Eastern Slopes Grizzly Bear Project

A CONSERVATION BIOLOGY APPROACH TO MANAGEMENT OF GRIZZLY BEARSIN BANFF NATIONAL PARK, ALBERTA

By Michael L. Gibeau

I examined movement patterns of adult female grizzly bears (Ursus arctos) in the Bow River Watershed, Alberta. Intensive movement data showed that habituated adult female bears did not take advantage of higher quality habitats in the same manner as wary bears. The combination of habituated bears using lower quality habitats and demonstrating higher movement rates suggests less energy available for growth and reproduction. Bears within an area of restricted human access used higher quality habitat and traveled less than bears in unregulated areas. I document the permeability of several highways in a landscape where human presence is widespread. One highway with 24 hour, year-round high traffic volumes served as a total barrier for adult female movement and a filtered barrier for males. Traffic volume appeared to be a key variable in highway permeability. Significant potential currently exists for permanent habitat and population fragmentation to occur along the Trans Canada Highway. I document the degree and magnitude of grizzly bear responses as a function of multiple interacting variables based on observed distances to roads, trails and development features. Bears were found closer to trails during the human inactive period when within high quality habitat and further from trails when distant from high quality habitat. Female bears remained further than males from paved roads regardless of the habitat quality or time of day. My data indicated an inverse relationship between the sexes in response to vehicles and traffic noise compared to the response to human settlement and encountering people. I developed a predictive GIS-based model of adult female grizzly bear security areas in the Central Canadian Rocky Mountains. Forty eight percent of the land surface area of the Banff, Yoho, and Kootenay National Parks were unsuitable for grizzly bears, primarily because of rock and ice. This is unfortunate because it is assumed that these national parks form productive core refugia for grizzly bears. Management of access and development are key to grizzly bear persistence in the region. An adaptive management approach, bringing in new knowledge of grizzly bear response to human activity, will be crucial, to support population connectivity and habitat security.

A very successful research project would not have been possible without the dedication of field biologists B. Benn, M. Jalkotzy, C. Mamo, C. Morgan, C. Mueller, J. Paczkowski, H. Robinson, J. Saher, S. Stevens, S. Stotyn, J. Theberge, M. Urquhart, and J. Wittingham. Their efforts were augmented through the volunteer support of K. Barkman, C. Campbell, P. Hoffer, L. Homstol, and M. Lacroix. Assistance in coordination of field staff was provided by A. Dibb, S. Donelon, T. Hurd and C. White. Trapping was conducted by R. Leblanc, C. Mamo, R. Riddell, and I. Ross. Veterinary care was provided by Dr. T. Shury. Several Alberta Fish and Wildlife Officers, Banff National Park Wardens and Peter Lougheed Park Rangers all provided invaluable safety backup and field assistance during trapping. The Banff Park Warden Service and Kananaskis Country Park Rangers provided logistical support through all stages of monitoring. Exemplary flying skills were provided by Alpine Helicopters of Canmore and fixed wing pilot M. Dupuis of Wildlife Observation Air Services. S. Jevons, J. Slemko, and J. Wierzchowski provided GIS and computer technical support. C. Kelly provided much needed editing.

The Eastern Slopes Grizzly Bear Steering Committee helped implement and guided this research. All steering committee participants contribute either money, time or both toward the objectives. Through the Steering Committee, governments, industry, business and conservation groups work together to support this and other research components (www.canadianrockies.net/grizzly). Finally, I thank my supervisory committee for the support, encouragement and critical review of this project.

ABSTRACT

ACKNOWLEDGMENTS

TABLE OF CONTENTS

BACKGROUND AND CONTEXT FOR THE RESEARCH

ABSTRACT

INTRODUCTION

STUDY AREA

METHODS

Landscape Characteristics

Analysis

RESULTS

DISCUSSION

MANAGEMENT IMPLICATIONS

CHAPTER TWO: EFFECTS OF HIGHWAYS ON GRIZZLY BEAR MOVEMENT IN THE BOW RIVER WATERSHED, ALBERTA.

ABSTRACT

INTRODUCTION

STUDY AREA

METHODS

Highway Crossings

Crossing Location Attributes

Analysis

RESULTS

Highway Crossings

Crossing Location Attributes

DISCUSSION

MANAGEMENT IMPLICATIONS

ABSTRACT

INTRODUCTION

STUDY AREA

METHODS

Analysis

RESULTS

TCH

High Use Paved Roads

High Use Trails

High Use Features

DISCUSSION

MANAGEMENT IMPLICATIONS

CHAPTER FOUR: MANAGING FOR GRIZZLY BEAR SECURITY AREAS IN BANFF NATIONAL PARK AND THE CENTRAL CANADIAN ROCKY MOUNTAINS.

ABSTRACT

INTRODUCTION

STUDY AREA

METHODS

Analysis

RESULTS

DISCUSSION

MANAGEMENT IMPLICATIONS

LITERATURE CITED

A variety of development pressures in the Central Canadian Rocky Mountains are currently accelerating habitat fragmentation, displacement and mortality of large carnivores including grizzly bears. The result is loss of connectivity as well as potential loss of viability for large carnivore metapopulations. As demands on the land increase, the cumulative effects from individually minor, yet collectively significant, uses occurring over space and time continue to mount.

Many scientists now recognize that human activity is so prevalent today that ignoring it is now not only almost impossible, but would result in excluding from consideration one of the dominant components in modern biological communities (Primack 1992). Given the existing and proposed human development in the Central Rockies area, multiple human-related impacts have been demonstrated to be a major issue especially for large carnivores (Noss et al. 1996, Mattson et al. 1996, Gibeau 1998). Thus, the process and factors which could lead to extirpation have begun (Shaffer 1981, Gilpin and Soule 1986).

Recent advances in our understanding of conservation biology, metapopulation dynamics and landscape ecology have focused our concern on the long term viability of large carnivores. As the landscape becomes ever increasingly human dominated, we incrementally impinge on a population's ability to acquire needed resources. The challenge ahead lies in trying to understand the process and not just understanding the population (J. Weaver pers. comm.)

One of the primary concerns with high levels of human presence is loss of habitat effectiveness which could have an influence on carrying capacity, hence recruitment and survivorship. This may have a more subtle but at least as profound an impact on animal populations as direct mortalities. Comparing spatial and temporal preference for, or avoidance of, different distributions of human activities may also give insights into levels of habitat alienation. In addition, one of the key factors in metapopulation persistence is dispersal (Hansson 1991). Chance of extirpation from a number of stochastic forces increases with loss of connectivity and associated fragmentation (see Hanski and Gilpin 1991 for overview). The effects of fragmentation have emerged as one of the important topics of the science of landscape ecology (Forman and Godron 1986).

In this study, I address the issues of habitat alienation and landscape fragmentation, through a detailed research program focusing on grizzly bears. This species requires immediate attention because of its lack of resilience in the face of change (Weaver et al. 1996). My focus on grizzly bears is viewed as one component nested within a much larger research agenda. That agenda seeks to apply a large carnivore conservation strategy to the Rocky Mountain Park Complex as outlined by World Wildlife Fund (Dueck 1990, Paquet and Hackman1995). Ultimately, if managers can identify principles concerning responses to certain human activities, they will be better able to predict and thus avoid impacts (Gutzwiller 1991).

Several different factors contributed to a growing concern for grizzly bears in Alberta. In 1990 the province of Alberta released its grizzly bear management plan (Nagy and Gunson 1990). This document clearly showed not only historic declines of grizzly bears in the province, but over hunting, especially during 1980-1988. In 1992 the Federal government enacted the Canadian Environmental Assessment Act which broadened the scope of traditional environmental assessment to consider the cumulative effects of developments at a landscape scale. The following year (1993) the Alberta Environmental Protection and Enhancement Act passed. It also included a provision for assessing the cumulative impacts of development proposals. Research based in Yoho and Kootenay national parks showed that individual grizzly bears may enter four different management jurisdictions in a year (Raine and Riddell 1991). Herrero (1994) suggested that grizzly bear populations in Canadian national parks by themselves were probably all too small for a high probability of long term persistence, and therefore integrated management with surrounding provincial or territorial lands would be required. These and other factors all demonstrated that grizzly bears would become a focal species for cumulative effects assessment (Herrero et al. 1998).

Because of the species’ biological characteristics, grizzly bears recover slowly if at all from population declines, and only if mortality factors have been brought under control (Mattson et al. 1996). These biological characteristics are part of the reason why human activities can have such a significant impact on grizzly bears. Whether land is managed as parks, commercial forests or privately, management practices must respond to the grizzlies’ needs if these bears are to survive. There is an urgent need for scientific data on grizzly bears to help land managers better understand the affects of human activities on this species. Previous research has focused on the effects of roads and road density on grizzly populations (IGBC 1998 and references therein). Until now, little research has been devoted to the effects of non-motorized, tourism-oriented activities on grizzly bears.

The overall goal of my research was to understand how access to resources and encounters with humans, influence grizzly bears. Specific research objectives included:

1. To detect spatial and temporal activity patterns of bears given various levels of human influences.

2. Determine how the distribution of human influences affects a bear's ability to use the landscape.

3. Determine if sufficient habitat connectivity exists to allow bears access to resources.

4. Determine what, if any adjustments to human activities would give bears better access to resources.

5. Suggest management alternatives for integrating land uses compatible with bear habitat needs for the study area.

The Bow River Watershed within the Central Rockies Ecosystem was chosen as the primary study area (Figure 1). This is 11,400 km²of mountainous terrain 50-180 km west of Calgary, in southwestern Alberta. The area includes a portion of Banff National Park (BNP) and adjacent Alberta Provincial lands known as Kananaskis Country. Neither jurisdiction allows grizzly bear hunting although bears are exposed to some hunting outside the Bow River Watershed. Differing agency mandates oversee preservation, industrial tourism, recreation, forestry, oil and gas extraction, mining and stock grazing. Native councils, towns and municipalities, commercial developers and

Figure 1.The study area, highlighting the Bow River Watershed within the Central Rockies Ecosystem.

residential owners diversify land administration even further. The combination of a highly developed transportation system and elaborate infrastructure makes this one of the most human dominated landscapes in the world where a population of grizzly bears still survives.

Between 1994-1998 I used radio telemetry to gather information on individual grizzly bears during the non-denning season. Radio-collared individuals provided information on movements, home range, habitat use and reactions to human activity. These bears were monitored from the air and ground wherever they went and my budget permitted. Aerial monitoring gave infrequent, but relatively unbiased location data. This facilitated understanding of home range, movements and habitat use. Ground-based telemetry allowed intensive monitoring of grizzly bear activities related to developments such as towns, highways, campgrounds and trails. A detailed description of capture and monitoring methods are documented in Gibeau and Herrero (1995), Stevens et al. (1999) and at www.canadianrockies.net/grizzly. While this research did not concentrate on population ecology parameters, I also gathered basic demographic information following Eberhardt et al. (1994) and Hovey and McLellan (1994) to enhance and strengthen overall study findings.

I examined movement patterns of adult female grizzly bears (Ursus arctos) in the Bow River Watershed where the combination of a highly developed transportation system and elaborate infrastructure makes it one of the most intensively developed landscapes in the world where a grizzly bear population still survives. I used 385 daily movement distances from 17 adult female bears for this analysis. I found no difference in adult female movement rate between the traditional day versus night, but substantial difference when dividing the data by when humans were most active. This supports the contention that differences in use patterns are attributable to human activity. Consistent differences in movement rates between wary and habituated adult females, although not statistically significant, further suggested the influence of humans. My intensive movement data showed that habituated adult female bears were not able to take advantage of higher quality habitats in the same manner that wary bears were. The combination of habituated bears using lower quality habitats and demonstrating higher movement rates suggests less energy available for growth and reproduction. Bears within an area of restricted human access used higher quality habitat and traveled less than bears in unregulated areas. Providing security allows bears access to the highest quality habitats, without competition from humans, thus maximizing the reproductive potential of the population. Identification of security areas and limiting human access to these areas should receive the highest priority for habitat conservation of grizzly bears in the Bow River Watershed and the Central Canadian Rocky Mountains.

It is important to understand whether human activity affects spatial and temporal activity patterns of bears, because displacement into sub-optimal habitats or reduced feeding efficiency can affect the net energy available for growth and reproduction (MacHutchon et al. 1998). Research focused on breeding age female grizzly bears. This segment of the population is most critical to population viability (Knight and Eberhardt 1985, Harris 1986) but is also one of the segments most vulnerable to displacement and/or habituation to people (Mattson et al. 1987, Mattson 1990). Human-caused displacement of bears from important foraging areas can result in reduction of habitat effectiveness and carrying capacity (Gibeau 1998). Habituation, the progressive waning of a response to a neutral stimulus (Thorpe 1956), allows some bears to take advantage of near-human environments, however, it may increase the chances of a bear being removed from the population because of human safety concerns (Meagher and Fowler 1989). Habituation in a national park setting can also be dangerous for people, and has been associated with some bear-caused human fatalities (Herrero 1985).

On Alaskan salmon streams, Olson (1993) and Olson et al. (1997) determined that the degree of tolerance of people better predicted the spatial and seasonal patterns of individual bear use than did the more traditional categories of age, sex, and maternal status. Their work demonstrated that non-habituated adult grizzly bears reduced activity at an Alaskan salmon stream in response to an extended lodge season. In contrast, habituated adult bear activity remained similar among years. Understanding female grizzly bear movement patterns is important for managing human activities to minimize their impacts on bears.

I documented daily movement distances of adult female grizzly bears as part of a larger research project investigating the effects of human development on grizzlies. I test whether differences in daily movement are correlated with the level of human access, habituation or habitat quality.

The primary study area encompassed the Bow River Watershed from its headwaters to approximately where it meets the prairies. This is 11,400 km²of mountainous terrain 50-180 km west of Calgary, in southwestern Alberta. This area includes a portion of Banff National Park (BNP) and adjacent Alberta Provincial lands known as Kananaskis Country. Neither jurisdiction allows grizzly bear hunting although bears are exposed to some hunting outside the Bow River Watershed. Differing agency mandates oversee preservation, industrial tourism, recreation, forestry, oil and gas extraction, mining and stock grazing. Native councils, towns and municipalities, commercial developers and residential owners diversify land administration even further.

People access the region using primarily the Trans Canada Highway, a major transcontinental transportation route, that bisects the study area east to west. Several high speed, two lane paved roads serve as arterial transportation routes. Numerous two lane paved secondary roads complete the transportation system through most of the low elevation valleys. I know of no other area within occupied grizzly bear habitat in North America that has such an extensive network of high speed, high volume highways.

Human presence is widespread both within and outside of BNP. Three towns, Banff, Lake Louise and Canmore are world-renowned tourist destinations that attract approximately four million visitors annually. Developments, in addition to the towns that support tourism and industry, include a transcontinental railway, numerous hotels, campgrounds and picnic areas, 5 golf courses, 5 downhill ski facilities and an extensive network of hiking, biking and equestrian trails. The combination of a well-developed transportation system and elaborate infrastructure make the Bow River Watershed one of the most intensively developed landscapes in the world where grizzly bears still survive.

Topographic features include rugged mountain slopes, steep-sided ravines and flat valley-bottoms. Valley bottoms have much of the most productive vegetation. The climate is continental with long, cold winters and short, cool summers. The aspect and elevation of the mountainous topography modifies climate somewhat. Topography, soil and local climate strongly influence plant communities.

Between 1994 and 1998 I captured and radio-marked grizzly bears in the Bow River Watershed, Alberta (Stevens et al. 1999) and monitored their movements. Individuals were equipped with either a conventional VHF radio collar (Lotek Engineering, Newmarket, Ontario) or aVHF ear tag transmitter (Advanced Telemetry Systems, Isanti, Minnesota). All radio collars were fitted with a breakaway cotton spacer (Hellgren et al. 1988) to ensure that collars would not be worn permanently.

I searched for collared bears at least once per week from the air, weather permitting, using a portable receiver, a right-left switchbox and paired 3-element yagi antennae attached to a Bell Jet Ranger III helicopter; or paired 2-element H antennae attached to a STOL equipped Cessna 337 Skymaster. Aerial tracking followed the techniques of Mech (1983). Aerial fixes were established from an aircraft mounted GPS unit and later transformed to UTM coordinates (North American Datum 1927) using the Geocalc Program (Blue Marble Graphics 1993). I also located bears from the ground opportunistically on a daily basis using a portable receiver, roof mounted omni-directional antenna and 3-element hand-held yagi antenna. Workers used either the loudest signal method or nulls (Springer 1979) to determine bearings from two or more positions (Nams and Boutin 1991). Bearings were plotted on 1:50,000 scale topographic maps with bear locations recorded to the nearest 100 m using the Universal Transverse Mercator (UTM) grid coordinate system. In addition to systematic radio tracking, I conducted periodic 24-hour monitoring of individual animals at hourly intervals to obtain daily movement patterns. Through testing with radio collars placed in known locations I recorded an average telemetry error of 150 m. Radio locations were supplemented by occasional direct observation or reports from the public.

I used both air and ground radio telemetry data from individual adult females to establish daily movement distance. Telemetry data were imported into a geographic information system (GIS) using MapInfo Professional® software (MapInfo Corporation, Troy, New York, USA) for analysis. I calculated daily movement distance (midnight to midnight) for individual female bears by measuring the largest linear distance of >2 radio locations per day. This measurement does not reflect total distance traveled over the course of a day which may be much greater. I also subdivided the data set and measured distance traveled based upon: 1) equal periods where the majority of time was during daylight conditions, hereafter called day (08:00-20:00) versus where the majority of time was during darkness, hereafter called night (20:00-0:800); and 2) differences in hours of peak human activity both on the highways and trail system (Parks Canada unpublished data) being human active (08:00-17:00) versus human inactive (17:00-0:800) periods. Based on field experience (M. Gibeau and C. Mamo pers. observation) I then assigned a level of habituation to each female bear. I used a two-tier (wary and habituated) classification scheme based on definitions by Herrero (1985:51). There were no food-conditioned bears in our sample.

Following Mattson et al. (1992) "bears that were known to exhibit considerable tolerance of humans were considered to be habituated. These bears remained in an area despite close approach of humans, or operated in areas near humans without apparent caution. The tendency of a bear to range near roads or developments by itself was not a basis for concluding that the animal was habituated". Several bears maintained ranges near humans but did not exhibit the close-range tolerance of humans that is considered to be characteristic of human habituation.

A more refined analysis of individual daily movement used a subset of data with >10 radio locations per day. Using program CALHOME (Kie et al. 1996), I calculated total distance moved and minimum convex polygons (MCP) to spatially define each daily movement event. I then used GIS (MapInfo® Corporation, Troy, New York, and Idrisi®, Clark Univ., Worcester, MA ) to summarize each of the MCP in terms of their landscape characteristics: 1) habitat quality, 2) compactness ratio, 3) road density, and 4) total access density.

In the absence of a habitat suitability map for the study area I derived surrogate habitat values using Landsat Thematic Mapper satellite images transformed into a greenness band using the tasseled cap transformation (Crist and Cicone 1984, Manley et al 1992). Mace et al. (1999) found a strong selection by grizzly bears for areas of high greenness. I categorized the image into 12 classes of increasing greenness as an indicator of grizzly bear habitat. Use and expected values for each greenness class calculated from my aerial telemetry data set indicated that the 4 highest classes were used more than expected based on availability (P = 0.0002). These 4 classes were combined into a single map layer to represent preferred or high habitat quality. Several different metrics of habitat quality were used in preliminary tests including raw greenness score, variability in habitat quality, and distance to high quality habitat. I found the most predictive variable to be the percent of high quality habitat within a 1.5 km radius of a radio location (an area that roughly corresponded to an average female daily movement distance in my study area).

Compactness ratio (Eastman 1997) is a measure of the shape of a polygon compared to the most compact shape of the same size, a circle. Compactness ratio (Cr) was determined using the following formula:

Cr = SQRT(Ap/Ac)

where SQRT is the square root function, Ap = the area of the polygon being calculated, and Ac = the area of a circle having the same perimeter as that of the polygon being calculated. The smaller the Cr the closer the shape of the polygon to a perfect circle.

I produced both road (motorized roads) and total (roads and trails) human access density maps using a moving window technique (Pereira and Itami 1991, Mace et al 1996, 1999) with a 1.5 km radius window. The moving window technique calculated linear road or total access kilometers per square kilometer. All unsuitable lands (rock, bare soil, and water bodies) were excluded in the density calculations.

I used summary statistics (SPSS 1998) as the basis for distance comparisons. A Mann-Whitney U statistic was used to test for differences between wary and habituated bears as an initial screening procedure. An unbalanced, multivariate analysis of variance (MANOVA) was used to test for differences in movement distance among habituation class, daily, day versus night, and human active versus human inactive movement distances.

For the more refined analysis of daily MCP events, I initially divided the data set by the level of human activity. I used the Spray River valley (250 km²) as a control area where human use has been restricted to less than 30 parties per month since 1992. This control area was used to compare restricted human access to all other areas with unregulated access. Less high quality habitat was available in the restricted area than in unregulated areas (P < 0.001). I treated the level of habituation and the restricted area as factors and tested their relationship with landscape characteristics and average distances traveled by the bears using an unbalanced, multi variate analysis of variance (MANOVA). Significance was accepted at P < 0.05. The assumption of equal variances was met for the telemetry data set.

I used step-wise discriminant function analysis to distinguish the relative importance of landscape characteristics. Mahalanobis distances criterion was used in a stepwise fashion for variable entry and removal. I estimated the overall power of the model by scrutinizing the eigenvalues, Wilk’s lambda, canonical correlation coefficients, and the percentage of correctly classified cases. To improve power, I opted for a binary model contrasting wary with habituated bears. I judged the relative contribution of the variables by analyzing the order in which the variables were entered or removed from the analysis, combined with the analysis of the structure matrix and the magnitude of the standardized canonical function coefficients.

I collected 385 daily movement distances from 17 adult female grizzly bears during the study. Slightly more than 60 % (n = 237) of the daily movement distances recorded were from wary females (n = 12) while the remainder (n = 148) were from females classified as habituated (n = 5).

The distribution of movement distances was found to be non-normal. However, the similar shapes of the distributions (symmetric to slightly right-skewed) resulted in the variables having similar variances (verified by the Levene's tests of the homogeneity of variance). Results of statistical tests were the same with both the original values and transformed values to normalize the data. Given the robustness of the chosen statistical tool (MANOVA) to departures from normality and the need to maintain clarity, I present results of original values.

No significant difference was detected in daily distance traveled (P = 0.191, power 0.275) between wary and habituated bears (_ = 3.2 km, SE = 0.12 km). I chose to continue with the distinction between wary and habituated bears given that the statistical power indicated a highbability of committing a Type II error and separation may provide biologically meaningful insights (Cherry 1998, Johnson 1999). Evidence is strong from other research that there are differences between wary and habituated bears (Jope 1985, Herrero 1985, Mattson et al. 1992).

I also found no significant difference in distance traveled (F _{3, 252}= 1.18, P = 0.279, power = 0.191) between day (08:00-20:00) and night (20:00-08:00). There were however, significant differences in distance traveled (F _{3, 207}= 9.247, P = 0.003) when the data set was divided between the human active period (08:00-17:00) and the human inactive period (17:00-08:00). Bears moved less during the human active period (1.3 km) than during the human inactive period (1.9 km). Differences were also detected in distance traveled between wary and habituated bears (Fig 1), although they were not statistically significant.

I identified 42 daily events with sufficient sample sizes (_ = 16 radio telemetry locations in 24 hours) from 10 different adult female bears to construct MCP’s. Wary bears (n = 7) contributed 24 daily events whereas habituated bears (n = 3) contributed 18 events. Ten events from 2 bears (both classified as wary) were within the control area where human activity is restricted.

In comparisons between the area of restricted human access and areas of unregulated human access, significant differences were detected in proximity to high quality habitat (F _{2, 39}= 4.42, P = 0.007), compactness ratio (F _{2, 39}= 5.71, P = 0.022), total access density (F _{2, 39}= 15.14, P < 0.001) and distance traveled (F _{2, 39}= 4.42, P = 0.042). MCP’s for bears within the restricted area had a higher percentage of high quality habitat within a 1.5 km radius (10%) than MCP’s for bears in unregulated areas (6%). The compactness ratio of MCP’s within the restricted area was smaller than in unregulated areas. Total access density in the vicinity of MCP’s in the restricted area (0.45 km/km²) was less than MCP’s in unregulated areas (2.1 km/km²). Average total distance traveled within MCP’s in the restricted area (4.0 km) was less than total distance traveled within MCP’s in unregulated areas (6.5 km).

Significant differences were also detected in the sample of daily MCP events between wary and habituated bears for proximity to high quality habitat (F _{2, 39}= 26.36, P < 0.001) and road density (F _{2, 39}= 10.37, P = 0.003). MCP’s for wary bears had a higher percentage of high quality habitat within a 1.5 km radius (18%) than MCP’s for habituated bears (5%) (Fig 2). Average road density in the vicinity of MCP’s from wary bears was less (0.0 km/km²) than MCP’s from habituated bears ( 1.1 km/km²).

Discriminant function analysis for wary versus habituated bears produced a 78.6% cross-validated correct classification rate. Wilk’s Lambda was low (0.49) denoting relatively high discriminating power. The high canonical correlation coefficient (0.71) indicated the model discriminated well between the groups. The structure matrix revealed the percentage of high quality habitat within a 1.5 km radius (-0.618) contributed most to the discriminant function. The negative sign indicated that habituated bears are found within a lower percentage of high quality habitat. The second most important variable was total access density (0.311) and thirdly, compactness ratio (0.284). Results of the standardized canonical function coefficients were similar. Based on the analysis, the most important landscape characteristics for wary bears were: 1) high percentage of high quality habitat within a 1.5 km radius, 2) low total access density, and 3) low compactness ratio.

I found no difference in bear movement rate for adult females between the conventional division of day versus night. However, I detected substantial difference when dividing the data by when humans were active (Fig. 1). This is consistent with the findings of Olson et al. (1997) that differences in use patterns are attributable to human activity. Consistent differences in movement rates between wary and habituated adult females (Fig. 1), although not statistically significant, further suggested an influence of humans. Another apparent influence was observed in my intensive movement data which showed that habituated adult female bears were not able to take advantage of higher quality habitats (Fig. 2) in the same manner that wary bears were. The combination of habituated bears using lower quality habitats and demonstrating higher movement rates has obvious implications for the net energy available for growth and reproduction.

While these implications on fitness and reproduction are most acute for habituated bears they are not limited to this subset of the population. Although the sample is small, movement patterns of the two adult females within the control area also demonstrated an impact of human activity. Bears within the area of restricted human access used higher quality habitat and traveled less than bears in unregulated areas despite the fact that less high quality habitat was available in the restricted area. The lower compactness ratio in the restricted area suggests that bears used the landscape more adeptly.

While it is obvious that total access density would be lower in the control area than unregulated areas, the differences were large (0.45 km/km²versus 2.1 km/km²). This is because, for the most part, there are few places in the Bow River watershed that do not have some kind of human access. Even more dramatic were the differences in road density between areas used by wary and habituated bears.

Overall, both wary and habituated adult female grizzly bears were affected by human presence as evidenced by my discriminant function analysis. In the relative absence of humans, wary bears were characterized by more efficient use of higher quality habitats with less movement. Increased human presence eroded this habitat optimization to a point where habituated bears traveled further in sub-optimal habitats. Females that have access to predictable and high value foods such as meat and berries attain greater adult size, mature earlier and have larger litters than those with access only to foods with low nutritional value such as roots (Hilderbrand et al. 1999, Mattson et al. 1999, Nagy and Haroldson 1989, Rogers 1977).

Daily activity patterns of grizzly bears have been found to vary widely. Some studies have found grizzly bears to be diurnal (Stemlock and Dean 1983, Wenum 1998, MacHutchon 1998). Others have found grizzly bears to be more crepuscular (Harting 1985, Gunther 1990, McCann 1991). Several authors have suggested this variability is due to grizzly bear’s ability to alter their temporal and spatial activity patterns in response to human activity. One study (MacHutchon 1998) found variation with age and sex classes as well as level of human activity. Mattson (1990) believed that response to human activity is a function of several factors including the nature and extent of historical interactions with humans, availability of human foods, demographics and size of the population, and distribution of habitats. Given this multitude of variables we might expect that all bears end up responding to human activity somewhere along the continuum between extreme wariness and habituated behavior.

Adult females are the reproductive engine of grizzly bear populations, and their success is the key to long term population persistence. A population’s resilience is determined by the resilience of the reproductively most important cohort, adult females. Providing adult female grizzly bears with the highest level of protection possible should be a management priority. Managing human impacts on individual grizzly bears and the population is key to this provision.

My findings reinforce those of Mattson (1993) in emphasizing the importance of security areas where female grizzly bears can meet their energetic requirements relatively free from encounters with humans. Security areas help to reduce the incidence of habituated bears, bears killed out of self-defense, and bears removed by management agencies because of unacceptable behavior.

Providing security allows bears access to the highest quality habitats without competition from humans, thus maximizing the reproductive potential of the population. Identification of security areas and limiting human access to these areas should receive the highest priority for habitat conservation of grizzly bears in the Bow River Watershed and the Central Canadian Rocky Mountains.

No information exists to evaluate the effects of high-speed, high-volume highways on grizzly bear movement. I document the permeability of several highways to a grizzly bear population in a landscape where human presence is widespread. One highway with 24-hour, year-round high traffic volumes served as an effective barrier for adult female movement and a filtered-barrier for males. Traffic volume appeared to be a key variable in the permeability of highways for grizzly bears. Highway crossings by grizzly bears were concentrated in specific locations and occurred during night as well as day. Zones of high frequency road crossings in my study area were characterized by lower than average total access density, closer to a major drainage, more rugged terrain and higher quality habitat. Significant potential currently exists for permanent habitat and population fragmentation to occur along the Trans Canada Highway. An adaptive management approach will be crucial, with population connectivity being of paramount importance, as we continue to gain knowledge of grizzly bear response to highways.

The ecological effects of roads have been the focus of many conservation biologists in the last decade (see Evink et al. 1996, Forman and Alexander 1998 for reviews). Studies show that roads affect mammal populations in numerous ways, from causing habitat loss and habitat alienation (i.e., sensory disturbance) to presenting physical barriers and causing road mortality (Adams & Geis 1983, Woodward 1990, Van der Zee et al. 1992, Belden and Hagedorn 1993, Brandenburg 1996, Romin and Bissonette 1996). Among these effects, habitat fragmentation and physical barriers pose what many conservation ecologists consider the greatest risk to maintaining species diversity, with demographic and genetic effects upon many species including grizzly bears (Wilcox and Murphy 1985, Ralls et al. 1986, Servheen and Sandstrom 1993, Dale et al. 1994, Mills and Smouse 1994, Forman and Alexander 1998).

Transportation routes cut across landscapes, fragmenting large areas for species such as grizzly bears. Vegetative hiding cover is always removed from the transportation corridor surface and along some portion of the right-of-way, thus making the corridor inhospitable or dangerous to grizzly bears. Previous research on response by grizzly bears to roads has been confined to interactions with tertiary or unimproved road systems (Archibald et al. 1987, Mattson et al. 1987, McLellan and Shackleton 1988, McLellan and Shackleton 1989a, Kasworm and Manley 1990, Mace et al. 1996). To my knowledge no information exists to evaluate the potential of high-speed, high-volume highways to disrupt or prevent movement within occupied grizzly bear habitat.

I document the permeability of several highways to grizzly bear movement in a landscape where human presence is widespread. In mountainous terrain throughout the world, valley bottoms are the preferred habitats for both humans and wildlife. The Bow River Watershed is no exception, with several high speed, high volume highways bisecting major valley systems. Using radio telemetry information, I tested the hypothesis that grizzly bear crossing rates do not differ between highway configurations, and that grizzly bears do not differentiate between crossing areas and random. I address the potential effects of avoidance of highways, as well as the implications of road and total access density on grizzly bears.

The study area encompassed the Bow River Watershed from its headwaters to approximately where it meets the prairies. This is 11,400 km²of mountainous terrain 50-180 km west of Calgary, in southwestern Alberta. This area includes a portion of Banff National Park (BNP) and adjacent Alberta Provincial lands known as Kananaskis Country. Neither jurisdiction allows grizzly bear hunting although bears are exposed to some hunting outside the Bow River Watershed. Differing agency mandates oversee preservation, industrial tourism, recreation, forestry, oil and gas extraction, mining and stock grazing. Native councils, towns and municipalities, commercial developers and residential owners diversify land administration even further.

Human presence is widespread both within and outside of BNP. Three towns, Banff, Lake Louise and Canmore are international tourist destinations that attract approximately four million visitors annually. Developments in addition to the towns that support the tourism industry include numerous hotels, campgrounds and picnic areas, 5 golf courses, 5 downhill ski facilities and an extensive network of hiking, biking and equestrian trails.

I know of no other area within occupied grizzly bear habitat in North America that has such an extensive network of high-speed, high-volume highways ( Fig 1).

People access the region using primarily the Trans Canada Highway (TCH), a major transcontinental transportation route, that bisects the study area east to west. It is a 4-lane divided highway through most of the study area with a summer (June - September) average traffic volume of 21,000 vehicles per day (Parks Canada, unpubl. data) and an observed average speed of 110-115 km/hr (Royal Canadian Mounted Police pers. comm). Thirty-five kilometers of this busy freeway have not yet been upgraded to a 4-lane divided highway along the western edge of the study area. Forty-five km of the TCH through BNP have been fenced to keep wildlife off the roadway. Wildlife crossing structures have been placed throughout the fenced section to facilitate movement across the highway (Clevenger and Waltho 2000). The remaining 40 km of the TCH along the eastern portion of the study area is a 4-lane divided highway but without a wildlife fence.

Several high speed, 2-lane paved highways serve as arterial transportation routes in the study area (Fig 1). Highway 40 intersects the TCH along the front range of the Rocky Mountains dissecting Kananaskis Country from north to south. It has a summer average traffic volume of 3,075 vehicles per day with an observed traffic speed of 105-110 km/hr (Royal Canadian Mounted Police pers. comm.). Highway 93 North intersects the TCH west of the town of Lake Louise paralleling the continental divide range north to Jasper. Highway 93 North has a summer average traffic volume of 3,530 vehicles per day with an observed traffic speed of 110-115 km/hr (Banff Highway Patrol pers. comm.). The Bow Valley Parkway (BVP) parallels the TCH on the opposite side of the valley between the towns of Banff and Lake Louise. Although this highway has no paved shoulders it is similar to other two lane highways with a summer average traffic volume of 2,230 vehicles per day with an observed traffic speed of 80-85 km/hr (Banff Highway Patrol pers. comm.). Numerous 2-lane paved secondary roads complete the transportation system through most of the low elevation valleys. There are very few gravel roads in the study area.

Topographic features include rugged mountain slopes, steep-sided ravines and flat valley bottoms. The climate is continental with long, cold winters and short, cool summers. The aspect and elevation of the mountainous topography modifies climate somewhat. Topography, soil and local climate strongly influence plant communities.

Between 1994 and 1998 I captured and radio-marked grizzly bears in the Bow River Watershed, Alberta (Stevens et al. 1999), and monitored their movements. I searched for collared bears at least once per week, weather permitting, from fixed-wing aircraft or helicopter. Bears were also searched for from the ground opportunistically on a daily basis using standard techniques (Kenward 1987, Samuel and Fuller 1996). In addition to systematic radio tracking, I conducted periodic 24-hour monitoring of individual animals at hourly intervals to obtain detailed information on fine-scale movement patterns in the vicinity of roads. Through testing with radio collars placed in known locations, I recorded an average telemetry error of 150 m. I supplemented radio locations by occasional direct observation reported by the public. Locations were plotted on 1:50,000 topographic maps, assigned a Universal Transverse Mercator (UTM) coordinate and later converted to digital Geographic Information System (GIS) maps using MapInfo Professional® software (MapInfo Corporation, Troy, New York, USA). For this analysis I used several subsets of the telemetry data (described below) to avoid biases of over sampling and to maximize independence between telemetry locations (Hurlbert 1984) required for some analyses.

To determine the minimum number of highway crossings by grizzly bears I used data from weekly aerial relocations because the sampling intensity was the same for all bears. I obtained a minimum estimate of crossing frequency by counting the number of times bears crossed the four highways in the study area.

The entire telemetry data set, including 24-hour monitoring, was used to identify areas on the highway where bears chose to cross. I identified highway crossings by plotting consecutive radio locations that were obtained within 24 hours and were <1 km from the highway. I selected radio locations within this distance in order to provide greater accuracy in determining the estimated crossing location. An estimated crossing location was identified as an intersection of a straight line between 2 consecutive radio locations and the highway.

To identify whether there was a pattern in highway crossings I created a crossing density map using a moving window technique with a 1.5 km radius. To facilitate statistical analysis I categorized the crossing density map into zones of high (>4) crossings and low (<4) crossing frequencies because the density map revealed crossing locations were either highly clustered or solitary.

The time of highway crossing was estimated by interpolating time from distance calculations following Brandenburg (1996). Assuming a straight line and constant rate of travel, highway crossing time (hct_i) was determined using the following formula:

hct_i= (r_i/d_i/) (et_i) + (t_i)

where r_iis the distance from the highway to the location occurring before the bear crossed the highway, d_iis the distance between sequential locations, et_iis the time elapsed between sequential locations, and t_iis the time of the location before the bear crossed the highway.

I obtained traffic volumes for each estimated time of highway crossing from hourly traffic counter data collected for each highway in the study area (Alberta Transportation and Utilities, and Parks Canada unpubl. data). There were no traffic volume data for Highway 40 in 1994, therefore I used 1995 volume data for the same date and time as the 1994 crossings (n=6). For analysis, hourly traffic volume assigned to each crossing location was categorized into either high volume (_ = 111 for BVP and 117 for Highway 40) or low volume (_ = 13 for BVP and 11 for Highway 40). Categories were based on the inflection point where significant change was observed in traffic volumes.

Using one radio location per day and program CALHOME (Kie et al. 1996), I constructed 99% minimum convex polygon (MCP) home ranges (Gibeau and Herrero 1999). I grouped individual bear home ranges into composite home ranges for male and female bears. Both road and total access density were calculated for each of the composite home ranges. Roads were defined as those passable by motor vehicle. Trails were restricted to non-motorized travel. Total access density was defined as all roads and trails.

Attributes associated with crossing zones were analyzed at two scales: 1) a fine site level scale within a 150 m radius of the highway crossing site location and, 2) a broader habitat scale within a 1 km radius of the crossing site location. I compared characteristics of high and low frequency crossing zones to the overall characteristics along the total length of the highway by stratified random sampling to extract landscape attributes in the two areas. Landscape attributes for the observed crossing zones and the random highway points (expected) were analyzed using Idrisi® (Clark Univ., Worcester, MA ) GIS map layers of: 1) proximity to high quality habitat, 2) proximity to nearest major drainage, 3) terrain ruggedness and 4) total human access density.

In the absence of a habitat suitability map for the study area I derived surrogate habitat values using Landsat Thematic Mapper satellite images transformed into a greenness band using the tasseled cap transformation (Crist and Cicone 1984, Manley et al 1992). Mace et al. (1999) showed a strong selection by grizzly bears for areas of high greenness. I categorized the image into 12 classes of increasing greenness as an indicator of grizzly bear habitat. Use and expected values for each greenness class calculated from my aerial telemetry data set indicated that the four highest classes were used more than expected based on availability (P = 0.0002). These four classes were combined into a single map layer to represent preferred or high habitat quality.

Proximity to nearest drainage measurements used digital hydrology data for the Bow River Watershed (Parks Canada and Alberta Environmental Protection). I eliminated the Bow and Kananaskis Rivers to focus the analysis on the drainage elements that, given their orientation with respect to the highways, might be conducive to the crossings of highways. A drainage was defined as a permanent watercourse mapped at a scale of 1:50,000.

Terrain ruggedness (Tr), an index capturing the level of complexity of terrain (Nellemann and Fry 1995), was calculated using the following formula:

Tr = (De^xAc)/(De+Ac)

where De = the density of contour lines within a given window, and Ac = an index of aspect variability (defined as the frequency of cardinal aspect change) within a given window. I used a circular window of 1.5 km radius that roughly corresponds to an average female daily movement distance. The resulting map of terrain ruggedness classified the landscape where the higher the value the greater the topographic diversity.

Both road (motorized roads) and total (roads and trails) human access density maps were produced using a moving window technique (Pereira and Itami 1991, Mace et al 1996, 1999) with a 1.5 km radius window. The moving window technique calculated linear road or total access kilometers per square kilometer. All unsuitable lands (rock, bare soil, and water bodies) were excluded in the density calculations.

An unbalanced multivariate analysis of variance (MANOVA) was used to test for the differences in landscape attributes between zones of high and low frequency crossing and the average conditions found along the highways. I used post hoc, multiple comparisons to identify differences in crossing zones for both the immediate vicinity (150 m) and broad scale (1 km) buffers around highways. Significance was accepted at P < 0.05. The assumption of equal variances was met for the telemetry data set.

I used discriminant function analysis to distinguish the relative importance of landscape characteristics. Mahalanobis distances criterion was used in a step-wise fashion for variable entry and removal. I estimated the overall power of the model by scrutinizing the eigenvalues, Wilk’s lambda, canonical correlation coefficients, and the percentage of correctly classified cases. To improve power, I opted for a binary model contrasting zone of high frequency crossing with the entire highway. I judged the relative contribution of the variables by analyzing the order in which the variables were entered or removed from the analysis, combined with the analysis of the structure matrix and the magnitude of the standardized canonical function coefficients.

I collected 7,380 telemetry locations from 54 grizzly bears (16 adult male, 11 subadult male, 19 adult female, 8 subadult female) between 1994-98. Twenty one of those 54 bears had home ranges in the same valley as a high speed highway (6 adult male, 2 subadult male, 10 adult female, 3 subadult female). Using the aerial telemetry data as a sample of equal-intensity monitoring, I recorded differences in permeability between highways (Table 1). Small sample size precluded meaningful statistical testing.

Three adult males, 2 subadult males, and 1 subadult female crossed the TCH during the 5 year period. Both subadult males were habituated and ultimately removed from the population. The subadult female was also habituated but was not removed.

Table 1. Minimum number of recorded grizzly bear highway crossings in

the Bow River Watershed, Alberta, 1994-1998.

- Bear’s home range was not in the same valley as the highway, thus no interaction.

B Bear’s home range was in the same valley, but did not cross the highway constituting a home range boundary.

In Kananaskis Country, 3 adult males, 5 adult females, 1 subadult male, and 2 subadult females crossed Highway 40. Two adult males, 2 adult females, 1subadult male and 1subadult female crossed the BVP. Both subadults are the bears that also crossed the TCH. One adult female and 1 subadult male crossed Highway 93, although these were the only bears in my sample in that vicinity.

There was a total of 33 crossings by 6 bears on the TCH for a median of 1.5 crossings per bear. Two bears crossed Highway 93, 17 times for a median of 8.5 crossings per bear. On Highway 40 there was a median of 12.0 crossings per bear, while on the BVP the median was 6.0. One adult male accounted for 66% of all TCH crossings. Of the four highways, the TCH formed a home range boundary for 6 adult females, while Highway 93 bordered the home range of one adult female.

Road density within the composite female home range was 0.16 km/km². Total access density within the composite female home range was 1.30 km/km². Road density within the composite male home range was 0.44 km/km². Total access density within the composite male home range was 1.54 km/km².

Only the BVP and Highway 40 had sufficient data that met my established criteria of telemetry locations within 24 hours and 1 km of the highway to analyze highway crossing zones. Three male grizzly bears crossed the BVP 18 times and 2 females crossed 13 times for a total of 31 crossings. Two male grizzly bears crossed Highway 40, 12 times and 6 females crossed 28 times for a total of 40 times. Discrete areas of high frequency grizzly bear crossing were identified for both the BVP (Fig. 2) and Highway 40 (Fig. 3).

Timing of highway crossing varied over the 24 hour period. There was no clear pattern of any nocturnal or crepuscular activity crossing roads as most crossings took place between 23:00-07:00 hours (38%) and between13:00-18:00 (38%) (Fig. 4). The limited amount of data precluded analysis of each highway separately.

The number of grizzly bear highway crossings varied in relation to traffic volume and intensity of highway use. Most single event crossing areas (identified as zones of low frequency crossing) were made during periods of high traffic volume. In contrast, areas of multiple crossings (zones of high frequency crossing) were used equally during periods of both high and low traffic volumes (Table 2).

Table 2. Number of grizzly bear highway crossings for different traffic volumes in the Bow River Watershed, Alberta, 1994-1998.

The distribution of distance measurements of crossing location attributes was found to be non-normal. However, the similar shapes of the distributions (symmetric to slightly right-skewed) resulted in the variables having similar variances (verified by the Levene's tests of the homogeneity of variance). Results of statistical tests were the same with both the original values and transformed values to normalize the data. Given the robustness of the chosen statistical tool (MANOVA) to the departures from normality and the need to maintain a clarity of presentation, I present results of original values.

High quality habitat was closer to zones of high frequency crossing than the entire highway in the 150 m buffer for both the BVP (F _{2, 560}= 6.92, P < 0.001) and Highway 40 (F _{2, 862}= 8.61, P < 0.001) but was not significantly closer within a 1 km buffer for either highway. No significant differences were detected in proximity to high quality habitat between zones of high and low frequency crossing in the immediate vicinity of either highway.

Significant differences were detected in the distance to a major drainage between zones of high frequency crossing, and both zones of low frequency crossing and the entire highway in the 150 m buffer for the BVP (F _{2, 560}= 51.16, P < 0.001) and Highway 40 (F _{2, 862}= 26.61, P < 0.001). Similar results were found for the 1 km buffer for both the BVP (F _{2, 1103}= 22.70, P < 0.001) and Highway 40 (F _{2, 2552}= 158.85, P < 0.001). In all cases, major cross-drainages were closer to zones of high frequency crossing than to either zones of low frequency crossing or the entire highway.

Significant differences in terrain ruggedness were evident along the BVP between zones of high frequency crossing, and both zones of low frequency crossing and the entire highway in the immediate vicinity of the highway (F _{2, 560}= 28.08, P < 0.001) and 1 km buffer (F _{2, 1103}= 14.84, P < 0.001). Significant differences in terrain ruggedness were also observed along Highway 40 between zones of high frequency crossing, and both zones of low frequency crossing and the entire highway in the immediate vicinity of the highway (F _{2, 862}= 143.86, P < 0.001) and 1 km buffer (F _{2, 2552}= 136.34, P < 0.001). In all cases, terrain ruggedness values were higher in zones of high frequency crossing than either zones of low frequency crossing or the entire highway.

Significant differences in total human access density were evident only for the BVP between zones of high frequency crossing, and both zones of low frequency crossing and the entire highway in both the immediate vicinity of the highway (F _{2, 560}= 126.54, P < 0.001) and 1 km buffer (F _{2, 1103}= 62.37, P < 0.001). Total access density in zones of high frequency crossing were 2.34 km/km²compared to 3.01 km/km²along the entire highway in the 150 m buffer. When analyzing both highways together, significance was detected between zones of high frequency crossing, and both zones of low frequency crossing and the entire highway in the immediate vicinity of the highway (F _{2, 1425}= 15.57, P < 0.001), but not for the 1 km buffer.

Discriminant function analysis for the BVP produced a 88.3% cross-validated correct classification rate. Wilk’s Lambda was low (0.51) denoting relatively high discriminating power. The high canonical correlation coefficient (0.69) indicated the model discriminated well between the groups. TheHighway 40 analysis had an 83.1% cross-validated correct classification rate. Wilk’s Lambda was again low (0.60) denoting relatively high discriminating power. The high canonical correlation coefficient (0.63) indicated the model discriminated well between the groups.

Some differences between the structure matrix and the standardized canonical function coefficients were evident due to a correlation between terrain ruggedness and distance to a major drainage. I report the standardized coefficients because of my desire to use discriminant function analysis as a predictive tool.

The most important contributors to calculating the discriminant score for the BVP were: 1) total access density (0.790), 2) terrain ruggedness (-0.694), 3) distance to a major drainage (0.654), and 4) proximity to high quality habitat (0.176). As the canonical discriminant functions evaluated at group means placed the group centroid for the zone of high frequency crossing on the negative side from the grand means, the negative sign on terrain ruggedness indicates higher ruggedness levels within zones of high frequency crossing.

The most important contributors to calculating the discriminant score for Highway 40 were: 1) terrain ruggedness (1.227), 2) total access density (0.789), and 3) distance to a major drainage (-0.213). As the canonical discriminant functions evaluated at group means placed the group centroid for the zone of high frequency crossing on the positive side from the grand means, the positive sign on terrain ruggedness indicates higher ruggedness levels within zones of high frequency crossing.

The relative crossing index of bears along the four highways indicated that the TCH was the least permeable for grizzly bears (Table 1). The facts that no radio marked adult females crossed this major highway during the study (Gibeau and Herrero 1999), and two-thirds of the crossings were by one adult male, are of the utmost concern. Black bears responded similarly to a high-speed highway in North Carolina by crossing only 14 times (n = 9 bears) in 4 years and one bear was killed while attempting to cross (Beringer et al. 1990). Highway 40 had the highest number of crossings per bear. This can be partially explained by the fact that the recorded Highway 40 traffic volume is most likely overestimated as a large proportion of the vehicles do not constitute through-traffic (P. Kilburn, Alberta Transportation and Utilities pers. comm.).

Road density has been proposed as a broad index of the ecological effects of roads in a landscape (Forman and Hersperger 1996, Forman et al. 1997, Forman and Alexander 1998). Besides constituting a source of mortality, roads can disrupt movements, cause habitat fragmentation and increase human access, the latter ultimately leading to increased mortality (Brody and Pelton 1989, Beringer et al. 1990). The reported threshold density for functioning landscapes with large carnivores is approximately 0.6 km/km²(Forman et al.1997) and is based on field studies of wolves, cougars and brown bears (Thiel 1985, Jensen et al. 1986, Mech et al. 1988, Van Dyke et al. 1986, Clevenger et al. 1997). Previous studies have emphasized that roads themselves are not the problem but rather the human access created by roads. This access can lead to greater disturbance, vulnerability to legal and illegal killing, and management removal due to conflicts (Mech et al. 1988, Brody and Pelton 1989, Beringer et al. 1990, Clevenger et al. 1997).

In the Bow River watershed, road density by itself does not encapsulate the overall effect of humans. I feel a better measure is total access density of roads and trails which in combination contribute to sensory disturbance. Composite female and male grizzly bear home ranges in this study area were well below the Forman et al. (1997) proposed threshold road density of 0.6 km/km², but well above that threshold when considering total access density. Brody and Pelton (1989) indicated that the threshold density for high traffic volume roads (e.g., interstate highways) in bear ranges is extremely low and will be much lower than 0.6 km/km². Contrary to other studies of road effects on large carnivores, I documented how one highway with 24-hour, year-round high traffic volumes can serve as an effective barrier for adult female movement and a filtered barrier for males.

My results, in part, support earlier findings that as traffic volumes increase, crossings by bears decrease (Brody and Pelton 1989, Beringer et al. 1990, Brandenburg 1996). Major highways such as the Trans-Canada in the Bow River Valley can severely disrupt movements by adult female grizzly bears and to a lesser extent, male bear movements. Most of the TCH crossings were by one adult male, suggesting that successfully traversing a major highway is likely a learned behavior that some individuals become adept at over time.

No literature exists on grizzly bear response to roads with traffic volumes in the order of magnitude I have documented. Brandenburg (1996) reported on how roads with traffic volumes between 100-10,000 vehicles per day affected movements of black bears. An interstate highway bisected black bear study areas in North Carolina (Brody and Pelton 1989, Beringer et al. 1990), however, the timing, frequency or distribution of crossings were not reported. Road crossings by black bears at another North Carolina study area were primarily nocturnal and/or during times of low traffic volume (Brandenburg 1996). I found that highway crossings by grizzly bears were concentrated in specific locations (Fig. 2 and 3) and occurred during night as well as day (Fig. 4). The rugged topography of this study area, characterized by glacier-carved valleys and steep-walled side drainages, probably limits the number of possible cross-valley and cross-highway locations. The relatively high number of daytime crossings by grizzly bears was surprising and difficult to explain. Bears had a strong tendency to use zones of high frequency crossings during low traffic volumes, however, during periods of high traffic volume, bears used zones of low frequency crossing as much as the zones of high frequency crossings. The scattered distribution of crossing areas during heavy traffic (single events) I attribute to random crossing site selection by bears, i.e., attempting to cross anywhere possible.

Zones of high frequency road crossing for grizzly bears in my study area were characterized by lower total access density, close proximity to a major drainage, more rugged terrain, and higher quality habitat. While these results are for the most part intuitive and consistent with our knowledge, this first attempt to characterize highway crossing zones identifies specific parameters that can be applied in an adaptive management approach.

Managers have known for some time the direct effect of high-speed roads (mortality) and indirect effect of unpaved or secondary road densities (disturbance, human access) on many wildlife species. However, major transportation systems pose problems that have been virtually disregarded in the past, the most serious being habitat fragmentation and barrier effects to wildlife movement. To date much of our experience related to grizzly bears and roads can be summarized, arguably, into management of human access (IGBC 1998). The effect of high speed, high volume highways on grizzly bears has not been addressed until now.

Permeability was significantly compromised for grizzly bears along the TCH even though this highway had 3 different configurations in the study area (40 kilometers of 4-lane no wildlife fence, 45 kilometers of 4-lane with a wildlife fence, and 35 kilometers of 2-lane). This leads me to believe that traffic volume appears to be a key variable in the permeability of highways for grizzly bears. The higher the traffic volume the less likely a bear will cross the highway. Management agencies in the Bow River Watershed now find themselves in a particularly difficult position with respect to maintaining a contiguous grizzly bear population in the Central Canadian Rocky Mountains. My results, along with home range analysis (Gibeau and Herrero 1999), suggest that the TCH both inside and outside of Banff National Park is a barrier to female grizzly bear movement, and a filtered barrier to male movement. The implications of a movement barrier are unknown, however it certainly points to the initial stages of islandization. Given grizzly bear’s large home ranges, significant potential currently exists for permanent habitat and population fragmentation to occur. Management agencies must maintain access to high quality habitat, especially for adult females.

My analysis demonstrates grizzly bears cross some highways in very site-specific locations, enabling us to predict crossing zones. This will be useful to managers contemplating crossing structures for wildlife passage in an attempt to mitigate the adverse effects of the TCH. While a significant amount of time, energy, and money has gone into the design, placement and construction of these crossing structures (Leeson 1996), the question remains as to how grizzly bears will travel through human access zones to get to the mouth of these structures in order that they may cross the highway. Further, is it possible to mitigate highway effects for grizzly bears if they are repelled by highways and fail to even get close to them? Currently, the combination of intense sensory disturbance from high traffic volumes and overall high human access density precludes most grizzly bears from being in the vicinity of the TCH. For bears to have the possibility of using these crossing structures overall human access density must be reduced, especially in areas identified as potential crossing zones. An adaptive management approach will be crucial as we continue to gain knowledge of grizzly bear response to highways.

National Parks by themselves cannot sustain a regional grizzly bear population (Herrero 1994). Some of the best chances for grizzly bear persistence come from outside National Parks (McLellan et al. 1999) and hence a cooperative and coordinated management approach is critical, with population connectivity being of paramount importance.

Few studies have reported the effects of multiple human activities on grizzly bears. Unlike most grizzly bear studies that report specific avoidance distances to various human developments from analysis of resource selection, I document the degree and magnitude of grizzly bear responses as a function of multiple interacting variables based on observed median distances to roads, trails and development features in a landscape where human presence is widespread. Female grizzly bears remained further than males from paved roads regardless of the habitat quality or time of day. Males were found closer to paved roads when within or adjacent to high quality habitat and during the period of least human activity. The combination of traffic volume and highway configuration, however, overrides a bear’s attraction to high quality habitats for high-speed, high-volume, highways. High human presence underlies the apparent unwillingness of most grizzly bears to use habitats near busy transportation corridors. This avoidance behavior is strongest in the adult segment of the population. Bears were found closer to trails during the human inactive period when within high quality habitat and further from trails when distant to high quality habitat. My data indicated an inverse relationship between the sexes in response to vehicles and traffic noise compared to the response to human settlement and encountering people. Female bears were found further away than males in relation to vehicles and traffic noise, yet found closer than males to human settlement and places where people may be encountered. Those males that were more willing to exploit high quality habitat near roads used both cover and darkness as part of their adaptive strategy. Adult females were the most risk-averse cohort, choosing to avoid humans instead of seeking out high quality habitats. Adult female grizzly bears were influenced most by human activities and development. Management agencies must maintain access to high quality habitat, especially for adult females, and create new opportunities to support the reproductive potential of the population.

Many wildlife populations have been reduced to small fractions of their former size during modern times due to anthropogenic pressures such as habitat loss and overexploitation. This phenomenon is increasing in Canada. Alberta has an expanding economy based significantly on the development of natural resources such as agriculture, oil and gas, forestry and nature-based tourism. Individual grizzly bears, having large home ranges, increasingly come into contact with all of these activities. Herrero (1994) showed that grizzly bear populations in Canadian national parks by themselves were probably all too small to have a high probability of long term persistence, and therefore integrated management with surrounding provincial or territorial lands would be required. Within the boundaries of Banff, Yoho and Kootenay National Parks, research by Gibeau (1998) showed that habitat effectiveness was significantly compromised by development. Whether land is managed as parks, commercial forests or privately, management practices must respond to the grizzlies’ needs if these bears are to survive. There is an urgent need for scientific data to help land managers better understand the effects of human activities on grizzly bears.

The response of grizzly bears to humans has been the focus of much research within the last 15 years. Most of these studies, however, have focused on one type of human activity such as roads (McLellan and Shackleton 1988, Mace et al. 1996), forestry or other industrial activity (Archibald et al. 1987, McLellan 1990), recreation (Jope 1985, Gunther 1990, Olson et al. 1990, Mace and Waller 1996), or facilities (Mattson et al. 1987, Reinhart and Mattson 1990). Few studies have reported the effects of multiple human activities (Mattson et al. 1987, McLellan and Shackleton 1989a, Kasworm and Manley 1990) and then, only using univariate analysis.

I document the distance of grizzly bears to different human activities and development features based on sex, age class and habitat quality in a landscape where human presence is widespread. The Bow River Watershed is one of the most intensively developed landscapes in the world where a grizzly bear population still survives. In this setting, grizzly bears may not be able to avoid humans and still find requisite resources. Using radio telemetry information, I tested the hypothesis that grizzly bears do not differ in their response to roads, trails, and major development features.

The study area encompassed the Bow River Watershed from its headwaters to approximately where it meets the prairies. This is 11,400 km²of mountainous terrain 50-180 km west of Calgary (a city of 800,000 people) in southwestern Alberta. This area includes a portion of Banff National Park (BNP) and adjacent Alberta Provincial lands known as Kananaskis Country. Neither jurisdiction allows grizzly bear hunting although bears are exposed to some hunting outside the Bow River Watershed. Differing agency mandates oversee preservation, industrial tourism, recreation, forestry, oil and gas extraction, mining and stock grazing. Native councils, towns and municipalities, commercial developers and residential owners diversify land administration even further.

Human presence is widespread both within and outside of BNP. Three towns, Banff, Lake Louise and Canmore are world-renowned tourist destinations that attract approximately four million visitors annually. Developments in addition to the towns that support the tourism industry include a multitude of hotels, campgrounds and picnic areas, 5 golf courses, 5 downhill ski facilities and an extensive network of hiking, biking and equestrian trails.

People access the region using primarily the Trans Canada Highway (TCH), a major transcontinental transportation route, that bisects the study area. It is a high-speed, high-volume (21,000 vehicles per day, summer average daily traffic volume; Parks Canada, unpubl. data), 4-lane divided highway through much of the study area. Forty-five km of the TCH through BNP has been fenced to keep wildlife off the road surface. Several high-speed, 2-lane paved roads serve as arterial transportation routes. Numerous 2-lane paved secondary roads complete the transportation system through most of the low elevation valleys. Traffic volumes on these arterial and secondary paved roads are high during the day (>300 vehicles per hour) but low at night (<50 vehicles per hour) which is significantly different than the TCH (Gibeau and Herrero 1998). There are few gravel roads in the study area.

Topographic features include rugged mountain slopes, steep-sided ravines, and flat valley bottoms. The climate is continental with long, cold winters and short, cool summers. The aspect and elevation of the mountainous topography modifies climate somewhat. Topography, soil, and local climate strongly influence plant communities. The landscape can be classified into major ecoregions: montane (1,300 to 1,600 m), subalpine (1,600 to 2,300 m) and alpine (2,300+ m). The montane region is dominated by dry grasslands, wet shrubland and forests of Lodgepole pine (Pinus contorta), Douglas-fir (Pseudotsuga menziesii), white spruce (Picea glauca) and aspen (Populus tremuloides). Subalpine areas are forested with mature stands of lodgepole pine, Engelman spruce (Picea engelmanii), subalpine fir (Abies lasiocarpa) and subalpine larch (Larix lyallii) interspersed by areas of wetland shrub. A mosaic of low shrubs and herbs characterize alpine areas.

Between 1994 and 1998 I captured, radio-marked and monitored grizzly bears in the Bow River Watershed, Alberta. Individuals were equipped with either a conventional radio collar (Lotek Engineering, Newmarket, Ontario) or an ear tag transmitter (Advanced Telemetry Systems, Isanti, Minnesota). All radio collars were fitted with a breakaway cotton spacer (Hellgren et al. 1988) to ensure that collars would not be worn permanently. I located bears from the ground opportunistically on a daily basis using a portable receiver, roof mounted omni-directional antenna and 3-element hand-held yagi antenna. Workers used either the loudest signal method or nulls (Springer 1979) to determine bearings from 2 or more positions (Nams and Boutin 1991). Bearings were plotted on 1:50,000 scale topographic maps with bear locations recorded to the nearest 100 m using the Universal Transverse Mercator (UTM) grid coordinate system. Through testing with radio collars placed in known locations I recorded an average telemetry error of 150 m. Radio locations were supplemented by occasional direct observation or reports from the public.

I searched for collared bears at least once per week from the air, weather permitting, using a portable receiver, a right-left switchbox, and paired 3-element Yagi antennae attached to a Bell Jet Ranger III helicopter or paired 2-element H antennae attached to a STOL equipped Cessna 337 Skymaster. Aerial tracking followed the techniques of Mech (1983). Aerial fixes were established from an aircraft mounted GPS unit and later transformed to UTM coordinates (North American Datum 1927) using the Geocalc Program (Blue Marble Graphics 1993). In addition to systematic radio tracking, I conducted periodic 24-hour monitoring of individual animals at hourly intervals to obtain daily movement patterns.

I used both air and ground radio telemetry data consisting of one relocation per day to avoid biases of over sampling and maximize independence between telemetry locations (Hurlbert 1984). Ground-based telemetry data can be biased in some cases towards only where workers can travel. In this analysis, however, I rely upon ground-based telemetry data because the analysis is specific to areas where workers could travel and sample sizes are larger than the aerial data set. Relocations were categorized by: (1) sex; (2) age: adult (>5 years old) and subadult; (3) season: preberry (den emergence through 15 July) and berry (16 July through den entrance); (4) differences in hours of peak human activity both on the highways and trail system (Parks Canada unpublished data) being human active (08:00-17:00) versus human inactive (17:00-0:800) periods; (5) distance to high quality habitat: within (<150 m, which is consistent with telemetry error), adjacent (150-300 m) or distant (>300 m). Telemetry data were imported into geographic information system (GIS) maps using MapInfo Professional® software(MapInfo Corporation, Troy, New York, USA) for analysis.

In the absence of a habitat suitability map for the study area I derived surrogate habitat values using Landsat Thematic Mapper satellite images transformed into a greenness band using the tasseled cap transformation (Crist and Cicone 1984, Manley et al 1992). Mace et al. (1999) found a strong selection by grizzly bears for areas of high greenness. I categorized the image into 12 classes of increasing greenness as an indicator of grizzly bear habitat. Use and expected values for each greenness class calculated from my aerial telemetry data set indicated that the four highest classes were used more than expected based on availability (P = 0.0002). These four classes were combined into a single GIS map layer to represent preferred or high habitat quality.

The percentage of available vegetative cover was calculated for each 30 m pixel within the study area based on classified Landsat Thematic Mapper satellite images using a moving-window routine. I chose a 1.5 km radius as a moving-window size which approximates the average daily feeding radius of an adult female grizzly bear in my study area. The resulting GIS map provided a measure of the percent cover in the vicinity of each telemetry location.

I quantified human presence using the most recent data of human activity across the region (Gibeau 1998). These GIS maps categorized vector, point and polygon data of all motorized and non-motorized human developments and facilities into high (>100) and low (< 100) users per month (USDA Forest Service 1990, Gibeau 1998) based on visitation records and expert opinion. These data became the basis for measuring the distance of telemetry locations to various human developments and features. I categorized human use by: (1) Trans Canada Highway, (2) high use paved roads, (3) high use trails, and (4) high use features (campgrounds, lodges, picnic area, etc.). Although there were several low use trails within the study area, most were located in association with either the TCH or a high use paved road. Any attempt to partition out the effects of low use trails was masked by these other roadways. There were too few gravel roads, low use roads, and low use features for meaningful analysis.

The nearest distance (>1 m) to each of the above 4 human use categories was calculated in a raster format with a 50 m pixel size using Idrisi® (Clark Univ., Worcester, MA ) GIS software for each telemetry location. The resulting spreadsheet provided distances to four types of human uses, and percent cover for every telemetry location categorized by sex, age, season, time of day, and habitat quality. For comparison, I generated a stratified systematic sample of 2765 random points and calculated the nearest distance to each of the four human use categories for these locations as well.

Analysis of the distance data was complicated by the rugged mountainous topography. Distance measurements did not take into account intervening mountain ranges, therefore I had to impose distance limits to avoid measurements that were in fact in adjacent mountain valleys. I found the average distance between ridge tops for the major valleys in the Bow River Watershed to be 6.5 km. Therefore, I used half that distance (3.25 km) as a maximum to give a high probability that measurements were within the same mountain valley.

I used a Mann-Whitney U statistic to test whether the aerial and ground telemetry data sets, and the stratified systematic sample of random points, came from the same population. An unbalanced analysis of variance (ANOVA) was used to test both main effects and interactions among sex, age, season, distance to high habitat quality, and time of day for the measured distances to each of the four types of human uses. Profile plots were used to visualize the relationship between variables using estimated marginal means. Significance was accepted at P < 0.05. The assumption of equal variance was met for the aerial telemetry data set.

I collected 4,359 daily telemetry locations from 49 grizzly bears (15 adult male, 7 subadult male, 19 adult female, 8 subadult female) during the study. Slightly more than half of the locations were obtained from the ground (n = 2,471, 57%) compared to aerial locations (n = 1,888). There were significant differences between the aerial and ground telemetry data sets (P= 0.002) in observed distances to high use paved roads, high use trails and high use features; however, not for the TCH (P = 0.101). Although the ground-based telemetry data was statistically different from the aerial data set, I report and interpret ground-based results because the analysis is specific to areas where workers could travel and ground based telemetry may provide further biologically meaningful insights (Cherry 1998, Johnson 1999). Significant differences also existed between the aerial data set and the stratified systematic random sample in observed distances to the TCH (P= 0.011), high use paved roads (P= 0.004), high use trails (P< 0.001) and high use features (P = 0.025). Reported results are from the aerial telemetry data set, except where I specifically denote the ground-based data was used.

Differences between the percent cover around telemetry locations (56%) and the random sample (22%) were evident (P< 0.001). While not statistically significant, males tended to use cover more (67%) than females (52%). Again, while no statistical significance was detected, adult males used cover more (68%) than subadult males (56%). Adult (52%) and subadult (58%) females had similar use of cover.

The distribution of distance measurements for most of the four types of human uses was found to be non-normal. However, the similar shapes of the distributions (symmetric to slightly right-skewed) resulted in the variables having similar variances (verified by the Levene's tests of the homogeneity of variance). Results of statistical tests were the same with both the original values and transformed values to normalize the data. Given the robustness of the chosen statistical tool (ANOVA) to the departures from normality and the need to maintain a clarity, I present results of original values. The corrected model for each of the four types of human uses was significant (P < 0.025). Due to model complexity, three and four way interactions were not considered. I also report median distances to various human developments and activities as they provide a better measure of central tendency than means when data are skewed (Table 1).

Table 1. Median distance (m) to human developments measured from aerial telemetry locations of grizzly bears in the Bow River Watershed, Alberta, 1994-1998.