation green-up from May to July, as indexed by NDVI, asserting that rapid green-up across an altitudinal gradient reduces the time that high quality forage is available for grazers that make annual altitudinal migrations. It is possible that the negative association between SpT and recruitment could be explained by the findings of Pettorelli et al. (2007). Lower temperatures from May 15th-June 15th could delay green-up at higher elevations, prolonging the time that high quality forage is available for lambs and lactating mothers. Another unexpected finding was that bighorn herds allopatric with mountain goats in this dataset showed lower lamb:ewe ratios than herds sympatric with mountain goats. This finding is likely due to habitat, management, or demographic characteristics associated with bighorn herds that are sympatric or allopatric with 85

mountain goats. For example, bighorn herds sympatric with mountain goats in this dataset tend to have lower populations than allopatric herds. Also, the dataset from which these results were obtained was relatively small (n=68), increasing the likelihood of obtaining spurious results.

MOUNTAIN GOAT ANALYSIS

Although it was interesting that the top exploratory model in the mountain goat analysis included the same interaction that was found in the winter bighorn sheep dataset, this analysis poorly explained variation in mountain goat recruitment in the GYA. This further demonstrates how little mountain goat demographics in the GYA is understood. The weak findings from the mountain goat dataset compared to the rather strong results from the bighorn analyses also highlights the difficulty in collecting basic demographic data on mountain goat populations compared to other ungulate species. Further research, resulting in a better knowledge of basic mountain goat demographics in the GYA would likely help determine the factors that drive demographic processes in this poorly understood species.

USE OF MANAGEMENT CLASSIFICATION DATA FOR DEMOGRAPHY STUDIES

The results of our bighorn sheep analyses provide evidence that bighorn sheep classification surveys conducted by wildlife management agencies in the GYA do capture trends in recruitment and that these data are capable of providing biological insights. The reduction in the observed lamb:ewe ratios from the summer bighorn dataset (48 lambs: 100 ewes) to the winter bighorn dataset (33.9 lambs: 100 ewes), to the spring bighorn dataset (26.3 lambs: 100 ewes) would be expected to be observed in the true lamb:ewe ratios, as very few, if any, births occur in this time period while juveniles experience greater annual mortality than adults (Gaillard et al. 1998). Additionally, the relatively high R2 values achieved in the winter dataset (Adj. R2: 0.28) and spring dataset (Adj. R2: 0.39), add more support for the usefulness of classification data when studying bighorn population dynamics.

The weak results from the analyses of the mountain goat classification data may be partially explained by the use of the kid:adult ratio rather than the kid:nanny ratio. The results of the analysis of the spring bighorn dataset using the lamb:adult ratio compared to the lamb:ewe ratio suggest that the using the kid:adult ratio to index recruitment in mountain goats, as opposed to the kid:nanny ratio, may reduce the explanatory power of statistical analyses. The use of the lamb:adult ratio (analogous to the kid:adult ratio) as the response variable in the spring bighorn dataset resulted in simple top ranked models with lower explanatory power than when the lamb:ewe ratio was used. Assuming the lamb:ewe ratio is the more accurate index of recruitment and that these findings from bighorn sheep classification data can be applied to mountain goat classification data despite the biological and social differences between the species, these findings suggest that at least some of the weak explanatory power in the top mountain goat models can be attributed to the fact that mountain goats cannot be reliably sexed

during aerial surveys. The poor performance of the mountain goat analyses may also be due to the use of climate covariates that are unimportant to mountain goat recruitment and not due to the quality of the classification data.

CLIMATE CHANGE IMPLICATIONS

The results of the bighorn sheep analyses provide evidence that bighorn demographics in the GYA are influenced by both regional and annual climatic variation. Therefore, it is likely that the changing climate will affect bighorn sheep in the GYA. However, since bighorn recruitment in different climatic regions appears to respond differently to annual climatic variation, we expect that bighorn populations in the different regions of the GYA will respond differently to climate change. With warmer future temperatures predicted worldwide, we expect a decrease in the depth and duration of snow cover in the GYA. The results of our analyses predict that such a decrease in depth and duration of snow cover will lead to increased bighorn recruitment in moist regions that experience high levels of snow accumulation and will result in decreased bighorn recruitment in arid regions that experience low amounts of snow accumulation. Similarly, if future climatic conditions result in increased moisture availability, our results predict that bighorn recruitment in arid regions will benefit and recruitment in moist regions will be adversely affected. Finally, if future climate conditions result in less moisture availability our results predict that bighorn recruitment in arid regions will be adversely affected and that bighorn recruitment in moist regions will benefit. It should be noted, however, that these extrapolations do not consider large-scale ecosystem changes, such as community composition, that would be expected with climate change and may have profound effects on bighorn sheep.

FURTHER RESEARCH

The results of this report have increased our understanding of bighorn recruitment in the GYA by synthesizing previously collected data. However, only the role of climate on recruitment was thoroughly assessed. Although the analyses attempted to detect gross differences in recruitment between bighorn sheep herds that are sympatric with mountain goats and herds that are allopatric with mountain goats, the data available were likely not adequate to detect differences in recruitment between allopatric and sympatric bighorn sheep herds. A simple comparison of recruitment in allopatric vs. sympatric bighorn sheep herds was confounded by varying levels of sympatry within herds through time that were not accounted for, as well as the fact that bighorn herds that are currently sympatric with mountain goats tend to occupy different regions of the GYA than allopatric herds. As mountain goats expand their range in Wyoming, there should be opportunities to examine the effects of colonizing mountain goats on native bighorn sheep populations. Improved knowledge on the interactions between introduced mountain goats and native bighorn sheep will aid wildlife managers in making decisions pertaining to the management of non-native mountain goats in the GYA.

Disease is an important aspect of bighorn demographics (Jorgenson et al. 1997, Monello et al. 2001, Cassirer and Sinclair 2007) that was not addressed by these analyses. Disease epidemics resulting in poor recruitment or all age die-offs have been reported in the Upper Yellowstone, Madison Range, and Absaroka-Beartooth regions within the time the data presented here were collected (Montana Bighorn Sheep Conservation Strategy 2010). Further, bighorn populations in the Wind River Range have suffered poor recruitment since a disease epidemic in 1991 (Wyoming Game & Fish Department 2010) We do not believe that these disease outbreaks reduce the validity of our findings, but rather highlight the need to examine the effects of disease as a driver of bighorn population demographics in the GYA.

SCOPE OF INFERENCE

The data used in these analyses were not collected randomly and were not from a controlled experiment. As such, the results of these analyses alone do not support causal inferences and are only explicitly applicable to the populations from which the data were collected from. Any casual statements made are used to present hypotheses regarding the ecologic mechanisms that drive the correlations observed in these data.

ACKNOWLEDGEMENTS

We would like to acknowledge employees of Idaho Department of Fish and Game (IFG), Montana Department of Fish, Wildlife and Parks (MFWP), Wyoming Game and Fish Department (WGF), Yellowstone National Park (YNP), and Grand Teton National Park (GTNP) for their efforts in collecting and providing to us the data used in these analyses. Specifically, Holly Miyasaki (IGF), Julie Cunningham (MFWP), Tom Lemke (MFWP), Karen Loveless (MFWP), Justin Paugh (MFWP), Shawn Stewart (MFWP), Doug Brimeyer (WGF), Stan Harter (WGF), Doug McWhirter (WGF), P.J. White (YNP), Sarah Dewey (GTNP) and Kevin Hurley (Wild Sheep Foundation) have contributed greatly the Mountain Ungulate Project. Also, we would like to acknowledge Phil Farnes (Snowcap Hydrology) for providing summarized Snotel data. Lastly, thanks to Braden Burkholder, Jesse DeVoe, Elizabeth Flesch, Megan O’Reilly and Jay Rotella of Montana State University for their help. Funding was provided by Yellowstone National Park, Cannon USA Inc-Yellowstone Park Foundation, Montana State University, Wyoming Game and Fish Department, Idaho Fish and Game Department, U.S. Forest Service.

LITERATURE CITED

Bailey, J. A. 1991. Reproductive success in mountain goats. Canadian Journal of Zoology 69:2956- 2961.

Bartlein, P. J., C. Whitlock and S. L. Shafter. 1997. Future climate in the Yellowstone National Park region and its potential impact on vegetation. Conservation Biology 11:782-792.

Bedrick, E.J. and C.L. Tsai. 1994. Model selection for multivariate regression in small samples. Biometrics 50: 226-231.

Bender, L. C., W. L. Meyers and W. R. Gould. 2003. Comparison of helicopter and ground surveys for North American elk Cervus elaphus and mule deer Odocoileus hemionus population composition. Wildlife Biology 9:199-205.

Bolker, B. 2011. Bbmle: tools for general maximum likelihood estimation. R package version 1.0.3. http://CRAN.R-project.org/package=bbmle

Bonenfant, C., J.M. Gaillard, F. Klein and J.L. Hamann. 2005. Can we use the young: female ratio to infer ungulate population dynamics? An empirical test using red deer Cervus elaphus as a model. Journal of Applied Ecology 42:361-370.

Caughley, G. 1974. Interpretation of age ratios. Journal of wildlife management 38:557-562.

Cassirer, E.F., A.F.E Sinclair. 2007. Dynamics of pneumonia in a bighorn sheep metapopulation. Journal of Wildlife Management 71: 1080-1088.

Chappel, R. W.. 1978. Winter bioenergetics of rocky mountain bighorn sheep. Canadian Journal of Zoology 56:2388-2393.

Cook, J. G., B. K. Johnson, R. C. Cook, R. A. Riggs, T. Delcurto, L. D. Bryant and L. L. Irwin. 2004. Effects of summer-autumn nutrition and parturition date on reproduction and survival of elk. Wildlife Monographs 1-61.

Côté, S.D. and M. Festa-Bianchet. 2001. Birthdate, mass and survival of mountain goat kids: effects of maternal characteristics and forage quality. Oecologia 127: 230-238.

Douglas, C. L. and D.M. Leslie, Jr. 1986. Influence of weather and density on lamb survival of desert mountain sheep. Journal of Wildlife Management 50:153-156.

Douglas, C. L. 2001. Weather, disease, and bighorn lamb survival during 23 years in Canyonlands National Park. Wildlife Society Bulletin 29:297-305.

Enk, T. A., H. D. Picton and J. S. Williams. 2001. Factors limiting a bighorn sheep population in Montana following a dieoff. Northwest Science 75:280-291.

Festa-Bianchet, M., K.G. Smith and M Urquhart. 1994. Mountain goat recruitment: kid production and survival to breeding age. Canadian Journal of Zoology 72:22-27.

Festa-Bianchet, M, J.M. Gaillard and S.D. Côté. 2003. Variable age structure and apparent density dependence in survival of adult ungulates. The Journal of Animal Ecology 72:640-649.

Gaillard, J. M., M. Festa-Bianchet and N.G. Yoccoz. 1998. Population dynamics of large herbivores: variable recruitment with constant adult survival. Trends in Ecology & Evolution 13:58-63. 89

Gaillard, J. M., M. Festa-Bianchet, N.G. Yoccoz, A. Loison and C. Toïgo. 2000. Temporal variation in fitness components and population dynamics of large herbivores. Annual Review of Ecology and Systematics 31:367-393.

Garrott, R. A., L.L. Eberhardt, P.J. White and J. Rotella. 2003. Climate-induced variation in vital rates of an unharvested large-herbivore population. Canadian Journal of Zoology 81:33-45.

Geist, V. 1971. Mountain sheep: A study in behavior and evolution. University of Chicago Press, Chicago, Illinois.

Gilbert, B.A., K.J. Raedeke. 2004. Recruitment dynamics of blacktail deer in the western Cascades. Journal of Wildlife Management 68: 120-128.

Grøtan, V, B.E. Saether, M. Lillegård, E.J. Solberg, S. Engen. 2009a. Geographical variation in the influence of density dependence and climate on the recruitment of Norwegian moose. Oecologia 161: 685-695.

Grøtan, V, B.E. Saether, F. Filli, S. Engen. 2009b. Effects of climate on population fluctuations of ibex. Global Change Biology 14: 218-228.

Hamel, S., S. D. Côté, K. G. Smith and M. Festa-Bianchet. 2006. Population dynamics and harvest potential of mountain goat herds in Alberta. Journal of Wildlife Management 70:1044-1053.

Hansen, A. J., R.P Neilson, V.H. Dale, C.H. Flather, L.R. Iverson, D.J. Currie, S. Shafer, R. Cook and P.J. Bartlein. 2001. Global change in forests: responses of species, communities, and biomes. Bioscience 51:765-779.

Harris, N. C., M.J. Kauffman and L.S. Mills. 2008. Inferences about ungulate population dynamics derived from age ratios. Journal of Wildlife Management 72:1143-1151.

Houston, D. B., and V. Stevens. 1988. Resource limitation in mountain goats – a test by experimental cropping. Canadian Journal of Zoology 66:228-238.

Jorgenson, J. T., M. Festa-Bianchet, J.M. Gaillard and W.D. Wishart. 1997. Effects of age, disease and density on survival of bighorn sheep. Ecology 78:1019-1032.

Lauenroth, W. K. and O. E. Sala. 1992. Long-term forage production of North-Armerican shortgrass steppe. Ecological Applications 2:397-403.

Laundre, J. W. 1994. Resource overlap between mountain goats and bighorn sheep. Great Basin Naturalist 54:114-121.

Lemke, T. O. 2004. Origin, expansion, and status of mountain goats in Yellowstone National Park. Wildlife Society Bulletin 32:532-541.

McCullough, D. R. 1994. In My Experience: What Do Herd Composition Counts Tell Us? Wildlife Society bulletin 22:295-300.

Monello, R. J., D.L Murray and E.F. Cassirer. 2001. Ecological correlates of pneumonia epizootics in bighorn sheep herds. Canadian journal of zoology 79:1423-1432.

Montana Bighorn Sheep Conservation Strategy. 2010. Montana Department of Fish, Wildlife and Parks, Wildlife Division. 313pp.

Mysterud, A., N.C. Stenseth, N.G. Yoccoz, R. Langvatn, G Steinheim. 2001. Nonlinear effects of large-scale climate variability on wild and domestic herbivores. Nature 410: 1096-1099.

Nippert, J. B., A. K. Knapp and J. M. Briggs. 2006. Intra-annual rainfall variability and grassland productivity: can the past predict the future? Plant Ecology 184:65-74.

Pettorelli, N.P., F Pelletier, A.V. Hardenberg, M. Festa-Bianchet, S.D. Côté. 2007. Early Onset of Vegetation Growth vs. Rapid Green-Up: Impacts on Juvenile Mountain Ungulates. Ecology 88: 381-390

Picton, H. D. 1984. Climate and the prediction of reproduction of three ungulate species. Journal of Applied Ecology 21:869-879.

Portier, C., M. Festa-Bianchet, J.M. Gaillard, J. T. Jorgenson and N. G. Yoccoz. 1998. Effects of density and weather on survival of bighorn sheep lambs (Ovis canadensis). Journal of Zoology 245:271-278.

Post, E. and D.R. Klein. 1999. Caribou calf production and seasonal range quality during a population decline. Journal of Wildlife Management 63:335-345.

Romme, W. H. and M. G. Turner. 1991. Implications of global climate change for biogeographic patterns in the Greater Yellowstone Ecosystem. Conservation Biology 5:373-386.

Sǽther, B. E. 1997. Environmental stochasticity and population dynamics of large herbivores: A search for mechanisms. Trends in Ecology & Evolution 12:143-149.

Sǽther, B.E., M. Lillegård, V. Grøtan, M.C. Drever, S. Engen, T.D. Nudds, K.M. Podruzny. 2008. Geographical gradients in the population dynamics of North American prairie ducks. Journal of Animal Ecology 77: 869-882.

Samuel, M. D., R.K. Steinhorst, E.O. Garton and J.W. Unsworth. 1992. Estimation of wildlife population ratios incorporating survey design and visibility bias. Journal of Wildlife Management 56:718-725.

Singer, F. J., A.Harting, K.K. Symonds and M.B. Coughenour 1997. Density dependence, compensation, and environmental effects on elk calf mortality in Yellowstone National Park. Journal of Wildlife Management 61:12-25. 91

Skogland, T. 1985. The effects of density-dependent resource limitations on the demography of wild reindeer. Journal of Animal Ecology 54:359-374.

Verme, L. J. 1969. Reproductive patterns of white-tailed deer related to nutritional plane. Journal of Wildlife Management 33:881-887.

Wickham, H. 2009. Ggplot2: elegant graphics for data analysis. R package version 0.9.0. http://had.co.nz/ggplot2/book.

Wyoming Game and Fish Department Job Completion Reports. 2010. Wyoming Game and Fish Department.

Appendix A. Map of regions defined in the analyses

Appendix B. Map and summary information of the bighorn sheep herds in the database

Number Range of Records Records Data Mean Lamb: Herd Summer Winter Snotel Sympatric/ Management in Initial Meeting Meeting 100 Ewe Ratio Number Herd ID Region Snotel Site Site Allopatric Agency Dataset Criteria Criteria ± 1 S.D. Sympatric Madison Warm 1993- 1 Spanish Peaks Range Lone Peak Lone Peak Season MFWP 25 10 2011 32.39 ± 12.49 Madison Sympatric 1988- 2 Hilgards Range Carrott Basin Beaver Creek Year Round MFWP 17 10 2009 32.95 ± 19.88 Sympatric Tom Miner to Upper Warm 1989- 3 Beatty Gulch Yellowstone Shower Falls Shower Falls Season YNP 23 22 2009 25.73 ± 12.73 Upper Yellowstone Upper Northeast Northeast 1988- 4 Complex 1 Yellowstone Entrance Entrance Allopatric MFWP,YNP 24 22 2009 22.93 ± 15.39 Absaroka- Monument Monument Sympatric 1982- 5 Monument Peak Beartooth Peak Peak Year Round MFWP 30 12 2010 31.99 ± 17.66 Sympatric Black Canyon to Upper Monument Northeast Warm 1995- 6 Soda Butte Yellowstone Peak Entrance Season MFWP, YNP 13 11 2009 30.55 ± 14.17 Sympatric Absaroka- Warm 1982- 7 Stillwater Beartooth Placer Basin Placer Basin Season MFWP 38 28 2009 35.25 ± 15.14 Absaroka- Sympatric 1980- 8 West Rosebud Beartooth Fisher Creek Placer Basin Year Round FMWP 28 20 2009 27.17 ± 15.77 Absaroka- Sympatric 1982- 9 Hellroaring Beartooth Beartooth Lake Beartooth Lake Year Round MFWP 26 11 2004 29.89 ± 13.65 Sympatric 10 Absaroka- Wolverine Warm 1982- Clark's Fork Beartooth Beartooth Lake Lake Season WGF 29 25 2009 41.94 ± 11.14 Absaroka Range 1984- 11 Trout Peak (Wyoming) Evening Star Evening Star Allopatric WGF 26 23 2009 35.46 ± 7.12 Absoraka Range 1982- 12 Wapiti Ridge (Wyoming) Blackwater Blackwater Allopatric WGF 31 26 2007 33.18 ± 14.18 Absaroka Range 1982- 13 Yount's Peak (Wyoming) Yount’s Peak Yount’s Peak Allopatric WGF 27 25 2007 36.51 ± 14.70 Absaroka Range 1983- 14 Franc's Peak (Wyoming) Kirwin Kirwin Allopatric WGF 36 23 2010 28.86 ± 9.81 95

Absaroka Range Burroughs Burroughs 1981- 15 Dubois Badlands (Wyoming) Creek Creek Allopatric WGF 27 25 2008 29.78 ± 9.04 Wind River 1989- 16 Whiskey Mountain Range Cold Springs Cold Springs Allopatric WGF 24 21 2010 23.54 ± 7.85 Wind River Townsend Townsend 1995- 17 Temple Peak Range Creek Creek Allopatric WGF 14 6 2002 21.44 ± 17.47 Southern Gros Ventre Gros Ventre 1982- 18 Jackson GYE Summit Summit Allopatric WGF 40 28 2009 43.99 ± 13.04 Southern 1991- 19 Targhee GYE Phillips Bench Phillips Bench Allopatric WGF, GTNP 27 2 1998 52.19 ± 7.51 Southern Darby Darby 1981- 20 Darby Mountain GYE Mountain Mountain Allopatric WGF 44 32 2007 49.64 ± 17.60

Appendix C. Map and summary information of the mountain goat herds in the database

Total Number Range of Mean Kid: Records Records Data 100 Adult Herd Summer Winter Sympatric/ Management in Initial Meeting Meeting Ratio ± 1 Number Herd ID Region Snotel Site Snotel Site Allopatric Agency Dataset Criteria Criteria S.D. Sympatric 1 Madison Warm MT_324_MG Range Lone Peak Lone Peak Season MFWP 1 1 1999 14.28 2 Madison 1994- 27.06 ± MT_325_MG Range Lone Peak Lone Peak Allopatric MFWP 6 6 2009 3.32

3 Madison 1994- 15.83 ± MT_326_MG Range Lone Peak Lone Peak Allopatric MFWP 6 6 2009 5.91 Sympatric 4 Madison Carrot Carrot Warm 1994- 19.88 ± MT_327_MG Range Basin Basin Season MFWP 4 4 2003 5.91

5 Madison Carrot Carrot Sympatric 1994- 24.16 ± MT_328_MG Range Basin Basin Year Round MFWP 3 3 1999 6.96 Sympatric 6 Madison Carrot Carrot Warm MT_362_MG Range Basin Basin Season MFWP 1 1 1993 17.31 Sympatric 7 Upper Shower Shower Warm 1997- 29.58 ± MT_314_MG Yellowstone Falls Falls Season MFWP/YNP 9 8 2009 14.70 8 Absaroka- Monument Monument 1990- 30.35 ± MT_330_MG Beartooth Peak Peak Allopatric MFWP 14 11 2007 13.89 Sympatric 9 Absaroka- Monument Monument Warm 1981- 33.10 ± MT_323_MG Beartooth Peak Peak Season MFWP 25 18 2007 10.85 Sympatric 10 Absaroka- Monument Monument Warm 1989- 32.46 ± MT_329_MG Beartooth Peak Peak Season MFWP 14 14 2007 9.69 Sympatric 11 Absaroka- Fisher Fisher Warm 1997- 29.53 ± MT_316_MG Beartooth Creek Creek Season MFWP/YNP 11 11 2009 6.09 Sympatric 12 Absaroka- Fisher Fisher Warm 1979- 21.32 ± MT_519_MG Beartooth Creek Creek Season MFWP 7 6 1986 7.88

Sympatric 13 Absaroka- Fisher Fisher Warm 1979- 20.72 ± MT_518_MG Beartooth Creek Creek Season MFWP 12 8 1990 8.67 Sympatric 14 Absaroka- Beartooth Beartooth Warm 1981- 19.96 ± MT_517_MG Beartooth Lake Lake Season MFWP 11 6 1990 9.35

15 Absaroka- Beartooth Beartooth Sympatric 1981- 33.99 ± MT_514_MG Beartooth Lake Lake Year Round MFWP 6 4 1988 17.87 31.53 ± 16 Beartooth Beartooth Beartooth Beartooth Sympatric 1989- 7.22 MG201 Range Lake Lake Year Round WGF 23 14 2006 17 Palisades Southern Pine Creek Pine Creek 1997- 35.98 ± (Wyoming) GYE Pass Pass Allopatric WGF 12 8 2006 9.77 18 Palisades Southern Pine Creek Pine Creek 2002- 23.00 ± (Idaho) GYE Pass Pass Allopatric IFG 18 3 2006 5.67

Development and Testing of Occupancy Surveys for Sympatric Bighorn Sheep and Mountain Goats in Northern Yellowstone Megan O’Reilly, Jay Rotella, and Robert A. Garrott

ABSTRACT

Both bighorn sheep and mountain goats are generalist herbivores that overlap extensively in broad food and habitat requirements, but there have been few studies examining the potential for competition between sympatric populations. One area in which native bighorn sheep are living in sympatry with non-native mountain goats is the southern Gallatin Mountain range within and adjacent to the northwest boundary of Yellowstone National Park. Existing data of bighorn sheep and mountain goat observations for the area vary in spatial precision and records of areas where observers looked for but did not detect animals are not available. To gain a better understanding of the relationship between bighorn sheep and mountain goats and their habitat, it is necessary to understand resource selection and the extent of overlap in resource use among sympatric populations on fine spatial and temporal scales. In order to meet this need we designed and implemented formal, ground-based occupancy surveys during the summer of 2011. A crew of four spent 113 observer days in the field resulting in 240 hours of occupancy surveys and approximately 210 miles hiked. Observers recorded presence-absence data for both mountain ungulates. A total of 6,932 sample units were surveyed, with 68 bighorn sheep and 95 mountain goat groups detected. Proportions of bighorn sheep and mountain goat groups detected by both observers during double observer surveys were 76.9% and 54.5%, respectively. We summarize the objectives and field design of this project and report on our efforts to develop enhanced habitat models which will provide managers with additional ecological insights. 100

INTRODUCTION

An organism’s ecological niche is defined by its ecological role and the abiotic and biotic factors within its habitat that limit where it can survive, develop, and reproduce (Elton 1927, Hutchinson 1957). In the absence of competitors, an organism can occupy a larger ecological niche than when the use of some primary resource is restricted by individuals of another species. When two species have overlapping niches, interspecific competition will occur resulting in one species using limited resources more efficiently and reproducing faster. In some instances interspecific competition may lead to local elimination of the inferior competitor, which is known as competitive exclusion (Gause 1934, Begon et al. 2006). The extent of interspecific competition is reliant on the extent of niche overlap and is expected to lead to evolution towards niche divergence (Lack 1947, Schluter et al. 1985) allowing both species to coexist.

In some situations interspecific competition can be caused by intentional or accidental introduction of non-native species by human activities (Cox 1999). When occupying a niche similar to that of native species, non-native species may compete for resources indirectly through resource exploitation, or directly through interference by individuals (Pianka 1981, Noss and Cooperrider 1994, Poling and Hayslette 2006). Native species competing with exotics may be forced to modify their use of resources, resulting in decreased individual fitness and less abundant populations (Gause 1934, Hobbs et al. 1990, Mack et al. 2000). Non-native species may also serve as vectors for disease transmission (Reed and Green 1994, Daszak et al. 2000) which may further limit the fitness of native species. In some cases, competition with non-native species may lead to competitive exclusion of native species. When assessing potential competition between native and non-native species, it is important to consider other potential determinants of the distribution and abundance of a native species such as human exploitation, predator effects, climate change and disease outbreaks (Fritts and Rodda 1998).

In some parts of North America mountain goats (Oreamnos americanus) are not native causing concern about their potential impact on native plant and animal populations, particularly Rocky Mountain bighorn sheep (Ovis canadensis canadensis) (Laundre 1994, Varley and Varley 1996). Mountain goats are native to North America with historic ranges throughout British Columbia, Alberta, the Yukon Territories, southern Alaska and the northwestern United States, west of the Continental Divide (Chadwick 1983). Native populations in the United States are restricted to Alaska, northern Idaho, western Montana and western Washington (Guenzel 1980, Picton and Lonner 2008). Mountain goat habitat is frequently characterized by remote, mountainous, steep and cliffy terrain. In an effort to increase hunting opportunities in the early and mid-1900s, state wildlife agencies in the northwestern United States transplanted mountain goats to numerous mountain ranges outside their historical range (Picton and Lonner 2008). Introduced populations of mountain goats have expanded their range substantially making it critical to examine the potential effects of mountain goats on bighorn sheep.

101

Native populations of Rocky Mountain bighorn sheep are distributed throughout the mountainous regions of western North America from Alberta and British Columbia south as far as central Arizona and northern New Mexico. Bighorn sheep prefer areas of open landscape with stable plant communities, dominated by grasses and sedges in close proximity to steep, mountainous terrain. Both mountain goats and bighorn sheep are generalist herbivores that overlap extensively in broad food and habitat requirements (Laundre 1994). Although bighorn sheep and mountain goats are adapted to various habitats, it appears they most frequently overlap in subalpine and alpine areas. Similarities in foraging and habitat use could lead to competition between the two species (Pianka 1981, Noss and Cooperrider 1994 , Mack et al. 2000, Poling and Hayslette 2006).

To date, there have been few studies examining the potential for competition between mountain goats and bighorn sheep. The majority of studies on these two species have examined feeding habits and habitat use of only allopatric populations and attempted to demonstrate the potential for competition based on broad use of similar resources. Resulting literature indicates possible competition between the two species as a result of dietary overlap in some seasons as well as dominance of mountain goats over bighorn sheep when direct interactions occur (Chadwick 1983, Reed 1986).

Both mountain goats and bighorn sheep have some distinctive physical adaptations which should allow them to successfully exploit different niches within alpine and subalpine habitats. Mountain goats have short, heavily muscled limbs and broad hooves with special traction pads that help grip on smooth surfaces, such as rock and ice. They are not well suited to outrun predators (Geist 1971, Adams et al. 1982, Chadwick 1983) and in many cases avoid predation by escaping to steeper, more rugged terrain. This requires mountain goats to forage alone or in small groups, where food resources are frequently patchy and/or sparse in cliff terrain (Adams et al. 1982). Bighorn sheep are morphologically well adapted to outrun predators over broken terrain (Geist 1971:257) with longer limbs and leaner bodies. As a result, they tend to feed in larger groups in areas of continuous, dense forage with unobstructed visibility and close proximity to open areas where they can outrun predators (Shannon et al. 1975, Adams et al. 1982). Bighorn sheep and mountain goats are naturally sympatric in some areas west of the Continental Divide and both species, in these areas seem to have partitioned their use of resources in such a manner that they are able to coexist. In other areas, bighorn sheep have persisted in the absence of competition from native mountain goat populations. The possibility exists that in these areas bighorn sheep have expanded their niche to encompass some of the resources that would typically be used by mountain goats in their native habitat. In these instances the potential for competition between the two species may be increased as a result of greater overlap in resource use. To gain a better understanding of the relationship between bighorn sheep and mountain goats and their habitat, it is necessary to understand resource selection and the extent of overlap in resource use among both allopatric and sympatric populations on a fine spatial scale.

102

Several types of data may be used to examine resource selection by animals; however, each has strengths and limitations. Most commonly, presence-only data are collected which consist of animal locations from radio-telemetry collars, transect surveys or opportunistic sightings and do not include records for areas where observers looked for but did not detect animals. These data may be used to produce coarse spatial representations of animal distributions (Elith et al. 2006, Gormley et al. 2011) but do not allow for predictions or comparisons between places where the species of interest is present and absent (Hirzel et al. 2006). In these situations, where non- detection points do not exist, random “available” points may be generated for the study area and compared to points of use; however, this method is less likely than formal survey methods to identify sites of true absence (Loiselle et al. 2003, Keating and Cherry 2004). If available points are used instead of non-detection points, the intercept cannot be interpreted, and only a resource selection function (RSF) can be estimated. A RSF provides only relative probabilities that are proportional to the actual probabilities of use for an area. When true probabilities of use across a study area are very small, relative probabilities may be misleading. One area may appear to be strongly selected for relative to another, when in fact both areas have very small actual probability of use (Manly et al. 2002, Lele and Kleim 2006). Presence-only data are frequently not collected as part of systematic, standardized surveys and, as a result, are likely subject to spatial and detection biases (Hijmans et al. 2000, Reese et al. 2005, and Gormley et al. 2011).

A second type of data used to assess resource selection is presence-absence data. These data are most often collected as part of formal occupancy surveys, during which observers record areas of both detection and non-detection of animals, within defined survey units. These data may be used to estimate resource selection probability functions (RSPF) which allow for estimation of the actual probability that a species of interest will use a given resource, based on some combination of ecological variables, such as food or escape terrain.

The most effective type of data for examining resource selection is occupancy data which consists of presence-absence data combined with estimates of the probability that an animal will be detected when present in a survey unit. Animals will not always be detected by observers when present, so we cannot interpret non-detection of an animal as a true absence. Failure to account for imperfect detection will tell us more about the observer’s ability to find animals on the landscape rather than true occupancy (MacKenzie 2005) and can result in the underestimation of occupancy and false inferences about the relationship between actual occupancy and habitat characteristics. There is typically variation in the probability among survey units that an animal will be detected when present. By visiting survey units repeatedly over a short period of time and conducting multiple surveys it is possible to use patterns of detection and non-detection to estimate detection probabilities (MacKenzie et al. 2002).

Our study area represents a portion of the southern Gallatin Mountain range within and adjacent to the northwest boundary of Yellowstone National Park (Figure 1) which provides year-round habitat for native bighorn sheep and non-native mountain goats. Based on historical observation 103

data, we know that bighorn sheep and mountain goats have been sympatric in the area since 1967; however existing animal location data have been collected in many different ways resulting in observations that vary greatly in spatial precision. Most of these observations were recorded by managers during annual population surveys and are comprised of presence-only data.

The objectives of this study are to: 1) use existing management presence-only data to develop habitat models to define regions within the general study area to be intensively sampled during formal occupancy surveys; 2) intensively sample individual regions within the study area in an effort to develop, test and refine ground-based field methodology for collection of spatially explicit occupancy data for bighorn sheep and mountain goats in mountainous terrain; and 3) develop preliminary habitat selection models to predict distributions of bighorn sheep and mountain goats by including additional habitat covariates, multiple scales of selection, and spatially explicit occupancy and detection data obtained during the 2011 field season.

METHODS

STUDY AREA

The general study area (1342 km2) used for the presence-only habitat selection modeling effort was located in the southern portion of the Gallatin Range in Montana and Wyoming and encompassed areas east of the Yellowstone River (Figure 1). The area was chosen based on historic data which indicated local sympatry of native bighorn sheep and introduced mountain goats. The area was further broken down into four survey regions (Region 1, Region 2, Region 3 and Region 4) to be used for the development of field methods and data collection for formal occupancy surveys. These four survey regions extended from Fortress Mountain south to Sepulcher Peak and were restricted to areas west of the Yellowstone River (Figure 1). Regions were defined based on the results of presence-only habitat suitability models and maps, as well as the ability of an occupancy survey field crew to cover a given area with the resources available during three to four day backpacking trips.

104

Figure 1 – Study area located in the southern Gallatin Range in Montana and Wyoming. The area is composed of four individual survey regions, each of which was surveyed for presence and absence of bighorn sheep and mountain goats during the summer of 2011.

Land ownership was a mosaic of U.S. Forest Service (Gallatin National Forest), National Park Service (Yellowstone National Park), Bureau of Land Management and private lands. Topography varied from rolling hills and flats with winding streams to abrupt, steep slopes. Elevations in the area ranged from 1,501 meters at the Yellowstone River to 3,334 meters on Electric Peak. The area experiences short summers and harsh long winters, snow frequently persisting in the higher portions of the study area into July. Average annual precipitation is 118.6 centimeters as measured at 2469 meters elevation by the Shower Falls weather station in the northern Gallatin Range, Montana. Mean annual temperature is 1.1 degrees Celsius.

PRESENCE-ONLY MODELING EFFORT

Management Point Data

In order to determine the area to be sampled via formal occupancy surveys in the summer of 2011, summer season habitat selection modeling was conducted using a subset of existing bighorn sheep and mountain goat presence-only observation data which were previously compiled into a point database of all available bighorn sheep and mountain goat locations for the Greater Yellowstone Area (GYA). Data were collected by multiple agencies including Montana Fish, Wildlife and Parks, Yellowstone National Park, Montana State University, and private

105

wildlife consultants beginning in 1967 and continuing to the present. Animals were observed from various survey platforms including airplanes, helicopters and ground locations. Many of these data were not collected as part of structured surveys, resulting in various spatial resolutions of animal locations. Records considered for these analyses were restricted to observations in our general study area and with quarter section or finer spatial resolution, resulting in a total of 160 mountain goat locations and 377 bighorn sheep locations (Figure 2a). There were insufficient mountain goat data to examine habitat use in winter or lambing seasons, so we opted to focus analyses on summer and rut location data for mountain goats and summer location data for bighorn sheep. Bighorn sheep appeared to move into lower elevation habitat during the rut, however mountain goats appeared to inhabit similar areas during both summer and rut seasons (Figure 2a). As a result, mountain goat location data from these two seasons were combined and included for development of summer habitat models. A total of 1000 available points were randomly distributed (Figure 2b) within the study area to define characteristics of available habitat and to compare available to used habitat locations. This number was chosen as it provided reasonable coverage of the study area and was anticipated to capture heterogeneity in the landscape. The same set of available points was employed in modeling for both species. Around each point, a 300-m radius buffer was created for data extraction from the related covariate layers to quantify habitat characteristics. The buffer size was selected because this is the best spatial resolution we felt was reasonable to assign to these data.

106

Figure 2 – (a) Bighorn sheep summer and mountain goat summer and rut locations used in summer season, presence-only habitat models in the general study area. (b) Randomly distributed “available” points used in summer season, presence-only habitat models for bighorn sheep and mountain goats in the general study area.

Covariate Development

All covariates used for the presence-only modeling effort were developed as part of preliminary analyses conducted in 2010 and were based on bighorn sheep and mountain goat biology, published literature and readily accessible GIS layers. Each of six habitat parameters (Table 1) was measured using 30-by-30 m grid-cell resolution, as this was the most common resolution of available data sources and allowed for reasonable spatial resolution and feasible data processing. All related layers were compiled and analyzed in a geographic information system (GIS). GIS analyses and modeling were completed using ESRI’s ArcMap Ver. 9.1 software (available from: http://www.esri.com). Hawth’s Analysis Tools extension was also used to allow extraction of statistics from various GIS layers concurrently.

107

Table 1 – List of covariates considered in presence-only modeling effort. Covariate List Description

Distance to escape terrain slope greater than 36 degrees

Elevation meters

East/West Aspect Gross et al. 2002

North/South Aspect Gross et al. 2002

Percent tree MODIS data re-sampled to 30 meters

Ruggedness Poole et al. 2009

TCAP wetness tasseled cap wetness (band 3), Landsat derivative (July 2000)

Habitat characteristics within the study area were represented using a Digital Elevation Model (DEM), Landsat Enhanced Thematic Mapper (ETM) satellite imagery and a MODIS scene. DEM layers were obtained from the Montana Natural Resource Information System (NRIS) (http://nris.mt.gov). Landsat imagery data were obtained from the Global Land Cover Facility website (www.landcover.org) and MODIS data were obtained from Montana State University’s Landscape Biodiversity Lab (http://www.montana.edu/hansen/). Distance to escape terrain (Escape), elevation (Elev), aspect (Asp) and ruggedness (Rug) were developed using DEM layers. Escape terrain has been defined in existing literature using a range of slopes from 27 degrees to greater than 40 degrees (Zeigenfuss et al. 2000, Gross et al. 2002, DeCesare and Pletscher 2006, Poole et al. 2009). For these analyses we defined escape terrain as slope greater than 36 degrees. A distance layer was created for the entire study area to determine the distance to the nearest escape terrain pixel. Two continuous covariates were derived from an aspect layer created for the study area. These covariates described N-S and E-W exposures and ranged from 0 to 180, with 0 indicating north or east, respectively, and 180 indicating south or west, respectively (Gross et al. 2002).

The data layer created from a MODIS scene covered the entire GYA and contained a percent tree cover (Tree) for each pixel. The original spatial resolution of the MODIS scene was 250 meters; however, it was resampled to 30 meters to ensure the resolution was comparable to other data layers used for analyses. Terrain ruggedness was calculated using the curvature function in ArcView 9.1. A curvature grid of 30 m resolution was generated and run through a moving window analysis for standard deviation within a 100 m radius of each grid cell. The result was a measure of the variability of the rate of change in slope for each grid cell. A high value was

108

considered to be more rugged habitat as it would be indicative of a high degree of change in slope and complexity of surrounding cliffs (Poole et al. 2009).

Landsat data were converted from digital numbers to at-sensor reflectance values for normalization across scenes prior to use in data transformation and covariate creation. Landsat imagery for the study area consisted of three scenes collected in July of 2000. Image bands were downloaded individually in digital number format and were converted to at-sensor reflectance to be used for creating covariates. Preprocessing of Landsat imagery and creation of remotely sensed variables was done using Research System’s Inc. ENVI. v4.1. (personal communication: M. Zambon). The TCAP transformation was used to convert the original covariant Landsat data into new bands making up three unrelated indices: brightness, greenness and wetness which can reflect the condition of vegetation and soil (Sheng et al. 2011). Tasseled Cap Transformation Band 3 from Landsat ETM + imagery (Wet) was included as a surrogate for the amount of moisture found across the study area (lower values indicate wetter habitat) (Crist and Kauth 1986).

Statistical Modeling

Including two or more strongly correlated covariates in the same model may confound results and make interpretation difficult (Farrar and Glauber 1967). All pairwise correlations between covariates were explored to ensure highly co-linear covariates were not included in the same candidate models. The remaining covariates included in the global model were: logit π = β0 + β 1 (distance to escape terrain) + β2 (elevation) + β3 (east/west aspect) + β4 (north/south aspect) + β5 (percent tree) + β6 (ruggedness) + β7 (TCAP wetness) which was used for subsequent modeling efforts.

Data were imported into program “R”. Density plots were created depicting the use and available location distributions for each of the seven covariates considered for mountain goat and bighorn sheep summer habitat models. These plots allow for easy interpretation of data distributions. If distributions of use and available points are comparable, it is unlikely the covariate in question represents a selection criterion. Where use and available distributions are different, there is evidence that the covariate may be influential to animal distributions.

We used logistic regression to relate the relative probabilities of use by bighorn sheep and mountain goats to the covariates of interest (Manly et al. 2002). Because these were exploratory analyses, all possible additive combinations of selected covariates were considered in modeling using the dredge function in program R. The strength of support that the data gave to each model was evaluated by ranking models with Akaike's Information Criterion and model weights (Burnham and Anderson 2002). Models within 4 ΔAIC units of the top model were considered to have received comparable support from the data. For each of the coefficient estimates resulting from top-models 95% confidence intervals were constructed.

109

Model outputs were in log odd units and were used for presentation of summer distribution maps. The ArcGIS raster calculator was used to apply regression coefficients from the top models to the covariates for the study area to create summer habitat suitability maps for mountain goats and bighorn sheep. The habitat suitability map constructed during this modeling exercise was used to identify regions within the general study area to be intensively sampled during formal occupancy surveys during the summer of 2011. Each survey area was stratified into four categories of species occurrence probability: very low (40%), low, medium and high (20%) probabilities of species occurrence. Due to the logistical and practical constraints of conducting ground surveys in mountainous terrain, occupancy survey sampling was limited to all areas predicted as high (20%) and medium (20%) suitability habitats and approximately half (10%) of the low suitability habitat defined by these models.

OCCUPANCY MODELING EFFORTS

Development of Formal Occupancy Surveys

In order to divide the landscape into discrete sampling units for occupancy surveys, a 100 x 100 meter grid cell system was placed over an aerial photo of each of the survey regions (Figure 3). The grid cell system was designed to allow observers to record animal locations with fine spatial resolution (within 100 meters) and to record areas where they looked for, but did not detect animals.

The field computers used for data collection were the Geo Mesa by Juniper Systems. ArcPad was loaded onto each of four field computers and custom modifications were made to ArcPad software to allow observers to enter animal locations, group sex-age composition, and behavior data accurately and efficiently into field computers during surveys.

Observers conducted occupancy surveys within each of the four survey regions. A survey event consisted of three to four day backpacking trips in each of these regions. Travel routes were placed along trails and ridges, based on what was logistically feasible and safe for observer travel, while affording observers a reasonable field of view. Observation points were systematically placed every three kilometers along travel routes using ArcGIS. A random number generator was used to select a starting point within the first three kilometers of the intended travel route. Observation points were then systematically selected every three kilometers from the starting point, as this distance allowed the survey crew to traverse a reasonable amount of the survey area during a three to four day trip.

110

Figure 3 – Screen shot of ArcPad in Juniper Systems Geo Mesa field computers. Grid overlay is made up of 100x100 meter cells, which define the individual units surveyed for bighorn sheep and mountain goat presence-absence. Green dots indicate grid cells where observers looked but did not detect bighorn sheep or mountain goats, blue triangles indicate bighorn sheep observations and red triangles indicate mountain goat observations.

A 500 meter radius buffer was placed around each observation point to allow observers to select a site for conducting an occupancy survey that afforded maximum visibility of the landscape.

The surrounding viewsheds were surveyed and observations recorded into field computers. If there were multiple options for survey travel routes (i.e., a fork in the trail) direction of travel was based on logistics. Upon arrival at each systematically selected observation point, a coin was flipped to determine which observers would survey a given viewshed (i.e., east or west side of a ridgeline) from the observation point. The exception was when an observer had already surveyed parts of the adjacent area. In order to avoid confusion about which areas were previously surveyed by another observer the original observer surveyed the aforementioned adjacent area. Observers attempted to visit each predetermined observation point during a survey event. Certain pre-selected observation points were not visited, due to logistical constraints in the field (i.e., weather, challenging terrain, etc). Observers made every attempt to visit these points and survey visible viewsheds on subsequent visits.

111

Animals will not always be detected when they are present in a sampling unit. Patterns of detection and non-detection accrued through repeat surveys in a short period of time may be used to estimate detection probabilities (MacKenzie et al. 2002, 2003, Mackenzie and Royle 2005, Mackenzie et al. 2006). In an effort to accrue adequate animal detections for estimation of detection probabilities, observers alternated between double and single observation methods. During double observer surveys, the observers determined the most effective way to conduct one independent survey per team member. All communication regarding visual observation of presence or absence of animals ceased. Observers positioned themselves on opposite sides of a natural barrier (i.e., rock, vegetation) when possible and did not view the field computer of the other observer once a survey period had begun. These measures helped to ensure that data collected by each observer were independent. After surveys at selected observation points were completed, team members reconvened and determined which animals were detected by both observers and which animals were only detected by one observer. Data were recorded onto detection probability worksheets. Observers then traveled together to the next systematically selected observation point until all possible observation points had been visited and the associated visible viewsheds had been surveyed. Efforts were focused on visiting more sampling units rather than conducting more visits per sampling unit during a survey period, as additional visits do not always notably increase the accuracy of detection probability estimates (MacKenzie and Royle 2005). If previously surveyed grid cells were visible on a future trip, those grid cells were re-surveyed.

Before scanning grid cells within a viewshed for animals, observers agreed on a reasonable topographic boundary defining the area to be surveyed. This was done at the start of both single and double observer surveys to ensure observers were covering a comparable amount of land so survey duration was not drastically different. Date, survey point ID, survey start and end times and observer location (UTM, WGS-1984) were recorded. Average wind-speed (meters/second) over a ten-second period and temperature (degrees Fahrenheit) were measured using a Kestrel 2000 Pocket Weather meter at the beginning and end of each survey period. Data were recorded into field notebooks. Observers then scanned all grid cells within the viewshed for mountain goats and bighorn sheep using 10x42 binoculars and 20x60 spotting scopes. When animals were detected, the predominant behavior of the group at first observation (feeding, resting, traveling, or other) was recorded and each group was assigned a unique number for the day. A group was considered a single animal or individuals of a species within approximately 250 meters of each other. Animals separated by more than 250 meters at first detection were considered separate groups. Group numbers allowed us to record animals as they traveled through multiple grid cells and minimize the chance of animals being counted multiple times. Animals were counted before any attempt at age/sex classification was made in an effort to increase the likelihood of recording all visible animals before they moved out of sight. Counts and point locations of detected animals were recorded directly into the field computers.

112

Using binoculars and spotting scopes, observers attempted to classify individual animals within a group by age and sex classes, including mature male, young male, female, yearling or young of year. Bighorn sheep classes were determined using methods described by Geist (1971, and were based on a combination of possible identifying features including horn size, body size and positive identifying features, including urination posture and external genitalia. It is frequently difficult to differentiate between yearling males and ewes. If an observer was unable to differentiate between the two, the animal was classified as a female. Methods described by Chadwick (1983) were used to classify mountain goats and were based on a combination of possible identifying features including horn mass and shape, body size, rump cleanliness and positive identifying features, including external genitalia and urination posture. If it was not possible to determine the age or sex of an animal, it was recorded as unknown.

Each grid cell surveyed by an observer was assigned a ranking from 1-4 based on the percentage of the grid cell visible to the observer: 1- 1%-25%, 2- 26%-50%, 3- 51%-75% and 4- 76%- 100%. This ranking system was based on topography (i.e. a cell going over a ridgeline or in a draw) and was not affected by cover or ruggedness. Due to high snow levels in 2011 and our inability to quantify snow cover using a GIS layer, we did not survey large areas of the landscape fully covered in snow. These areas would not be truly representative of the landscape features available at the time of the survey.

All surveyed grid cells where no bighorn sheep or mountain goats were detected were identified and recorded in the field computer by placing a non-detection point in approximately, the middle of the grid cell. If a group of animals moved into a previously unoccupied grid cell during a survey, the observer changed the status of the grid cell and included relevant animal information. If groups of animals were encountered while observers were traveling between the predetermined observation points, they were counted and classified and their point location recorded as an opportunistic sighting. Surveys were not conducted during periods of extreme inclement weather (i.e., high winds, heavy rain) due to a decrease in observer ability to locate animals on the landscape. Upon returning from the field, all supplementary data recorded in field notebooks was logged onto data sheets to ensure consistency of recorded data and to streamline data entry.

Survey Data Entry

Data from field computers were downloaded after each field data collection event, upon return to the office. The Access database designed for this study consisted of four main tables to store the point and demography data for bighorn sheep and mountain goat observations (Figure 4). The first table, Field Computer, stored all of the survey data downloaded from field computers. The second table, Notebook, stored all of the survey information not recorded directly into field computers. The third table, Detection Probability, stored survey data related to groups of animals detected during double observer surveys. The fourth table, Opportunistic, stored the attribute data and the demography data associated with each opportunistic animal point location recorded 113

outside of a survey event. The Field Computer table was linked to the Notebook table through the Survey ID, a unique ID code generated for each survey conducted. The Detection Probability table was also linked to the Notebook table through the Survey ID.

Figure 4 – A screenshot of three of the four main tables (Field Computer, Notebook Data and Detection Probability) are displayed as well as the relationships between them. The Opportunistic table was not related to the other tables in the database.

Development of Additional Covariates

Species distributions and habitat selection depend on various biological processes, many of which are affected by solar radiation. For example, Keating et al. (2007) demonstrated high concentrations of ungulates during the winter in locations receiving relatively high levels of solar radiation. Frequently, habitat studies include covariates such as aspect as a proxy for solar radiation, rather than including it as a distinct covariate (Keating et al. 2007). We found it difficult to determine the biological significance of the aspect covariates as used in previous modeling efforts and as a result, we decided to exclude these covariates from further modeling efforts and instead we examined other indices of solar radiation. We used equations developed by McCune and Keon (2002) that were easily executed in a spreadsheet to estimate direct incident radiation and an index of heat load using slope, aspect and latitude. Although potential direct incident radiation is symmetrical around a north/south axis the same is not true for heat load. A slope receiving afternoon sun will likely be warmer than an equivalent slope receiving 114

morning sun. To account for this McCune and Keon rescaled aspect so that a value of zero represents the coolest aspect (NE) and a value of one represents the warmest aspect (SW). There is no way to convert the results into a collective measure of temperature so the output must be considered a unit-less index of heat load (McCune and Keon 2002). Although it appears a useful way to examine relative heat load across the study area, this method fails to take into account cloud cover and adjacent topography. As a result, we also decided to examine solar radiation as calculated using the Solar Radiation toolset in ArcGIS 9.3 Spatial Analyst Extension based on methods developed by Fu and Rich (2002). The tools available allow the user to process a huge amount of information including atmospheric effects, site latitude, elevation, slope, aspect, daily sun angle shift and adjacent topographical shading, which would otherwise be too time consuming. The related calculations can be performed for individual points or over large geographic areas and are carried out in a multi-step process. The process begins with calculation of an upward-looking hemispherical viewshed and is calculated based on topography. This viewshed is then laid over a direct sunmap and a diffuse skymap, allowing for estimation of direct radiation and diffuse radiation, respectively. This process is then repeated for every point or area of interest and used to produce a map of insolation (Huang and Fu 2009). Our hope is that inclusion of these site specific features may help capture some of the factors affecting spatial distribution patterns of air and soil temperatures, patterns of snow melt, moisture content of soil and the amount of light available to enable plant productivity at a small scale.

The ruggedness measure used in previous presence-only modeling included only the variability of the rate of change in slope for each grid cell. When examining landscape ruggedness and how it may affect resource selection by animals it is important to include changes in both aspect and gradient of slope to truly capture features of surrounding topography important to animals. It is likely that bighorn sheep and mountain goats perceive ruggedness and slope differently when assessing escape and forage terrain and therefore it seems logical to quantify heterogeneity in both aspect and slope (Sappington et al. 2007). Using layers previously generated in ArcView 9.3 including slope, aspect and contour and an ArcView Script developed by Sappington et al. (2007) we created a new layer of vector ruggedness measures which combined variation in aspect and slope into a single measure (script available online from the Environmental Systems Research Institute ArcScripts website: www.esri.com/arcscripts). This method may capture more heterogeneity in the landscape than indices based only on changes in slope or elevation.

Animals select landscape features at various spatial scales ranging from geographic areas to individual plants (Johnson 1980). Scale is defined here as spatial extent or area as opposed to grain or resolution of measured landscape features (Kie et al 2002). The scale at which an animal selects a resource is dependent on vulnerability to predation, the cost of foraging and the spatial distribution of the resource (Senft et al. 1987, Gustafason 1998, Johnson et al. 2002, Kie et al. 2002). Our ability to capture differences in resource selection and to make associations between animals and their habitat will vary along with scale (Boyce 2006). If data are not explored at the

115

scales that are important to animals, we are likely to misinterpret the results of habitat suitability analyses (Wiens 1989). Examining too large of an area may preclude detection of habitat heterogeneity and examining too small of an area may cause under sampling of variance in local habitat characteristics (Boyce et al 2006). In an effort to account for heterogeneity of the habitat and local variation of habitat characteristics, we decided to examine habitat selection at two different spatial scales by placing a buffer with 100 meter radius and 500 meter radius around observation points for both detection and non-detection records and examining landscape features within each buffer.

RESULTS

PRESENCE-ONLY MODELING EFFORT

The use versus available covariate distributions for mountain goats during the summer season suggested mountain goats selected for areas closer to escape terrain in drier, rugged, higher elevation terrain than what was available. There was also some evidence for selection of areas with greater tree-cover. Interpretations of the density plot patterns were supported by results of model selection. The two top-models accounted for nearly all of the model weight (Table 2). Both of these models included five of the seven covariates, with only east/west aspect not appearing in both models. Coefficient estimates were similar for both models and the only coefficient confidence interval that spanned zero was wetness, in one of the two top-models (Table 3). The use versus available covariate distributions for bighorn sheep during the summer season suggested bighorn sheep and mountain goats selected for similar locations in rugged habitat, closer to escape terrain and at high elevations. Bighorn sheep also appeared to select for wetter areas with fewer trees compared to areas selected by mountain goats. Model comparisons resulted in one top model receiving 100% of the model weight (Table 2). The top model included six of the seven covariates considered in models. Point estimates and associated 95% confidence intervals for the six covariate coefficients appearing in the model did not span zero (Table 3). The resulting habitat suitability maps were very similar for both species so we only present the bighorn sheep suitability map in this document (Figure 5). A total area of 250km2, comprised of four regions, was identified for conducting occupancy surveys in the summer of 2011.

116

Table 2 – Model selection results for resource selection probability function models examining the effects of 7 landscape covariates on bighorn sheep and mountain goat summer and rut habitat selection. All models are presented along with the number of parameters (K), ∆AICc value and Akaike weight (ωi). Bighorn Sheep

Model ID Model AIC K AICc ∆AICc ωi

Escape + Elev + EWAsp + NSAsp + Rug 1 3508 7 3494 0 0.91 + Wet

Escape + Elev + EWAsp + NSAsp + 2 3515 8 3499 5.196 0.07 Tree + Rug + Wet

Mountain Goat

Model ID Model AIC K AICc ∆AICc ωi

Escape + Elev + EWAsp + Tree + Rug 1 1537 7 1523 0 0.52 + Wet

2 Escape + Elev + Tree + Rug + Wet 1535 6 1523 0.174 0.48

3 Escape + Elev + Tree + Rug + Wet 1544 8 1528 5.047 0.04

117

Table 3 – Coefficient estimates of all models within 4 AIC units of the top model for bighorn sheep and mountain goat summer habitat selection. Coefficient estimates in bold font include confidence intervals that do not span zero. Bighorn Sheep Mountain Goat

Model ID 1 1 2

Covariate

Intercept 17.3 -23.07 -27.69

(-18.9, -15.7) (-37.96, -18.35) (-37.97,-18.44)

Escape -0.006 -0.005 -0.005

(-0.009, -0.005) (-0.008,-0.003) (-0.007,-0.003)

Elev 0.003 0.003 0.003

(0.003,0.004) (0.002,0.004) (0.003,0.004)

EWAsp 0.008 0.002 -

(0.005,0.011) (-0.004,0.008)

NSAsp -0.005 - -

(-0.009,-0.002)

Tree - 0.027 0.026

(0.018,0.039) (0.017,0.034)

Rug 2.66 3.856 3.887

(2.02,3.24) (3.093,4.685) (3.109,4.709)

Wet 122.6 -17.56 -34.06

(122.5,138.7) (-31.82, 18.72) (-50.17, -14.53)

118

Figure 5 – Bighorn sheep relative habitat suitability map for the study area, resulting from application of regression coefficients from the top models to the covariates for the study area. Also depicted are the four survey regions within the area (Region 1, Region 2, Region 3, and Region 4) that were sampled during occupancy surveys during summer 2011.

FORMAL OCCUPANCY FIELD SURVEYS

From 8th of July to the 20th of September 2011 the total number of observer days in the field was 113 resulting in approximately 240 hours of occupancy survey effort. Fifty-seven viewsheds were surveyed by a single observer and an additional 77 viewsheds were surveyed by two, independent observers. A total of 6,909 100x100 meter grid cells were surveyed (Figure 6) on at least one occasion with 3,392 of those grid cells visited on multiple occasions during the season. Fifteen groups of bighorn sheep were detected during occupancy surveys with an average group size of 10.3 individuals, a median group size of 7.0 individuals, and a group size range of 44.0 individuals with a SD of 10.9 individuals. One hundred fifty four individual bighorn sheep were observed and classified during occupancy surveys with 81 females, 57 young of the year, 1 mature male, 9 yearlings, and 6 unknown. Fifty grid cells were occupied by bighorn sheep during surveys and observers recorded bighorn sheep in an additional 18 grid cells while traveling between surveys.

Thirty-four groups of mountain goats were detected during surveys with an average group size of 2.3 individuals, a median group size of 1.0 individual, and a group size range of 17.0 individuals with a SD of 3.2 individuals. Seventy-nine individual mountain goats were observed and

119

classified during occupancy surveys with 17 females, 13 young of the year, 1 mature male, 5 yearlings and 43 unknown. Fifty-nine grid cells were occupied by mountain goats during surveys and observers recorded mountain goats in an additional 36 grid cells while traveling between surveys.

A total of 13 groups of bighorn sheep were detected during double observer surveys and 10 of these groups were detected by both observers (76.9%). A total of 22 groups of mountain goats were detected during double observer surveys and 12 groups were detected by both observers (54.5%).

120

Figure 6 – Maps of each individual survey region (Region 1, Region 2, Region 3 and Region 4) and the observations recorded in each, during 2011 field season. Green dots indicate neither bighorn sheep nor mountain goats were observed, blue triangles indicate a bighorn sheep observations and red triangles indicate mountain goat observations. 121

Figure 6 – (continued). 122

FUTURE EFFORTS

We are currently preparing the data collected during the summer of 2011 as well as the new covariate GIS layers for analyses. The results of these efforts will be presented in a final thesis product in August of 2012. Occupancy surveys will be continued for another two years by a new master’s student, Jesse DeVoe, and his crew beginning in the summer of 2012. By increasing the data available for analyses we hope we will be able to construct habitat models for both bighorn sheep and mountain goats and validate these models in additional study areas.

ACKNOWLEDGEMENTS

Management data for both bighorn sheep and mountain goats were generously provided by the Montana Department of Fish, Wildlife and Parks, including Julie Cunningham, Tom Lemke, and Karen Loveless and Wild Things Unlimited via Steve Gehman. Special Thanks to Dan Tyers for providing expertise about the study area as well as many biological insights. Logistic support was provided by Christie Hendrix, Mike Zambon and Mike Sawaya. Several landowners were gracious enough to allow us access to their land and to take an interest in our work including the B-Bar Ranch, Royal Teton Ranch, and Grizzly Creek Ranch. Special thanks to Tiffany Allen, Braden Burkholder and Jesse DeVoe for assistance with GIS and editing of this document and to the field crew, Carson Butler and Elizabeth Flesch.

LITERATURE CITED

Adams, L.G., K.L. Risenhoover, and J.A. Bailey. 1982. Ecological relationships of mountain goats and Rocky Mountain bighorn sheep. Northern Wild Sheep and Goat Council 3:9- 22.

Begon, M., C.R. Townsend, and J.L. Harper. 2006. Ecology from Individuals to Ecosystems, 4th edition. Blackwell, Cambridge.

Beyer, H.L. 2004. Hawth's Analysis Tools for ArcGIS. Available at http://www.spatialecology.com/htools.

Boyce, M.S. 2006. Scale for resource selection functions. Diversity and Distributions 12:269- 276.

Burnham, K.P., and D.R. Anderson. 2002. Model selection and multimodel inference: A practical information-theoretic approach. 2nd edition. Springer Science and Business Media, Inc., New York, New York.

Chadwick, D.H. 1983. A Beast the Color of Winter. University of Nebraska Press. Lincoln, Nebraska.

123

Cox, G.W. 1999. Alien species in North America and Hawaii: impacts on natural ecosystems. University of Chicago Press. Chicago & London. 201 pp.

Crist, E.P., and R.J. Kauth. 1986. The tasseled cap de-mystified. Photogrammetric Engineering and Remote Sensing 52:81-86.

Daszak, P., A.A. Cunningham, and A.D. Hyatt. 2000. Emerging infectious diseases of wildlife: threats to biodiversity and human health. Science 287:443-449.

DeCesare, N.J., and D.H. Pletscher. 2006. Movements, connectivity, and resource selection of Rocky Mountain bighorn sheep. Journal of Mammalogy 87:531-538

Elith, J., C.H.Graham, R.P. Anderson, M.Dudik, S. Ferrier, A. Guisan, R.J. Hijmans, F. Huettmann, J.R. Leathwick, A. Lehmann, J. Li, L.G. Lohmann, B.A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J. McC., M. Overton, A.T. Peterson, S.J. Phillips, K. Richardson, R. Scachetti-Pereira, R.E. Schapire, J. Soberon, S. Williams, M.S. Wisz, and N.E. Zimmermann. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29:129-151.

Elton, C. 1927. Animal Ecology. Sidgwick and Jackson, London. 207 pp.

Farrar, D.E., and R.R. Glauber. 1967. Multicollineanty in regression analysis: the problem revisited. The Review of Economics and Statistics 49:92-107.

Fritts, T.H., and G.H. Rodda. 1998. The role of introduced species in the degradation of island ecosystems: A case history of Guam. Annual Review of Ecology and Systematics 29: 113-140.

Fu, P., and P. M. Rich. 2002. A Geometric Solar Radiation Model with Applications in Agriculture and Forestry. Computers and Electronics in Agriculture 37:25–35

Gause, G. F. 1934. The Struggle for Existence. Williams and Wilkins, Baltimore (reprinted 1964, Hafner, New York).

Geist, V. 1971. Mountain Sheep: A Study in Behavior and Evolution. University of ChicagoPress, Chicago, Illinois.

Gormley, A.M., D.M. Forsyth, P. Griffioen, M. Lindeman, D.S.L. Ramsey, M.P. Scroggie, and L. Woodford. 2011. Using presence-only and presence-absence data to estimate the current and potential distributions of established invasive species. Journal of Applied Ecology 48:25-34.

Gross, J.E., M.C. Kneeland, D.F. Reed, and R.M. Reich. 2002. GIS-based habitat models for mountain goats. Journal of Mammalogy 83:218-228. 124

Guenzel, R.J. 1980. A population perspective of successful mountain goat transplants. Biennial Symposium Northern Wild Sheep and Goat Council 2:403-457.

Gustafson, E.J. 1998. Quantifying Landscape Spatial Pattern: What Is the State of the Art? Ecosystems 1:143-156.

Hijmans, R.J., K.A. Garrett, Z. Huaman, D.P. Zhang, M. Schreuder, and M. Bonierbale. 2000. Assessing the geographic representativeness of genebank collections: the case of Bolivian wild potatoes. Conservation Biology 14:1755-1765

Hirzel, A.H., G. LeLay, V. Helfer, C. Randin, and A. Guisan. 2006. The ability of habitat suitability models to predict species presences. Ecological Modeling 99:142-152.

Hobbs, N.T., M.W. Miller, J.A. Bailey, D.F. Reed, and R.B. Gill. 1990. Biological criteria for introductions of large mammals: using simulation models to predict impacts of competition. Transactions North American Wildlife and Natural Resource Conference 55:620-632.

Huang, S., and P. Fu. 2009. Modeling small areas is a big challenge, using the solar radiation analysis tools in ArcGIS Spatial Analyst. ArcUser, Spring 2009:28-31.

Hutchinson, G.E. 1957. Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology 22:415-427.

Johnson, D.H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71.

Johnson, C.J., K.L. Parker, D.C. Heard, and M.P. Gillingham. 2002. Movement parameters of ungulates and scale-specific responses to the environment. Journal of Animal Ecology 71:225-235.

Keating, K.A., and S. Cherry. 2004. Use and interpretation of logistic regression in habitat selection studies. The Journal of Wildlife Management 68:774-789.

Keating, K.A., P.J.P. Gogan, J.M. Vore, L.R. Irby. 2007. A simple solar radiation index for wildlife habitat studies The Journal of Wildlife Management 71:1344-1348.

Kie, J.G., R.T. Bowyer, M.C. Nicholson, B.B. Boroski, and E.R. Loft. 2002. Landscape heterogeneity at differing scales: effects on spatial distribution of mule deer. Ecology 83:530-544.

Lack, D. 1947. Darwin's Finches. Cambridge University Press, New York, New York.

125

Laundré, J.W. 1994. Resource overlap between mountain goats and bighorn sheep. Great Basin Naturalist 54: 114-121.

Lele, S.R., and J.L. Keim. 2006. Weighted distributions and estimation of resource selection probability functions. Ecology 87: 3021-3028.

Loiselle, B.A., C.A. Howell, C.H. Graham, T. Brooks, P.H. Williams, J.M. Goerck, and K.G. Smith. 2003. Avoiding pitfalls of using species distribution models in conservation planning. Conservation Biology 17: 1591-1600.

Mack, R.N., D. Simberloff, W.M. Lonsdale, H. Evans, M. Clout, and F.A. Bazzaz. 2000. Biotic Invasions: causes, epidemiology, global consequences and control. Ecological Applications 10:689-710.

MacKenzie, D.I., J.D. Nichols, G.B. Lachman, S. Droege, J.A. Royle, and C.A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248-2255.

MacKenzie, D.I., J.D. Nichols, J.E. Hines, M.G. Knutson, and A.B. Franklin. 2003. Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology 84:2200-2207.

MacKenzie, D.I. and J.A. Royle. 2005. Designing occupancy studies: general advice and allocating survey effort. Journal of Applied Ecology 42:1105–1114.

MacKenzie, D.I. 2005. What are the issues with presence-absence data for wildlife

managers? The Journal of Wildlife Management 69:849-860.

MacKenzie, D.I., J.D. Nichols, J.A. Royle, K.H. Pollock, L.L. Bailey, and J.E. Hines. 2006. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Academic Press, New York.

Manly, B.F., L.L. McDonald, D.L. Thomas, T. L. McDonald, and W.P. Erickson. 2002. Resource Selection by Animals: Statistical Design and Analysis for Field Studies. Second edition. Kluwer Press, New York, New York, USA.

McCune, B., and D. Keon. 2002. Equations for potential annual direct incident radiationand heat load. Journal of Vegetation Science 13: 603-606.

Noss, R.F., and A.Y. Cooperrider. 1994. Saving Nature's Legacy: Protecting and Restoring Biodiversity. Washington, DC: Island Press.

126

Pianka, E.R. 1981. Competition and niche theory. Pages 167–196 in R.M. May, editor. Theoretical ecology: principles and applications. Second edition. Sinauer, Sunderland, Massachusetts, USA.

Picton, H.D., and T.N. Lonner. 2008. Montana’s Wildlife Legacy Decimation to Restoration.Media Works Publishing, Bozeman Montana. 286 pp.

Poling, T.D., and S.E. Hayslette. 2006. Dietary overlap and foraging competition between mourning doves and Eurasian collared-doves. Journal of Wildlife Management 70:998- 1004.

Poole, K.G., K. Stuart-Smith, and I.E. Teske. 2009. Wintering strategies by mountain goats inthe interior mountains. Canadian Journal of Zoology 87:273-283.

R Development Core Team. 2011. A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3- 900051-07-0, URL http://www.R-project.org

Reed, D.F. 1986. Alpine habitat selection in sympatric mountain goats and mountain sheep. Biennial Symposium North American Wild Sheep and Goat Council 5:421-422.

Reed, D.F., and K.A. Green. 1994. Mountain goats on Mount Evans, Colorado - conflicts and the importance of accurate population estimates. Biennial Symposium Northern Wild Sheep and Goat Council 9:139-143.

Reese, G.C., K.R. Wilson, J.A. Hoeting, and C.H. Flather. 2005. Factors affecting species distribution predictions: a simulation modeling experiment Ecological Applications 15:554-564.

Sappington, J.M., K.M. Longshore, and D.B. Thompson. 2007. Quantifying landscape ruggedness for animal habitat analysis: a case study using bighorn sheep in the Mojave Desert Journal of Wildlife Management 71:1419-1426.

Senft, R.L., M.B. Coughenour, D.W. Bailey, L.R. Rittenhouse, O.E. Sala, and D.M. Swift. 1987. Large herbivore foraging and ecological hierarchies. Bioscience 37:789-799.

Schluter, D., T.D. Price, and P.R. Grant. 1985. Ecological character displacement in Darwin’s finches Science 227:1056-1059.

Shannon, N.H., R.J. Hudson, V.C. Brink, and W.D. Kitts. 1975. Determinants of spatial distribution of rocky mountain bighorn sheep. Journal of Wildlife Management 39:387- 401.

127

Sheng, L., J. Huang, X. Tang. 2011. A tasseled cap transformation for CBERS-02B CCD data. Journal of Zhejiang University-SCIENCE B (Biomedicine & Biotechnology) 12:780- 786.

Varley, N.C., and J.D. Varley. 1996. Introductions of mountain goats in the greater Yellowstone ecosystem: they’re here to stay or are they? Northern Wild Sheep and Goat Council 10:113-117.

Weins, J.A. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397.

Zeigenfuss, L.C., F.J. Singer, and M.A. Gudorf. 2000. Test of a modified habitat suitability model for bighorn sheep. Restoration Ecology 8:38-46.

128

GPS-VHF Telemetry Studies of Mountain Ungulates in the Greater Yellowstone Area Jesse DeVoe and Robert A. Garrott

ABSTRACT

Telemetry studies on bighorn sheep (Ovis canadensis) and mountain goats (Oreamnos americanus) in the greater Yellowstone area (GYA) are relatively rare, especially in comparison to other large mammals. There is therefore a significant dearth of detailed information on mountain ungulate demographic and spatial ecology as well as competition dynamics between the non-native mountain goat and the native bighorn sheep. The Mountain Ungulate Research Initiative is seeking to gain this valuable management and conservation information by initiating GPS and VHF radio telemetry studies across the GYA. We have selected ten study sites that represent the varying ecological settings of this ecosystem with differences in climate, geology, herd size, disease history, land use and management, migratory and non-migratory herds, sympatric and allopatric herds, and high and low elevation ranges. In addition, we have developed a dual collar, multiple deployment strategy to efficiently maximize collection of ecological data and support long-term research goals. This includes the deployment of a GPS collar simultaneously with a VHF collar for each animal instrumented. After two years of fine spatial- and temporal-scale data collection the GPS collars will release for recovery while the VHF collars will remain on animals to obtain an additional five years of demographic data. The recovered GPS collars will then be refurbished and redeployed with new VHF collars on additional animals.

129

INTRODUCTION

The greater Yellowstone area (GYA) embodies a variety of mountain ranges with diverse geology, topography, and climate regimes that result in differing ecological settings for the two mountain ungulates that occupy the region, bighorn sheep and mountain goats. Both species are important large mammals with high intrinsic value to the people occupying the region as well as the millions of visitors each year that come from throughout the world to enjoy the natural beauty and wildlife of this last relatively intact temperate ecosystem.

Basic research to understand the ecology of these two prominent large mammals in the GYA have been quite limited compared to all the other large mammals that occupy the region, with most agency resources expended on routine population monitoring. The GYA Mountain Ungulate Research Project was initiated in fall 2009 with funding from Yellowstone National Park to begin filling this scientific void in order to inform policy and enhance management and conservation of these species. We have developed a strong collaboration with all the state and federal agencies responsible for wildlife and public land management in the mountainous regions of the GYA including the National Park Service (Yellowstone and Grand Teton NP), US Forest Service, Wyoming Game and Fish Department, Idaho Department of Fish and Game, and Montana Fish Wildlife and Parks, and have also received support from several non-government sportsman organizations and the Canon Corporation.

All of these organizations have varying and overlapping interest for initiating comparative telemetry studies of bighorn sheep and mountain goats in the region to address a myriad of important ecological questions. Basic ecological information is needed on population dynamics and vital rates (survival, reproduction) and the factors that influence annual variation in demographic performance. There is also a dearth of information on spatial ecology including seasonal distribution patterns, movement dynamics, and migration corridors. Better seasonal habitat models are also needed to inform policy on land use management and conservation, as well as to aid in identifying appropriate sites for restoration of native bighorn sheep. Little is known about metapopulation dynamics and the potential for genetic exchange and disease transmission. The region is also experiencing unprecedented climate warming that is predicted to have particularly strong impacts on alpine and subalpine species and communities. In addition, the presence and continued range expansion of non-native mountain goats that were introduced into some mountain ranges of Montana and Idaho between the late 1940s and early 1970s has generated an array of sites with varying levels of occupancy, both allopatric (one species) and sympatric (both species) with bighorn sheep populations. With mountain goat occupancy of sympatric sites across the GYA ranging from 5 to 60 years, a valuable and unique opportunity exists to collect information on competition and population dynamics for both mountain ungulates. The challenge, then, is to implement a large-scale study to sample across these various allopatric and sympatric sites while also capturing the varied environmental attributes that are present in mountain ungulate habitat across the GYA. 130

The Mountain Ungulate Project, through significant collaborator effort and input, has selected several bighorn sheep and mountain goat herds throughout the mountain ranges of the GYA for the implementation of telemetry studies to gain ecological knowledge that will enhance conservation and management of these species by all the natural resource agencies with administrative authority in the region. To address these broad objectives, we intend to simultaneously collar individuals of both species in several sympatric herds as well as a number of allopatric herds. The sympatric herds selected will have varying durations of mountain goat occupancy, from areas where bighorn sheep ranges have only recently been colonized by mountain goats to areas where mountain goats have shared ranges with bighorn sheep for many decades. While this large-scale, multi-species, and multi-study site research plan is ambitious, we have strong support from our collaborators, have secured adequate resources to initiate our research plan, and have received encouragement from a number of organizations and individuals that additional funding may be secured as we continue to develop the research program and produce creditable and high quality scientific products. We are committed to make this research program successful and are confident that it will result in valuable ecological knowledge that will contribute substantially to the conservation of local and regional populations of both bighorn sheep and mountain goats throughout the GYA.

Our specific objectives include the collection of information on 1) population demographics and dynamics, 2) seasonal distributions, 3) movement dynamics, including pathways of genetic exchange and disease transmission, and 4) competition dynamics. The fine spatial and temporal resolution data collected from these studies will also greatly expand and compliment the development of rigorous habitat models for both mountain ungulate species that are the focus of separate occupancy studies currently being initiated in many of the same areas.

The purpose of this section is to provide a detailed description of the telemetry studies, including the methodology, study site and herd selection, capture operations, and database overview. We also report on the progress and status of current and planned telemetry study efforts.

CAPTURE AND INSTRUMENTATION METHODOLOGY

In order to collect information on mountain ungulate populations, demographic characteristics, competition, and movement dynamics, we have designed and begun to implement telemetry studies incorporating multiple study sites, dual collaring (2 types of collars on same animal) and multiple deployment (same collar deployed consecutively on different animals) strategies. This integrated research plan will maximize field and funding efficiencies and collection of both demographic and spatial data. If we can maintain the strong collaborations that have been built during the first few years of this initiative and broaden our funding base, we are confident that this large-scale long-term research plan will provide important ecological insights for both bighorn sheep and mountain goats that will significantly enhance conservation and management of these iconic mountain ungulates. The methodology for this strategy is outlined here. 131

DUAL COLLAR STRATEGY

The dual collaring strategy involves the deployment of both a GPS and VHF radio collar on a single individual. The advantage of GPS technology is that it provides fine-scale (precise) spatial data at regular, relatively short, time intervals. Such data are optimal for addressing questions of spatial ecology. Spatial studies will provide insights into movement dynamics at the scale of individuals important in defining discrete populations, identifying migration pathways and corridors, and describing patterns of fidelity, dispersal, and metapopulation dynamics. GPS technology is the most appropriate method for this effort as detailed spatial studies would require intensive and extensive aerial surveys if VHF telemetry were used. The unpredictability of flying weather and the inherent hazards of flying in mountainous terrain would limit both the spatial and temporal resolution of the data and, thus, erode the potential ecological insights that can be gained from such an effort.

The disadvantage, however, of GPS technology is that deployment on animals is limited to approximately 1 to 2 years due to short battery life which limits their utility for collecting demographic (survival, reproduction) data. The VHF collars, on the other hand, have the capacity for long term deployment (about 5-8 years) and are optimal for addressing questions of population dynamics. Understanding and estimating the basic vital rates of the populations, that is survival and reproduction of adults and survival and recruitment of young-of-the-year, is important knowledge for managing and conserving populations. In ungulates, these demographic processes are age-dependent and can vary from year-to-year depending on variability in warm and cold season weather which, in turn, influences forage quantity, quality, and availability. VHF telemetry is a simple, reliable, and economical tool for long term survival and reproduction studies of individual animals. Thus, the combined instrumentation of GPS and VHF collars on individuals will serve to integrate and maximize ecological insight in an efficient manner.

This dual collar strategy will be similar to those used successfully in studies on the endangered Sierra Nevada bighorn sheep (personal communication -Tom Stephenson, Calif. Dept. Fish and Game) and is currently being used for long-term mountain goat studies in southeastern Alaska (personal communication – Kevin White, Alaska Dept. Fish and Game). This strategy is also presently being implemented in new field studies of the Henry Mountain bison herd in Utah (personal communication – David Koons, Utah State Univ.).

For our study, adult bighorn sheep ewes and adult mountain goats will be equipped with the two collars. The GPS collar uses store-on-board data collection and archiving technology and is scheduled to drop off after two years. The VHF collar, which will remain on the animal for life, will be in a semi-dormant state for the first two years during which the GPS unit is archiving data and sending signals from its own VHF beacon; the VHF collar will then switch to its programmed working duty cycle (after the GPS collar has dropped off the animal) at the start of 132

year three and is expected to transmit for an additional 4 years before its battery is depleted (Figure 1).

Figure 1 – Transmitting schedule for the dual GPS and VHF collars for instrumentation on individuals of adult bighorn sheep ewes and mountain goats.

Based on discussions with numerous biologists throughout the western US, Canada, and Alaska working on mountain ungulate studies, we have chosen the Telonics TGW-4400 GEN 4 GPS transmitter for the GPS collar to be used for all bighorn sheep and mountain goats in the GYA (see Figure 2). Products produced from this vendor have the best track record for performing as advertised, with reliable releases of the timed drop off units and relatively high success rate for acquiring spatial locations. The collars will be programmed to collect location information at 6 hour intervals with a projected battery life of two years from the deployment date. If all spatial data are successfully acquired during the two years, the unit will record approximately 3,000 locations on its non-volatile flash memory (which is capable of storing a total of 100,000 locations). Based on results from other biologists studying mountain ungulates, we can expect approximately an 85% fix success rate.

The GPS unit will have a built-in VHF beacon with a duty cycle programmed for eight hours on (starting at 0900) and sixteen hours off, allowing for a projected battery life of two years plus about three months for the recovery of the collar. The collars will be equipped with the Telonics CR-2A breakaway unit, programmed to drop the collar off two years after deployment. The TGW-4400 weighs about 500 grams with a belt width of two inches. We will have the hardware and software to fully reprogram these GPS units in the event of deployment delays and for redeployment of collars that have been retrieved in the field after dropping off the animals.

We have chosen the Telonics MOD-401 MK9 system for the VHF collar to be deployed in concert with the GPS collars for all bighorn sheep and mountain goats in the GYA (see Figure 2). At the beginning of year three, after its period of being semi-dormant (regular transmission of a signal for short periods will maintain battery and function performance), the collar will transmit a duty cycle of ten hours on (starting at 0900) and fourteen hours off. The collar will continue to transmit for a projected battery life of six years from deployment and the collars will remain on the animal for life. A mortality sensor will be programmed to send a mortality signal (slightly higher pulse rate) when the collar has not moved for six hours. The collar weighs approximately 300 grams with a belt width of one inch. 133

Figure 2 – As part of a dual collaring technique, mountain goats (left) and bighorn sheep (right) will be instrumented with a GPS collar (Telonics TGW-4400 GEN 4), shown on the right of each pair, and a VHF collar (Telonics MOD-401 MK9), shown on the left of each pair.

Because the project spans across Yellowstone and Grand Teton National Parks, we have chosen to use colors for the belting that matches each species’ pelage to reduce the visual impact of instrumentation: white for mountain goats and brown for bighorn sheep (Figure 2).

Due to the significant amount of ongoing wildlife research in the GYA employing telemetry, we have identified a list of available frequencies that can be used for the built-in VHF beacon on the GPS collar and for the VHF collar without causing frequency interference or duplication. Ideally, the frequencies for the VHF beacon on the GPS collar and the VHF collar will be identical for the same individual. Some VHF and GPS collars used for initial deployments on bighorn sheep in the Gros Ventre, WY study area transmit on different frequencies as existing collars from previous ungulate studies were rebuilt.

For the initial deployment of collars, we will target adult females to gain demographic information from multiple herds across the GYA. We anticipate mountain goat captures will be more difficult with limited, if any, ability to correctly identify and select females for the first 134

collar deployment. We therefore expect to be opportunistic during captures, instrumenting both female and male mountain goats as they are caught and processed. The number of animals chosen for instrumentation will vary by study site and will depend on the priorities of administrative management agencies, availability of resources to purchase collars and instrument animals, the size of the herds, and the success of the captures.

MULTIPLE DEPLOYMENT STRATEGY

After retrieving and downloading the data from the dropped GPS collar, it will be reprogramed and redeployed on another animal, either in the same or different herd depending on the herd size and circumstances. Ideally, we would like to instrument additional female adults for the second collar deployment to enhance the demographic studies, however, if an adequate sample of females for a herd was initially collared, we may select adult males of that herd for instrumentation if desired by the administrating management agency. Collaring males will provide information on a segment of the population that has a higher dispersal probability, influencing patterns of genetic exchange and disease transmission. For the second deployment, we will use the same tactic as the first; two years of spatial data collection by the GPS collar, followed by four years of transmission from a new VHF collar. After the GPS collar drops two years later, it will then be redeployed a third time with a new VHF collar on a male or female either in the same or different herd depending on the size of the herd and/or priorities of the administrating management agency. Thus, we will efficiently combine GPS spatial studies and VHF demographic studies into a long-term telemetry project (see Figure 3).

Figure 3 - Multiple deployment strategy for dual GPS and VHF instrumentation of bighorn sheep and mountain goats in herds across the GYA.

Gaining these additional years of data collection by using this strategy will increase the number of animals sampled, allowing us to obtain considerable spatial and demographic detail, as well as to gain insight on potential annual variation in movement patterns due to variation in plant phenology and snowpack dynamics caused by annual variation in weather.

135

STUDY SITES

We have identified 10 study sites across the GYA (Figure 4) based on discussions with collaborators and in consideration of available resources. Each site has a unique combination of attributes, including administration agencies, regulatory limitations, access, physical setting, abundances of mountain goats and bighorn sheep, and seasonal occupancy (Table 2). The selected sites represent a variety of ecological settings that exist across the GYA, which will allow us to gain as broad an understanding of mountain goat and bighorn sheep ecology and interactions as possible. Eight sites are sympatric, being occupied by both bighorn sheep and mountain goats, but with varying durations and strengths of sympatry. Three sites are allopatric: two bighorn sheep sites (in the Gros Ventre, WY and Mt. Everts, YNP) and one mountain goat site (in the Palisades area, ID). The chosen sites not only reflect biologically significant areas, but also reflect the strong support received from local biologists and natural resource agencies to combine resources for improving efficiency and enhancing the field efforts and data collection, thus maximizing scientific productivity and benefiting local wildlife managers.

136

3 6 7 9 4

Figure 4 - Study sites where telemetry studies have been initiated (green), planned (orange), and potential (purple) across the GYA superimposed on locations of bighorn sheep (blue) and mountain goat (red) contained in the GYA Mountain Ungulate Project’s point database, showing all mountain ungulate observations from 1947 to 2010. Refer to Table 2 and 3 for study site descriptions.

At present, telemetry studies have been initiated at three sites. Fifteen bighorn sheep were collared by Wyoming Game and Fish Department in winters 2010, 2011, and 2012 within the Gros Ventre study site. Three mountain goat nannies were instrumented in August 2011 by Idaho Department of Fish and Game within the Palisades study site, with plans to instrument an additional nine mountain goat nannies during fall 2012 or winter 2013. In February 2012, five bighorn sheep ewes were collared by Montana Fish Wildlife and Parks within the Upper Yellowstone study site, with plans to instrument an additional eight ewes in winter 2013. In association with these collared bighorn sheep in the Upper Yellowstone, plans are in place to instrument a sample of mountain goat nannies during summer 2012. Additional plans and collaborations are currently underway to instrument animals in the remaining study areas (Table 3).

137

Table 2 – Attributes of 10 current and potential telemetry sites across the GYA for the Mountain Ungulate Project.

ry ry

No. No.

Land Land

State

Agency

Site No. Site

Strength

Seasonal Seasonal

Sympat Sympatry

Site Name Site

Restriction Occupancy

Ownership

Mtn. Goats Mtn.

Mtn. Range Mtn.

Elevation (m) Elevation

Management

Duration (yrs) Duration

Bighorn Sheep Bighorn

Gros year-round 1 Gros Ventre WY WGFD USFS Wilderness 2500 0 ? none 0 Ventre (sheep) IDFG year-round 2 Palisades Palisades ID-WY USFS none 1200 50 0 none 0 WGFD (goats) Upper Wilderness year-round 3 Gallatin MT MTFWP USFS 2000 160 180 strong 20 Yellowstone Study Area (both species) WGFD USFS Wilderness year-round 4 Pilot-Cache Absaroka WY 3300 20 50 strong 10-15 YNP YNP Park (both species) year-round Sunlight WGFD USFS Wilderness 5 Absaroka WY 3300 20 200 weak 0-10 (sheep) Basin YNP YNP Park unk (goats) year-round MTFWP 6 Line Creek Beartooths MT-WY USFS none 3000 70 20 strong 20-30 (goats) WGFD winter (sheep) Gallatin/ 7 Mt. Everts MT-WY YNP YNP Park 2000 0 ? none 0 year-round Absaroka Grand Teton GTNP GTNP year-round 8 Tetons WY Park 3300 10 125 weak <5 Nat'l Park WGFD USFS (both species) USFS 9 Taylor Hilgard Madison MT MTFWP Wilderness 1900 30 125 strong >20 year-round Private Spanish year-round 10 Madison MT MTFWP USFS Wilderness 2500 85 200 strong >50 Peaks (both species) 138

Table 3 - Status of 10 current and anticipated telemetry sites across the GYA for the Mountain Ungulate Project.

Site Capture & Instrumentation No. Sheep No. Goats Site Name No. Date Instrumented Instrumented Current Telemetry Sites 1 Gros Ventre January 2010, 2011, & 2012 15 0 2 Palisades August 2011, winter 2013 0 12 3 Upper Yellowstone February 2012, summer 2012 5 5-10 Planned Telemetry Sites 4 Pilot-Cache January 2013 5-10 5-10 5 Sunlight Basin January 2013 5-10 5-10 6 Line Creek Summer 2012 5-10 5-10 7 Mt. Everts January 2013 8-10 0 Potential Telemetry Sites 8 Grand Teton Nat'l Park uncertain 9 Taylor Hilgard uncertain 10 Spanish Peaks uncertain

Gros Ventre, Wyoming (site 1)

The Gros Ventre study area provides an excellent opportunity to collect information on a population of bighorn sheep in an allopatric environment where mountain goats are not present in the system. This area is located in the Gros Ventre mountain range east of Jackson, Wyoming and Grand Teton National Park. The fifteen ewes net-gunned by helicopter and instrumented by Wyoming Game and Fish were captured on their winter range in the Millers Butte area located in the National Elk Refuge and along the Gros Ventre River in the Gros Ventre National Forest during 2010-2012.

Palisades, Idaho (site 2)

The Palisades study area, located in the Palisades Mountains of southeast Idaho (the southern- most study area in the GYA), is also an important site due to the allopatry of mountain goats in the system. Mountain goats have been present in this area for 40 years, during which bighorn sheep have been entirely absent. The three mountain goats ground darted and instrumented by Idaho Fish and Game in August 2011 were captured up the Big Elk Creek drainage. Further plans to capture an additional nine mountain goats are planned for winter 2012-2013 in this same area.

139

Upper Yellowstone, Montana (site 3)

The Upper Yellowstone area (Figure 5) is important due to relatively long term (20 years) sympatry of mountain goats and bighorn sheep. Targeted animals include mountain goats and bighorn sheep that inhabit areas just north of Yellowstone National Park, centered around Ramshorn Peak and Sheep Mountain. Four bighorn sheep ewes were captured on their wintering grounds in February 2012, with plans to instrument an additional eight ewes. For mountain goats, the captures will occur during the summer of 2012 in the area around Ramshorn Peak (located within the Hyalite-Porcupine-Buffalo Horn Wilderness Study Area) and the area around Sheep Mountain (partially located within this Wilderness Study Area).

Figure 5 - The Tom Miner Basin, Montana site for initiating telemetry studies as part of the Mountain Ungulate Project.

140

Pilot-Cache, Wyoming (site 4)

The Pilot-Cache study area (Figure 6) is important due to the relatively recent (10-15 years) overlap of mountain goats and bighorn sheep. This area is located south of Cooke City, Montana and includes a portion of the northeast section of Yellowstone National Park. Targeted animals here include mountain goats that inhabit the Guitar Lake and Pilot Creek areas of the Shoshone National Forest, as well as The Thunderer and Abiathar Peak areas of Yellowstone National Park. Mountain goats on the Shoshone National Forest are rarely observed outside of the North Absaroka Wilderness.

Targeted bighorn sheep include individuals that summer in the upper reaches of Pilot Creek, Tough Creek, and North Crandall Creek drainages within the North Absaroka Wilderness. These animals winter near Pilot Peak, and on the Pilot Creek/One Mile Creek Divide as well as the One Mile Creek/Tough Creek Divide. With the exception of bighorn sheep using the Tough Creek winter range, some of these sheep can occasionally be found outside the wilderness depending upon winter conditions. Bighorn sheep that reside year-round in the Mount Norris, The Thunderer, Abiathar Peak, and Amphitheater Mountain areas would also be targeted for capture.

Figure 6 - The Pilot-Cache site for initiating telemetry studies as part of the Mountain Ungulate Project. Figure courtesy Wyoming Game and Fish.

141

Sunlight Basin, Wyoming (site 5)

The Sunlight Basin area (Figure 7) is important due to the very recent (5 years) overlap of mountain goats and bighorn sheep. Of equal importance is the fact that this area represents a leading edge in the current expansion of mountain goats in this area. This area is located Targeted animals here include mountain goats that inhabit the drainages of Upper Sunlight Creek, Hoodoo Creek, Temple Creek, One Hunt Creek, Sweetwater Creek, and the North Fork of the Shoshone River. Mountain goats residing in the Saddle Mountain and/or Castor and Pollux Peak areas within Yellowstone National Park would also be targeted for capture. Mountain goats on the Shoshone National Forest are rarely observed outside of the North Absaroka Wilderness.

Targeted bighorn sheep include individuals that summer in these same areas (upper reaches of Sunlight Creek, Hoodoo Creek, Temple Creek, One Hunt Creek, Sweetwater Creek, and the North Fork of the Shoshone River within the North Absaroka Wilderness). Many bighorn sheep winter at high elevations on Black Mountain and Dike Mountain, essentially remaining on their summer ranges. Some bighorn sheep winter on ridges along the Temple and Hoodoo Creek drainages, still within the North Absaroka Wilderness. These animals are rarely found outside of wilderness. Bighorn sheep that summer south of Sunlight Creek descend to winter ranges along the North Fork of the Shoshone River, where they reside either in or out of the North Absaroka Wilderness depending upon winter conditions.

142

Figure 7 - The Sunlight Basin site for initiating telemetry studies as part of the Mountain Ungulate Project. Figure provided by Wyoming Game and Fish.

Line Creek, Montana (site 6)

The Line Creek area (Figure 8) is important due to the relatively long-term (20-30 years) overlap of mountain goats and bighorn sheep, and an apparent shift from a predominantly bighorn sheep winter range to one dominated by mountain goats. Targeted animals here include mountain goats and bighorn sheep that inhabit the following areas of the Shoshone and Custer National Forest: Line Creek peninsula, Wyoming Creek, and the tributaries of Rock Creek drainage above Hellroaring Creek. Mountain goats reside in this area year-round, while bighorn sheep are essentially found here only in winter. 143

Figure 8 - The Line Creek site for initiating telemetry studies as part of the Mountain Ungulate Project. Figure provided by Wyoming Game and Fish.

Mount Everts, Wyoming (site 7)

The Mount Everts study area, located east of Mammoth Hot Springs in Yellowstone National Park and southeast of Gardiner, Montana, also contains a population of allopatric bighorn sheep. About ten of these bighorn sheep will be targeted for instrumentation, likely during winter 2012- 2013. The herd that occupies this area can often be seen from the North Entrance Road in the Park and we predict that ground-darting will be possible because of their tolerance to human presence.

Grand Teton National Park, Wyoming (site 8)

Grand Teton National Park provides an excellent opportunity to study the two mountain ungulates in an area with very recent expansion of mountain goats into bighorn sheep range. This sympatry has only existed for the past five years or less.

144

Taylor Hilgard, Montana (site 9)

The Taylor Hilgards are a study area with long-term sympatry (20 years or more) between bighorn sheep and mountain goats. Bighorns have survived two all-age die-off events and have been augmented from Thompson Falls, Anaconda, and Wild Horse Island herds. Counts in the late 2000’s suggested recovery of this herd. The two winter ranges of this herd are well-defined: the larger (140+ individuals) exists near Quake Lake, and the smaller (~40) exists at Moose and Squaw Creek. Summer ranges of these bighorn are largely unknown, although bighorn have been seen on both sides of the Taylor-Hilgard mountain divide at very high elevation and near the alpine lakes (Cradle Lakes, Finger Lakes, Coney Lakes, Avalanche Lake, Blue Danube, and Lake Eglise). Mountain goats are frequently and abundantly seen throughout the suspected summer range of the bighorn sheep. In winter, the two are often viewed together, especially at the Moose and Squaw Creek bighorn wintering site, and less so at Quake Lake.

Spanish Peaks, Montana (site 10)

The Spanish Peaks are a study area with long-term sympatry (50 years or more) between native bighorn sheep and mountain goats. Bighorn and mountain goat populations appear healthy, overall. Curiously, there are areas where bighorn and mountain goats overlap to higher degrees, and areas where they overlap less, if at all. Mountain goats and bighorn sheep are readily observable together on their summer ranges from Mirror Lake, Thompson and Chilled Lakes, and Gallatin Peak. However, bighorn also summer on the southeastern side of the Spanish Peaks (Dudley Lake, Deer and Moon Lake, Jumbo Mountain and Wilson Peak) where mountain goats are more rarely observed. Mountain goats are also numerous on the northwestern side of the Spanish Peaks (Jerome Rock Lakes, Chiquita Lakes, and Spanish Lakes) where bighorns are rarely if ever observed. It is possible that these more eastern zones are more closely aligned with original transplant locations, explaining the mountain goat affinity to the west side.

On their winter ranges in this area, bighorn and mountain goat seem to seldom overlap. The bighorn winter range is tightly defined between Big Sky and Durnam Meadows (north of Asbestos Creek), whereas mountain goats either winter at high elevation or low along Jack Creek road in the Madison Valley.

CAPTURE TECHNIQUES

Extensive animal capture operations are necessary to provide the breadth of information desired for bighorn sheep and mountain goats from across the GYA. The Mountain Ungulate Project has been seeking the advice of biologists in Alaska, Canada, and the western United States that have expertise in trapping and chemical immobilization of mountain goats and bighorn sheep and feel there are several techniques that can be employed with a reasonable expectation of success. We have selected four capture techniques: drop-netting, Clover trapping, helicopter net-gunning, and ground darting. The method chosen will be dependent on the administrating land and wildlife 145

agencies, available resources, animals being captured, and access and location of the capture site. Bighorn sheep are readily baited on winter range and can be captured in clover traps and drop nets on low-elevation ranges with reasonable wheeled-vehicle access. Mountain goats do not respond well to forage baits, but can be successfully attracted to salt from spring through late summer, with biologists successfully using both drop nets and clover traps. There are a number of field sites in the GYA where animals are not extremely wary that could provide opportunities for stalking and chemical immobilization via dart rifles. Regardless of the method chosen, chemical immobilizing drugs will be used to reduce the amount of stress to the animal, decrease the risk of injury or death, and allow easier fitting of the telemetry collars and collection of biological samples (see Chemical Immobilization). Captures will primarily be scheduled for mid-summer to early fall for Clover trapping and ground darting of mountain goats and for winter seasons for drop netting, helicopter net-gunning, and ground darting of both species.

We have obtained the required authorization and permits from the necessary agencies for capture and instrumentation in several areas, including the Upper Yellowstone area (Gallatin National Forest) as well as areas within Yellowstone National Park and the North Absaroka and Absaroka-Beartooth Wildernesses. For the captures in the Palisades, ID study site, authorization was obtained internally by Idaho Fish and Game. For the captures performed in the Gros Ventre, WY study site, permitting was obtained internally by Wyoming Game and Fish. Permits for capture in other study sites are pending. Our capture and handling methods have also been approved by the Institutional Animal Care and Use Committee (through Montana State University).

Figure 9 - Capture of a bighorn sheep ewe by Montana Fish Wildlife and Parks, February 2012, as a part of the collaboration with the GYA Mountain Ungulate Project. 146

The following outlines the four methods that will be used to capture bighorn sheep and mountain goats in the greater Yellowstone area:

1) Drop nets In anticipation of animal capture operations in the next 6-10 months, a custom drop net system has been constructed for the trapping of multiple animals in one capture event (Figure 11). The drop net is a heavy, 80 feet x 80 feet net, with a 5 inch mesh size, that will be suspended about 8 feet above the bait site using a center pole and eight perimeter poles. The net is attached to each pole via an electronic (DC) solenoid-release mechanism. Each solenoid mechanism is connected by an electrical cable to a waterproof and portable case containing a 24-volt battery system and a custom-built wireless receiver that will allow an operator to drop the net from a remote monitoring site up to a half mile from the bait site. When a wireless transmission is received, the electrical circuit is completed, which causes each of the solenoid mechanisms to release simultaneously, dropping the net to the ground.

Figure 10 - Drop net system developed for the capture of multiple individuals of bighorn sheep.

147

Drop nets are typically used for bighorn sheep on their winter range, but may be used for mountain goats depending on the terrain and situation. Areas of frequent bighorn sheep or mountain goat use will be selected for capture, and potential capture sites will be pre-baited and monitored either directly or with the aid of camera traps until adequate numbers of animals are routinely visiting the salt or hay bait. The net will then be suspended over the bait site for 2-6 days to allow the animals to become accustomed to feeding under the net. To capture the animals, a technician hidden in a distant blind will wirelessly activate the solenoid-release mechanisms causing the net to fall on the animals. Typically 4-12 animals will be captured with each net drop. The weight of the net will prevent any significant movement or escape. Animal(s) will be immediately hobbled and blindfolded by several technicians to completely restrain the animal to prevent escape or injury to the animal or technicians. Immobilizing drugs will be administered to the captured animals (see Chemical Immobilization). Because multiple animals may be caught, this capture method requires about 12-15 people to make sure all animals are immediately restrained and processed efficiently. Animals will be captured, fitted with the dual GPS and VHS collars, weighed, and biological samples collected (see Health and Disease Sampling). Animals will be typically processed and released at the trap site within one hour of capture. Due to the large number of animals that can be caught, not all will be instrumented; however, the capture events will provide an excellent opportunity to collect samples for health and genetic screening and uncollared animals may be individually marked with number ear tags or potentially passive integrated tags (PIT).

2) Clover traps Initially developed for deer capture (Clover 1956), Clover traps have undergone many manipulations for different applications, but are generally netting surrounding a metal box frame (i.e., a netted cage), about 4 feet tall, 3 feet wide, and 6 feet long. These traps have been used successfully for several ungulate species, including mountain goats (Garrott and White 1982, Millspaugh et al. 1994, Cote et al. 1998, Haviernick et al. 1998, Hiller et al. 2010). A custom Clover trap is currently being developed and tested so that it can be broken down into small enough components to allow 2-3 traps to be transported into remote locations using a pack horse. We expect to produce a prototype this fall and possibly test the trap on deer during the winter. If successful, we would build an appropriate number of traps for use on remote mountain goat and bighorn sheep ranges by summer 2012. Areas of frequent bighorn sheep or mountain goat use will be selected for capture, and potential capture sites will be pre-baited and monitored either directly or with the aid of camera traps until the animals are routinely visiting the salt or hay bait. The trap will then be placed at the bait site for 2-6 days to allow the animals to become accustomed to feeding inside the trap. When the trap is set for capture, a trip-wire in the back of the cage allows the animal to get fully inside the trap before a netted door is released, capturing the animal inside. A VHF radio transmitter designed to begin transmitting when the door is released will be attached to each trap. The signal will immediately inform a capture crew of 2-4 technicians staged within 15 minutes of the trap. The captured animal will be chemically 148

immobilized, then blindfolded, fitted with the dual GPS and VHF collars, weighed, and biological samples collected. Animals will be typically processed and released at the trap site within one and an half hours of capture.

3) Helicopter net-gunning Net-gunning from a helicopter has been used successfully for both bighorn sheep and mountain goats and is a highly effective and safe capture technique for individual animals (Andryk et al. 1983, Krock et al. 1987, Jessup 1988, Poole and Heard 2003). Helicopter net-gunning will be conducted in cooperation with commercial contractors with extensive experience working with regional natural resource agencies, and will be in conjunction with interested agencies to maximize capture efficiency and cost effectiveness across multiple study sites. For some of these captures, the Montana State University research team will contract wildlife capture companies, while for some study sites the MSU team will provide support for the capture effort, but the management agencies will directly contract with the wildlife capture companies and underwrite the cost of the operation.

Figure 11 - Helicopter net-gunning capture by the Wyoming Game and Fish in the Gros Ventre study area, January 2011, as a part of the collaboration with the GYA Mountain Ungulate Project. Photo courtesy Mark Gocke.

149

Helicopter net-gunning allows single individuals from a herd to be selected for capture. When fired (using blank rounds), a net with weighted ends envelopes the animal and prevents further movement and escape. Technicians from the helicopter will be deployed as quickly as possible to restrain, hobble, and blindfold the animal. Captured animals will be chemically immobilized, and then fitted with the dual GPS and VHF collars, weighed, biological samples collected, and released at the capture site within about one hour. Due to the high risk of helicopter flights in mountainous terrain, flights will be limited to favorable weather days (i.e. high flight ceiling and moderate wind speed). Captures will not be attempted in areas where steep terrain puts the helicopter crew or animals in danger.

4) Ground darting Ground darting may also be used to capture and immobilize bighorn sheep and mountain goats. This technique will only be useful in areas where animals are accustomed to human presence and can be approached to a sufficiently close range for darting (20-60 meters). Capture will occur in areas that do not pose a risk to the crew or animals during induction of the drugs (i.e. deep rivers or streams or steep terrain). Once immobilized, the animal will be fitted with the dual GPS and VHF collars, weighed, biological samples collected, and released at the capture site within about one hour.

CHEMICAL IMMOBILIZATION

As part of all the capture protocols, the handling and processing of the animals will be facilitated with the use of chemical immobilizing drugs. Although animals can usually be released quicker without drugs, anesthetizing them will reduce the amount of stress the animal experiences and the chance of injury to the animal and the capture crew. It will also expedite the fitting of radio collars and collection of biological samples.

Through discussions with biologists with extensive experience using chemical immobilization, we have selected four potential drug combinations, outlined in Table 4. The drugs chosen will be dependent on the supervising veterinarian and their experience with the various drug options. We suspect that BAM will be the preferred drug as other wildlife biologists have found it extremely effective and have experienced no drug-related mortalities in either bighorn sheep or mountain goats.

150

Table 4 - Drug options for chemical immobilization of adult bighorn sheep and mountain goats.

Drug Per Adult Animal Species Mix/Amount Reversal Cocktail Dosage 15.0 mg/ml bighorn Butorphanol, sheep 10.0 mg/ml BAM 1.8-2.2 ml 2cc Tolazine, 5cc Antisedan & Azaperone, 6.0 mtn. goat mg/ml Medetomidine

2.4-2.7 mg Carfentanil mtn. goat ~0.02 mg/kg 100 mg Naltrexone/mg Carfentanil (females), 2.7-3.0 mg only Carfentanil Carfentanil Carfentanil (males)

bighorn 0.045 mg/kg 100 mg Naltrexone/mg Carfentanil + 3.5-4.5 mg Carfentanil, sheep Carfentanil, 0.2 Carfentanil, 1.0-3.0 mg/kg Xylazine 15-20 mg Xylazine only mg/kg Xylazine Tolazoline bighorn 0.05 mg/kg 50 mg Naltrexone/mg Etorphine 4.5-5 mg Etorphine, 20 sheep Etorphine, 0.2 Etorphine, 1.0-3.0 mg/kg (M9) mg Xylazine only mg/kg Xylazine Tolazine

Whenever possible, a veterinarian will be present at the capture events to deal with any medical emergencies.

Drugs will be administered by several methods, depending on the capture technique. For drop- netting, net-gunning, and possibly Clover trapping, we intend to administer the drugs using an intranasal syringe or intramuscularly with a needled syringe. Intranasal administration works with similar effectiveness as intravenous administration, and would reduce the risk of accidental needle jabs to the capture crew and tissue damage associated with injections. This methodology is routinely being used for net-restrained large mammals in Alaska and northwestern Canada where the biologist have refined the delivery device and its use. A syringe-mounted jab stick will be used for animals in Clover traps. For ground darting, a dart will be fired into the rear hind quarter (intramuscularly) of the animal using Pneudart cartridge-fired disposal darts. Once the drug has been administered, the capture crew will remain at a distance while the drug is taking effect. When the animal has been fully anesthetized, the technicians will blindfold the animal, remaining quiet to reduce sensory stimuli, and continue to process the animal. After the collars have been attached and samples collected, the antagonist drug will be administered intramuscularly or intravenously, and recovery should take about 5-20 minutes.

151

COLLARING TECHNIQUE

Prior to capture and instrumentation, collars will be programmed and properly prepared for deployment. For demographic investigations, the visual relocation and positive identification of individual collared animals is a necessity. For this reason, the collars will be patterned with horizontal stripes on the sides of the collars, arranged in unique combinations for each individual (Figure 12). We have developed eleven combinations, readily visible in the field, which can be deployed in herds or on individuals that are located nearby each other. In cases where several instrumented animals are located together and they cannot be identified by signal alone, this amount of unique patterning minimizes confusion of markings and should allow the observer to identify each individual. This set of combinations may be used for other herds if they occupy sites separated by considerable geography. The stripes for marking will be colored to contrast with the belting color: black on the white mountain goat collars and white on the brown bighorn sheep collars. Duct tape will be the product used for the striped markings. Collaborating agencies may prefer different methods of marking, but this method provides a standard for the

entire GYA that can be easily adopted and implemented.

Bighorn sheep Bighorn

Mountain goat Mountain

Figure 12 - Marking patterns for bighorn sheep and mountain goats collars to aid in the identification of individuals for survival and reproduction surveys.

152

Because the animals will be wearing both collars for two years and one collar for the remainder of their lives, proper fitting of the collars is critical in reducing the risk of unnecessary injury, stress, and hair loss. After the animal has been captured and adequately anesthetized, the collars will be fit snuggly onto the animal’s neck with the VHF collar placed anterior to the GPS collar. This will permit the VHF to remain in a snug, well-fitted position after the GPS has fallen off. A “snug” fit allows a flattened hand to be slipped between the collar and the animal’s neck without room for significant hand or finger movement but loose enough that there is no pressure on the hand. This will ensure that the collars do not slide up and down the neck, reducing hair loss while also preventing abrasion and collision between the two collars. This fit is also critical for males when head clashing during the rut to prevent injury to the mandibles. If younger rams are collared, room for growth will be considered, with the possibility of incorporating degradable padding to keep the collar snug over time.

Figure 13 - Bighorn sheep ewe being instrumented by Wyoming Game and Fish in the Gros Ventre study area, January 2011, as a part of the collaboration with the GYA Mountain Ungulate Project. Photo courtesy Mark Gocke.

153

HEALTH AND DISEASE SAMPLING

Biological Samples and Morphological Measures

Standard biological specimens will be collected and morphological measures will be recorded during the handling process. The samples collected from each animal will include: blood (~20 mL), skin biopsy, hair, fecal pellets, nasal swab, a tooth, and horn measurements, photos and samples (Table 5). If logistically feasible, each animal will also be weighed.

Table 5 - Samples to be collected during capture and handling operations, their purpose, and individuals/agencies performing the analysis in cooperation with the Mountain Ungulate Project.

Samples Collected Purpose Analysis by

Jim Berardenelli (MSU), Stable isotopes, pregnancy, disease, Blood serum Seth Newsome (UofWY) and metabolites State agencies Jim Berardenelli (MSU), Red blood cells Stable isotopes, genetics, Seth Newsome (UofWY) Skin biopsy Genetics Hair Stable isotopes and genetics Seth Newsome (UofWY) Fecal pellets Parasites Nasal swab Bacteria and parasites State agencies Tooth Aging Matson Lab, Missoula, MT Horn shaving Stable isotopes Seth Newsome (UofWY)

The blood will be drawn from the jugular vein using a 30cc syringe with an 18 G 1.5” needle and immediately transferred into non-heparinized serum-separating vacutainer tubes. The blood samples will be allowed to clot and the serum and red blood cells will be separated using a centrifuge after 2-3 hours from collection. The serum will then be decanted into a storage tube, and both elements will then be stored frozen at -20 degrees Celsius until analyzed. The blood samples will be analyzed for stable isotopes, genetics, pregnancy, disease serology, and a panel of metabolic function assays.

Skin biopsies for genetic analysis will be taken from the inside of one ear using a sterile 4-6mm biopsy punch. For genetic and stable isotope analysis, a small tuft of hair (about 5-10 hairs with roots intact) will be collected with latex gloves and stored in a whirl-pak. The fecal samples will be collected from the rectum of each animal using latex gloves and placed in a whirl-pak for the purpose of parasite screening. Nasal swabs for bacterial culture will be collected with latex gloves and also placed in a whirl-pak or appropriate culture medium container. Animals will be aged based on tooth irruption patterns thru age 3 for mountain goats and 4 for bighorn sheep with

154

tooth cementum analyses used to age older animals. A single vestigial premolar tooth from mountain goats or an incisorform canine tooth from bighorn sheep will be extracted using a veterinary dental elevator and forceps and stored in a whirl-pak, with tooth preparation and aging performed at the Matson Laboratory in East Missoula, Montana. All storage containers will be properly labeled with date, species, and collar frequencies.

Figure 141 - Blood sample being taken from a bighorn sheep ewe captured by Wyoming Game and Fish in the Gros Ventre study area, January 2011, as a part of the collaboration with the GYA Mountain Ungulate Project. Photo courtesy Mark Gocke.

Analysis of Biological Samples

Samples will be sent to Dr. James Berardenelli (Department of Animal and Range Sciences, Montana State University). Blood will be assayed for the following metabolites and metabolic and gonadal hormones:

Metabolites: 1) Glucose – the main energy source from diet, glycogenolysis, and gluconeogenesis

155

2) Non-esterified Fatty Acids (NEFA) – an indicator/marker for mobilization of fat reserves 3) -hydroxybutyrate – in ruminants, this is synthesized by anaerobic fermentative bacteria in the rumen and can be used as an energy source via gluconeogenesis and fat mobilization

Metabolic hormones: 1) Insulin – an indicator of energy utilization 2) Cortisol – an indicator of overall stress 3) Thyroxine (T4) and tri-iodothyronine (T3) – an indicator of basal metabolic rate

Gonadal steroids: 1) Progesterone – an indicator of ovulatory/luteal activity and pregnancy status in females 2) Testosterone – an indicator of seasonal breeding activity in males

Urinary metabolites: 1) Allantoin/creatinine ratios – indicators of nutritional status

The stable isotopes derived from the collected hair, horn shavings, and blood samples (both serum and red blood cells) will be used to better understand regional bighorn sheep and mountain goat ecology and competition. Stable isotopes obtained from plant tissues can be compared to those found in the tissues of the mountain ungulates, allowing us to evaluate the diet and dietary overlap between bighorn sheep and mountain goats. These samples will be sent to Dr. Seth Newsome for analysis (Department of Zoology and Physiology, University of Wyoming).

Teeth samples will be analyzed at the Matson Lab in Missoula, MT. The remaining samples will be assayed by the appropriate wildlife management agencies that routinely analyze samples from animal captures and/or assayed or archived for future work by the MSU researchers.

FIELD STUDIES FOR POPULATION DYNAMICS

With both the built-in VHF beacon on the GPS collar and the VHF collar programmed to transmit signals in a tandem fashion, each instrumented animal can be monitored for survival, reproduction, and kid/lamb survival for a total of six years. The unique markings on each collar (see Collaring Technique) will also greatly help the positive identification of animals. Depending on accessibility and frequency of flights conducted in the study areas for other wildlife-related activities, monitoring will be done either from the ground or aircraft.

Instrumented adults will be located for survival, and field crews will hike in on all mortality signals as soon as practical to retrieve the instruments, collect whatever samples we can (teeth, skulls, etc.) and evaluate evidence for the cause of death. Since we won’t arrive at most of the mortalities immediately after death, we expect that cause of death will not be determined with

156

much certainty for most animals. However, we anticipate that some deaths due to accidents (falls, avalanches), starvation (concentrated in late winter/early spring), hunter mortalities, and perhaps some predation will be able to be determined with some certainty.

During the spring lambing/kidding season, all reproductive-age instrumented females will be located to determine if they produced a young that year. The females with young will be observed again in the fall to determine if the young survived the first 6 months of life. Monitoring events will occur for a minimum of one week in the spring and one week in the fall. Further monitoring will be performed opportunistically and as resources from management agencies allow, however, obtaining several observations throughout the season can be critically important in confirming whether single monitoring observations were accurate. For example, a female may have produced a kid or lamb in the spring, but if it was not detected in the first monitoring event, we would incorrectly conclude that the female did not reproduce that year. Similarly, if a kid or a lamb monitored for survival in the fall was present but not seen, we may incorrectly conclude that it was a mortality. Monitoring for winter survival of young-of-the-year may also occur if resources and interest from management agencies allow.

If adequate numbers of animals can be instrumented, these simple surveys, in conjunction with the routine annual surveys conducted by biologists, will provide a wealth of data on adult and juvenile survival and mortality, reproduction, and recruitment, as well as coarse temporal scale information on seasonal movement patterns, fidelity, and dispersal following the scheduled drop off of the GPS collars after two years.

DATABASE OVERVIEW

A considerable amount of information will be obtained from these telemetry studies, primarily spatio-temporal data from the GPS devices, demographic information from monitoring observations of instrumented individuals, and results from biological samples collected during the capture and handling process. We have chosen to use Microsoft Access 2010 to organize and manage this information as it can store different types of data and is able to perform custom queries to answer important biological questions. Microsoft Access is commonly used for biological data and provides an interface allowing the data to be entered and shared easily. Microsoft Access 2010 is the most recent version, and permits users to save the database in earlier versions of the program if necessary.

The data will be stored in two separate databases (i.e., two separate Access files) for the purpose of keeping the general types of data manageable and clean. The information obtained from the capture operations and field monitoring of the collared animals will be stored in a Capture and Monitoring database, and the data obtained from the downloaded collars after retrieval will be stored in a Spatiotemporal GPS database.

157

The sharing of data obtained from other capture and monitoring activities for inclusion in these two databases maintained by Montana State University is at the discretion and willingness of the collaborating agencies performing those captures.

CAPTURE AND MONITORING DATABASE

The following five tables were created to house the information in the capture and monitoring database (Figure 15):

1) The Animal Information table will store all the information pertaining to the capture and handling process (Figure 16). This includes fields for the date of capture, capture UTM locations, capture agency, sex, species, herd, an animal identification number, radio frequencies of both the VHF collar and the built-in VHF beacon on the GPS collar, the collars’ duty cycles, the samples collected, collar markings or

Figure 15- Screenshot of the tables and fields contained in the numbers, and any other individual Capture and Monitoring Database for the GYA Mountain identification information. The Ungulate Project. status of the animal (dead, alive, alive-off air, or unknown), collar (transmitting, dead, recovered, or unknown), and frequency will also be maintained and updated here as well. Each instrumented animal will be contained in one column, and any changes to the status of the animal, collar, or frequency (i.e., drift in frequency) will be updated in that same column. By keeping the status of each instrumented animal updated, preparation for field monitoring activities is made much quicker and more efficient. If an individual is recollared, a new column will be created with the new information, however the animal identification number will be retained.

158

Figure 16 - Screenshot of the Animal Information table in the Capture and Monitoring Database for the GYA Mountain Ungulate Project. The most recently obtained information from capture operations performed by the Wyoming Game and Fish are displayed.

2) The Collar Information table will contain information for each individual collar. This means two rows will be created for each dual-collared animal; one row for the GPS collar information and another for the VHF collar information. All the specifications of the collar will be maintained here, including fields for the frequency, serial number, duty cycle, drop- off date, status (deployed or retrieved), responsible agency, and maintenance log.

3) The Monitoring Data table stores the information obtained from field monitoring activities (Figure 17). Each observation event of a collared animal will be contained in one row. Date, survey type (ground or aerial), animal identification number, study area, location, status (dead or alive), presence and number of offspring, and group size and composition (females, males, young males, yearlings, and young-of-year) will be stored in this table.

Figure 17 - Screenshot of the Monitoring Data table in the Capture and Monitoring Database for the GYA Mountain Ungulate Project. The most recently obtained information from observations performed in collaboration with Wyoming Game and Fish are displayed.

4) A separate Reproduction table will house the presence, number, and fate of any offspring produced by each individual for each year. This table is essentially the final assessment of each individual’s fecundity and their offspring’s survival for each year, with fields for the year, number of offspring, offspring sex, and offspring fate.

5) The Health table will contain the results of the analyses performed on the biological samples collected during the capture and handling operations (see Health and Disease Sampling).

SPATIOTEMPORAL GPS DATABASE

The spatial and temporal data collected and downloaded from the retrieved GPS collars will be processed or “cleaned” (any unnecessary data collected by the GPS removed) and stored in the spatiotemporal GPS database. Each downloaded collar will be contained in its own table (Figure 159

18). Fields for this table include an animal identification number (which relates to the animal number in the Capture and Monitoring Database), the VHF beacon frequency, fix number (consecutive count of each fix), failed fixes, date, time, latitude, longitude, altitude (m), and PDOP, HDOP, VDOP, and TDOP values (for discernment of accuracy). For future analysis, all the individual tables containing each collar’s data will be merged into a single aggregate table.

Figure 182 - Screenshot of a table containing data from a downloaded GPS collar in the Spatiotemporal GPS Database for the GYA Mountain Ungulate Project. The data was retrieved from a bighorn sheep ewe in the Gros Ventre study area by the Wyoming Game and Fish.

ACKNOWLEDGEMENTS

We would like to sincerely thank the following agency personnel who provided insight and support for developing and implementing the telemetry studies: Hollie Miyasaki (IDFG); Julie Cunningham (MFWP), Tom Lemke (MFWP), Karen Loveless (MFWP), Justin Paugh (MFWP), Shawn Stewart (MFWP), Doug Brimeyer (WGFD), Kevin Hurley (WGFD), Doug McWhirter (WGFD), P.J. White (YNP), Sarah Dewey (GTNP), and Andy Pils (USFS). Successful capture operations thanks to Hollie Miyasaki, Karen Loveless, Doug Brimeyer, Doug McWhirter, Neil Anderson (MTFWP), and Jennifer Ramsey (MTFWP). Financial support for this project was provided by the National Park Service, Canon USA Inc. through the Yellowstone Park Foundation, Wyoming Game and Fish Department, Idaho Department of Fish and Game, U.S. Forest Service, Montana Wild Sheep Foundation, and Wyoming Wild Sheep Foundation.

160

LITERATURE CITED

Andryk, T.A., L.R. Irby, D.L. Hook, J.J. Mccarthy, and G. Olson. 1983. Comparison of mountain sheep capture techniques: helicopter darting versus net-gunning. Wildlife Society Bulletin 11:184-187.

Clover, M.R. 1956. Single-gate deer trap. California Fish and Game 42: 199.

Cote, S.D., M. Festa-Bianchet, and F. Fournier. 1998. Life-history effects of chemical immobilization and radiocollars on mountain goats. Journal of Wildlife Management 62: 745-752.

Garrott, R.A. and G.C. White. 1982. Age and sex selectivity in trapping mule deer. The Journal of Wildlife Management 46: 1083-1086.

Haviernick, M., Cote, S.D., and M. Festa-Bianchet. 1998. Immobilization of mountain goats with Xylazine and reversal with Idazoxan. Journal of Wildlife Diseases 34: 342-347.

Hiller, T.L., J.P. Burroughs, H. Campa III, M.K. Cosgrove, B.A. Rudolph, and A.J. Tyre. Sex- age selectivity and correlates of capture for winter-trapped white-tailed deer. 2010. Journal of Wildlife Management 74: 564-572.

Jessup, D.A., R.K. Clark, R.A. Weaver, and M.D. Kock. 1988. The safety and cost-effectiveness of net-gun capture of desert bighorn sheep (Ovis canadensis nelsoni). The Journal of Zoo Animal Medicine 19: 208-213.

Kock, M.D., D.A. Jessup, R.K. Clark, C.E. Franti, and R.A. Weaver. 1987. Capture methods in five subspecies of free-ranging bighorn sheep: an evaluation of drop-net, drive-net, chemical immobilization and the net-gun. Journal of Wildlife Diseases 23:634-640.

Millspaugh, J.J., G.C. Brundige, and J.A. Jenks. 1994. Summer elk trapping in South Dakota. Prairie Naturalist 26:125-129.

Poole, K.G., and D.C. Heard. 2003. Seasonal habitat use and movements of mountain goats, Oreamnos americanus, in east-central British Columbia. Canadian Field-Naturalist 117(4): 565-576.

161