UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Testing habitat suitability models for Roosevelt elk Campbell, Karen Lea 1995

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1995-0324.pdf [ 6.1MB ]
JSON: 831-1.0086768.json
JSON-LD: 831-1.0086768-ld.json
RDF/XML (Pretty): 831-1.0086768-rdf.xml
RDF/JSON: 831-1.0086768-rdf.json
Turtle: 831-1.0086768-turtle.txt
N-Triples: 831-1.0086768-rdf-ntriples.txt
Original Record: 831-1.0086768-source.json
Full Text

Full Text

TESTING HABITAT SUITABILITY MODELS FOR ROOSEVELT ELK by KAREN LEA CAMPBELL B.Sc. The University of British Columbia, 1991 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Animal Science) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA June 1995 © Karen Lea Campbell, 1995 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Ar\iVY\oJ S>CA&4\C^ The University of British Columbia Vancouver, Canada Date tJLon^ , 3 ° . l°l0i^> DE-6 (2788) II ABSTRACT Seasonal habitat suitability models for Roosevelt elk, Cervus elaphus roosevelti, (Brunt 1991) were tested using two distinctly different study areas on northern Vancouver Island, British Columbia. Locations for 5 radio-collared elk, obtained approximately twice per week by telemetry from May 1992 to August 1993, were used to test the models. I used the adaptive kernel method to estimate seasonal home ranges from each elk's locations. Habitat suitability values for summer, mild winter, and severe winter were calculated across each study area on a geographic information system (GIS), using input variables from forest cover, topology, and understory coverages. I compared the suitability values of elk locations to values within home ranges and across study areas, and the suitability values of seasonal home ranges to those across the study areas. The home ranges were further compared with equal sized, circular areas randomly placed in the study areas. Because elk generally used areas of higher suitability than expected, I concluded that the model had some ability to predict areas which contained suitable elk habitat. I also identified limitations of the model. iii Table of Contents ABSTRACT ii Table of Contents iii List of Tables v List of Figures vi Acknowledgements vii INTRODUCTION 1 METHODS 4 Study Areas 4 Benson 4 Nahwitti 4 Elk 6 Radio Telemetry 6 Home Ranges 8 Geographic Information System (GIS) 9 Model 10 Forage 10 Cover 11 Interspersion 14 Aspect/Elevation 15 Habitat Suitability Calculations 16 Validation of Brunt's Models 17 Use vs. Availability, Level 2 Selection 20 1) Locations vs. Study Area 20 2) Home Ranges vs. Study Area 20 3) Home Ranges vs. Random Circles... 20 Use within the Home Range, Level 3 selection 22 RESULTS 23 Population Estimates 23 General Habitats and Movements of Elk 23 Telemetry Accuracy 24 Seasonal Ranges 24 Seasonal Models 24 Tests of the Model 28 Level 2 selection 28 1) Locations vs. Study Area 28 iv 2) Home Ranges vs. Study Area 29 3) Home Ranges vs. Random Circles 29 Level 3 selection 35 Locations vs. Seasonal Ranges 35 DISCUSSION 38 Sample Size 38 Independence 38 Telemetry Error 39 Seasonal Ranges 40 Model Validation 41 Problems with the Models 46 Management Implications 48 LITERATURE CITED 51 Appendix 1. Understory-based habitat types 56 Appendix 2. Common and scientific names of plant species listed 58 List of Tables v Table 1. Total number of radio locations for each collared adult female Roosevelt elk between May 1992 and August 1993 .6 Table 2. Cover and potential forage suitability values by habitat type (Brunt 1991, Lewis, unpubl.) 11 Table 3. Forage modifier values from Brunt (1991) 13 Table 4. Interspersion modifier values from Brunt (1991) 14 Table 5. Aspect/elevation modifier values used in the mild and severe winter models from Brunt (1991) 15 Table 6. Sizes (ha) of study areas and seasonal 95% ranges used by collared elk 24 Table 7. Results of the HSI summer and mild winter models for the 2 study areas (the overall availability in the study areas) 25 Table 8. Results of chi-square tests for locations vs. study area. Proportions and total number of observed locations are included. Expected locations = total locations X proportions of study area in HSI class (from Table 7) 28 Table 9: Comparison of mean HSI for seasonal ranges and grand mean of all corresponding circles. (P = probability of finding circles with a higher mean HSI than the seasonal range) 34 Table 10: Results of chi-square tests of locations vs. seasonal ranges for each animal/season. (Observed locations are in Table 9. Expected locations = total locations X proportions of range in HSI class from Figures 8 and 9) 35 vi List of Figures Figure 1. Location of study areas on northern Vancouver Island, B.C 5 Figure 2. Diagram of input variables for Brunt's (1991) summer suitability model. The thickness of the lines relates to the relative input of the variable 18 Figure 3. Expected difference between the proportion of habitats used by elk and that of the expected use if elk are using habitat classes in proportion to their availability. 19 Figure 4. Available areas in each HSI class for the Benson study area from the summer, mild winter, and severe winter models 26 Figure 5. Available areas in each HSI class for the Nahwitti study area from the summer, mild winter, and severe winter models 27 Figure 6. Examples of seasonal ranges on suitability maps for (a) summer and (b) mild winter (Benson) 30 Figure 7. Examples of seasonal ranges on suitability maps for (a) summer and (b) mild winter (Nahwitti) 31 Figure 8. Used (seasonal range) and available (study area) areas in each HSI class for the Benson study area. Numbers in graph titles are elk ID'S. VL=very low; L=low; M=medium; H=high; VH=very high 32 Figure 9. Used (seasonal range) and available (study area) areas in each HSI class for the Nahwitti study area. Numbers in graph titles are elk ID'S. VL=very low; L=low; M=medium; H=high; VH=very high 33 Figure 10. Illustrative examples of elk use within seasonal ranges in (a) summer and (b) mild winter (Benson) 36 Figure 11. Illustrative examples of elk use within seasonal ranges in (a) summer and (b) mild winter (Nahwitti) 37 Figure 12. Mean HSI values of elk #441's summer 1992 adaptive kernel and MCP ranges (black) and corresponding random circles (white) 43 Acknowledgements I would like to thank Dr. David Shackleton, my supervisor, and Steve Wilson for endless hours of discussion and advice, as well as Drs. Michael Pitt and Tom Sullivan of UBC, and Doug Janz of the Ministry of Environment, Lands and Parks for participating in my supervisory committee. Jerry Maedel, Jennifer Morrison, Pierre Vernier, and Steve Wilson provided help with the GIS component, and Doug Janz and Kim Brunt provided advice throughout the study. I would like to thank Terry Lewis for his advice, instruction, and continued support with the habitat information. Thanks to Rob Dreimal, Roger Briscoe, Marco Passeri, John Foster and other MacMillan Bloedel personnel for their help in locating the elk. I would also like to thank Cindy Fox, Kerry McGourlick, Bill Dumont, and John Foster for making me feel welcome in Port McNeill, and Ernie Springer for keeping my truck going. I would like to thank Marco Passeri for his friendship, love and support throughout the study, and his companionship outside the study. I would also like to thank my parents, for without their love and faith in me, none of this would have happened. This research was supported by a Science Council of British Columbia G.R.E.A.T. award, Western Forest Products, MacMillan Bloedel, Richmond Plywood, the North Island Sportsmen's Association, the Vancouver Island Wildlife Enhancement Fund, and the Habitat Conservation Fund. 1 INTRODUCTION In British Columbia, forest management occurs throughout the province and often conflicts with wildlife management. Such is the case with Roosevelt elk (Cervus elaphus roosevelti) on Vancouver Island, British Columbia, whose preferred habitat on rich valley floors also includes the most valuable timber. Although elk bring far less to the economy than do wood products, they are still a major economic benefit for the province. The total monetary value of elk in B.C. in 1993-94, including actual expenditures and willingness to pay by both resident and non-resident hunters, was $22.7 million in 1994 dollars (D. Janz, pers. comm.). In addition, many people enjoy the chance to see or photograph these and other wild animals. Herd sizes are monitored by the B.C. Ministry of Environment, Lands and Parks (MOELP) to relate population response to environmental and habitat conditions, and to set quotas for limited entry hunting on the Island. Both elk and forestry are important to British Columbians, and the need for elk habitat should be considered in logging plans. Habitat models could provide the necessary information to help forest managers maintain adequate elk habitat so that elk and forestry may coexist. Habitats used by elk in particular areas have been determined by radio-telemetry studies (Witmer and deCalesta 1983, Marcum and Scott 1985, Edge et al. 1987, Brunt etal. 1989, Edge and Marcum 1989, Merrill 1991, McCorquodale 1993), by examining elk sign (Schwartz and Mitchell 1945; Marcum and Scott 1985; Edge and Marcum 1989), or by observing tame (Irwin and Peek 1983; Parker et al. 1984) or wild (Schwartz and Mitchell 1945; Knight 1970; Hanley 1984) elk. However, all these methods are labour intensive and site-specific. For this reason, Brunt (1991) 2 developed habitat suitability index (HSI) models for Roosevelt elk on Vancouver Island. Theoretically, these models could be used anywhere on Vancouver Island to predict how suitable the habitat is for elk. The development of HSI models began in the mid-1970s so that wildlife managers could determine which species and habitats were present on lands within their jurisdictions and make more informed decisions about land use (Berry 1986). HSI and other wildlife-habitat models have been developed for species which are managed for hunting, such as black bear Ursus americanus (Clark et al. 1993), brown bear Ursus arctos (Schoen et al. 1992), and Columbian black-tailed deer Odocoileus hemionus columbianus (Eng and Schieck 1992). Other HSI models have been developed to manage endangered species such as the northern spotted owl Strix occidentalis caurina (Laymon and Reid 1986), and to predict how habitat will be affected by resource industries such as mining (Lancia etal. 1986). The HSI models for Roosevelt elk were first tested by Brunt (1991) in the southern part of Vancouver Island using newly transplanted migratory elk (low elevation in winter, and high elevation in summer), and the summer model was investigated in central Vancouver Island with non-migratory elk (Sovka 1993). Brunt's (1991) transplanted elk would not be familiar with their surroundings; therefore, they may not have selected the optimum habitat, thus affecting the results of model testing. Further testing was needed in both summer and winter, in different areas than where the models were developed, and with established elk herds. My study was designed to test both the summer and winter models on the northern end of the Island (the limit of the subspecies' natural range) in 2 areas with contrasting topography, and where the elk 3 were believed to be non-migratory. I believe that these 3 studies were sufficient to validate (or invalidate) the models for Vancouver Island because they include elk with both migration strategies and they sample areas from different parts of the island. Validation of a model tests its ability to predict states or events (Marcot et al. 1983, Bunnell 1989). Ideally, it should be validated with new data (Lancia et al. 1982) to assess model performance and reliability (Berry 1986, Laymon and Barrett 1986, Schamberger and O'Neill 1986, Bunnell 1989). Although models can improve understanding or predictive abilities, they are still incomplete pictures of reality (Bunnell 1973) because mathematical equations cannot explain all the variables in the real world. Thus, in model testing we merely look at the model's validity - its outcome. There is no way to test for the model's veracity - whether the entire model is true. The HSI value is determined from an assessment of the physical and biological attributes of the habitat in a patch and is assumed to be proportional to the patch's carrying capacity for an animal species (Berry 1986). With large animals such as elk, their home ranges generally include more than 1 habitat patch, so carrying capacity of small individual habitat patches cannot be measured. The HSI value given to a habitat patch thus represents the expected value of that patch to the animal, relative to other habitat types. Animals should select higher quality habitat whenever possible to increase their fitness (Schamberger and O'Neill 1986). To obtain an indication of habitat use by elk, I examined the locations and home ranges of radio-collared elk and compared their HSI values to those of the available habitat types in the study areas. The null hypotheses for the validation tests stated that the HSI values of areas used by elk did not differ significantly from those of the available area. I investigated Johnson's (1980) second and third order of selection (i.e. the area selected by the elk for their home range, and patterns of use within that home range). 4 METHODS Study Areas Benson The Benson River valley lies on northern Vancouver Island about 40 km south of Port McNeill (Figure 1). My study area extended from the headwaters of the river to Kathleen Lake, bounded approximately by the height of land on either side of the valley. The elevation rises from 80 m to 1400 m above sea level. The valley lies within the Very Wet Maritime Coastal Western Hemlock Zone (CWH) below 800-900 m, and within the Mountain Hemlock Zone (MH) at higher elevations (T. Lewis pers. comm.). Logging has occured in the valley for half a century, with substantial harvesting activity in the past 2 decades. Part of the timber on the valley bottom has been set aside as Elk Winter Range, and the remaining area is a mosaic of clearcuts, second growth, mature timber, and old growth. There is an extensive network of logging roads. Nahwitti The Nahwitti River valley lies in the Very Wet Maritime CWH zone, on the north coast of Vancouver Island about 90 km west of Port Hardy (Figure 1). My study area extended from Nahwitti Lake to the ocean, and into the plateau on either side of the valley at least half way to the next major valley. The valley is surrounded by the low hills of the Nahwitti plateau, with elevation rising from sea level to 660 m. Logging has 5 occurred mostly in the past 15 years, and road access is limited. Here also, a section of old growth has been set aside as Elk Winter Range. Figure 1 Location of study areas on northern Vancouver Island, B.C. Elk Roosevelt elk are a popular game species and hunting in both valleys is managed by MOELP. Wolf predation in the Nahwitti valley may have depleted the number of Roosevelt elk. When wolf numbers declined and access due to logging increased, hunting was suspended by MOELP to allow the population to recover. In the Benson, I estimated the population from an aerial census. I used maximum age-sex class counts from sightings to estimate the population in the Nahwitti, where I believe the entire population sometimes came together in one group. Radio Telemetry MOELP personnel darted 3 adult (> 3 years) female elk in the Benson from a helicopter and fitted them with Telonics (Telonics Telemetry-Electronics Consultants, Mesa, Arizona) or Lotek (Lotek Engineering, Newmarket, Ontario) radio collars. The Nahwitti elk were captured in a corral trap and 2 adult females were collared, also by MOELP. I used a Telonics Yagi antenna and receiver to locate animals approximately twice per week from May 1992 to August 1993 (Table 1). Beyer and Haufler (1994) Table 1: Total number of radio locations for each collared adult female Roosevelt elk between May 1992 and August 1993. Study Area ID# # Locations Benson 241 150 Benson 249 148 Benson 849 79 * Nahwitti 441 132 Nahwitti 461 127 * elk #849 died in January 1993. 7 suggest varying the time of day (or night) for locations; however, for safety reasons I was constrained primarily to hours when MacMillan Bloedel or Western Forest Products (WFP) personnel were in radio contact, or to times at night when I could be accompanied by someone else. I used the loudest signal method (Springer 1979) to determine the direction to take a compass bearing. The loudest signal, however, usually came from an arc of < 30°, so the bearing was taken at its centre. Triangulation, with a minimum of 3 a maximum of 10 bearings, was used to determine an animal's location. The time between each bearing was kept short to reduce error due to animal movement (Schmutz and White 1990; Saltz 1994). I continued to take bearings, and draw them on a map in the field, until I had at least 3 strong signals with bearings which converged on an area approximately < 2 ha. The location was taken as the approximate centre of this area, with more weight given to bearings with stronger signals, and less to probable outliers. An estimate of triangulation precision (error in the locations) was made using the average distance from the mapped point to the known location of all collared animals sighted just after triangulations were completed. I expected the true location of the animal to be within a circle of that radius, centred on the triangulated point. For 41 locations in the Benson and 63 in the Nahwitti, the collared elk were sighted before triangulation was complete, and the locations were drawn on the field map as points. Errors in these locations were from 2 sources: my ability to place the point on the map, and any inaccuracies in the map itself. 8 The locations were categorized as summer 1992, winter, or summer 1993. I used a combination of migration times from other Roosevelt elk herds on the Island, and plant phenology to determine the shifts between seasons. Summer began on April 15, and winter began on November 15. Home Ranges The home range of an animal is the area traversed by that individual in its normal activities of food gathering, mating, and caring for young (Burt 1943). Several methods of calculating the home range from a set of animal locations exist. I compared the adaptive kernel method (Worton 1989) in the program KERNELHR (Seaman and Powell 1991) to the harmonic mean method (Dixon and Chapman 1980) in Program Home Range (Ackerman et al. 1990). I believe the home ranges that I calculated using the adaptive kernal method fit the data and the topography better than did those using the harmonic mean method (which included high elevation areas that elk never were observed to use). The adaptive kernel method is a robust estimator of the utilization distribution because it does not assume any underlying distribution, and it is also very good at estimating areas of concentrated activity (Worton 1987). Furthermore, Program Home Range required input of an arbitrary smoothing constant which affected the resulting home range size (D. Sovka, pers. comm.). In KERNELHR, the optimal smoothing constant was chosen automatically by cross validation (Worton 1989), which provides an obvious advantage over the harmonic mean method (Larkin and Halkin 1994). I used KERNELHR to calculate home ranges from the animal locations for each individual, in each season. I avoided pooling locations across individuals (Aebischeref 9 al. 1993) and seasons (Schooley 1994) so that behavioural variation among individuals and annual variation among seasons would not be lost. Programs written by F. Hovey (SFU) were used to import the resulting density probability grid into Stanford Graphics 2.1 as a matrix, and to provide the density levels for 95% contours. These contours enclosed an area within which the probability of locating the animal was 95%, and I considered this to be the home range. The contours were saved as polygons in text files (DXF) readable by a geographic information system. Elk #849 died at the beginning of January, after only 14 winter locations. I included the last 6 summer locations in the home range calculation because the minimum number of locations for the adaptive kernel method was 20. Therefore, results for elk #849's winter home range should br interpreted conservatively. Geographic Information System (GIS) To analyze the spatial data, I used the GIS Terrasoft 10 (Digital Resource Systems Limited, Nanaimo, B.C.) and Idrisi 4.0 (Clark University Graduate School of Geography, Worcester, Massachusetts). For the Benson and Nahwitti, topographic features including roads, rivers, streams, lakes, and contours were available on 1:20 000 digital "Terrain Resource Information Management" (TRIM) maps from MOELP. Dr. Terrence Lewis (Consulting in Soils and Land Use, Courtenay, BC) created understory based habitat maps of the study areas from 1:15 000 aerial photographs and site visits. For the Benson, I digitized the map into the GIS, and for the Nahwitti, Lewis' photo-interpretation linework was digitally mapped with a stereo analytic plotter (Darren Ham, Dept. of Geography, UBC) and entered into the GIS. MacMillan Bloedel and MOELP provided digital forest cover maps for the Benson and Nahwitti 10 respectively, both of which I updated to 1993 and adjusted to the same coordinate system (UTM NAD83) as the TRIM map using The Geographic Calculator 2.0 (Blue Marble Geographies, Gardiner, Maine). Finally, I digitized elk locations and home ranges. To calculate the suitability values, I used raster (grid cell based) rather than vector (line based) analysis for 4 reasons: 1) the accuracy of the elk locations was about 30 to 40 m, and with my chosen pixel size of 20 m, no accuracy was lost; 2) the accuracy of the habitat and forest cover maps after being merged with the TRIM map was probably no better than 20 m; 3) two of the variables (aspect and elevation) originated as raster layers; and 4) it is a much faster method, requiring less disk space. Model Separate model algorithms were created for summer, mild winter, and severe winter. Basically, the models use 5 habitat factors (forage, cover, interspersion, aspect and elevation) as input variables, the last 2 of which are used as modifiers in winter. Forage Preferred elk forage (Brunt 1990) consists of a combination of trees (the most important species are western hemlock and amabilis fir), shrubs (e.g., dull oregon grape, Pacific ninebark, red elderberry, huckleberry/blueberry, willow, salmonberry), ferns (e.g., deer and sword), and herbs (e.g., bunchberry, twinflower, sedges, and grasses) [Latin names for plants are given in Appendix 2]. Season dictates the 11 proportions of each group found in the diet. For example, conifers are nearly absent from the summer diet, but considered to make up almost half the diet in winter (Brunt 1990). Forage values (Table 2) varying between 0.0 and 1.0 (with 1 being ideal) were based on habitat types, and then modified by overstory conditions. The modifier values (Table 3) are an index of canopy closure. Cover Elk use cover for at least 3 reasons: to avoid extreme temperatures, to take advantage of lesser snow accumulations in winter, and to avoid predators or human harassment. Brunt (1991) combined snow interception cover with thermal cover, and I will refer to the combination as thermal cover. Because cover is either present or absent, cover values (Table 2) were either 0.99 or 0.0 (the value 1 would cause division by zero in the algorithm). Brunt (1991) assumed that thermal cover was adequate in forested stands >10 m in height and with >70% mean canopy closure, while security cover was adequate in any stand >3 m in height. I used age of a forest stand to estimate tree height and canopy closure; security cover values in those stands < 8 years old, and thermal cover values in stands < 20 years old were multiplied by 0.0 to indicate the absence of cover. The habitat mapping done by T. Lewis identified more habitat types than did Brunt's understory mapping. Lewis correlated his system to Brunt's where possible, and estimated the forage and cover values for those habitat types not listed by Brunt (see Table 2). One discrepancy between the two systems is that whereas Brunt uses 12 Table 2. Cover and potential forage suitability values by habitat type (Brunt 1991, Lewis, unpubl.). See Appendix 1 for explanation of habitat types. Lewis' Habitat Type Brunt's Primary Understory A * B c D S11 Bog/Wetland 0.00 0.00 0.99 0.90 S1 H A , MH1, M1, M1c, M1 s Huckleberry-Moss 0.99 0.99 0.20 0.30 P Lichen-Pink mountain heather 0.99 0.00 0.10 0.20 S2F,S12F Lichen-Salal 0.99 0.00 0.20 0.10 M3,M3C,M5 Rosy twistedstalk-5 leaved bramble 0.99 0.00 0.30 0.30 S1C H,S2,S12, Salal-Huckleberry 0.99 0.00 0.40 0.20 S10,S12CH,M2 S3,S3B,S4 Salmonberry 0.99 0.00 0.80 0.80 A Slide Complex 0.00 0.00 0.50 0.99 S1C,S13,S13L Sword Fern 0.99 0.99 0.80 0.70 S5.S7 ** Skunk Cabbage 0.99 0.00 0.75 0.80 S6 Skunk Cabbage 0.99 0.00 0.40 0.60 M4 Skunk Cabbage 0.99 0.00 0.10 0.60 MH2 - 0.00 0.00 0.14 0.20 AT - 0.00 0.00 0.10 0.20 S8 Sphagnum [with Labrador tea] 0.00 0.00 0.25 0.50 S9 Sphagnum [treeless] 0.00 0.00 0.90 0.90 S15 Deer Fern 0.99 0.99 0.50 0.60 CP Sphagnum-deer fern 0.99 0.00 0.70 0.75 W - 0.00 0.00 0.40 0.50 R,NV, Talus - 0.00 0.00 0.00 0.00 * A=Security Cover, B=Thermal/Snow Interception Cover, C=Potential Winter Forage, D=Potential Summer Forage ** Forage and cover values for remaining habitats, which were not given by Brunt (1991), were estimated by T. Lewis. Table 3: Forage modifier values from Brunt (1991). 13 HABITAT FORAGE MODIFIER VALUE Logged < 2 years previously 0.40 Logged 3 -5 years previously 0.75 Logged 6-15 years previously 1.00 Logged 16-20 years previously1 0.50 Logged 21-50 years previously1 0.10 Logged > 50 years previously1 see unlogged Unlogged - deciduous dominated overstory2 0.75 Unlogged - conifer dominated overstory3 0.50 Bog/Wetland, Rock outcrop, and Slide Complex 1.00 1From Brunt (pers. comm.). 2Deer fern, Salmonberry, and Swordfern understory types. 3AII other understory types with the exception of Bog/Wetlands, Rock outcrops, and Slide complexes. elevation/aspect modifiers to identify snowpack zones, Lewis used montane (M) areas to make this distinction. In some cases, Lewis' classification would more accurately identify certain habitats. For example, snow accumulation zones and areas with regular thermal inversions are distinguished as montane by Lewis, whereas Brunt,who uses elevation to identify montane habitats, might not separate these areas from the surrounding sub-montane habitat (T. Lewis, pers. comm.). For polygons which contained combinations of habitats listed in Table 2, I calculated mean forage values, taking into account the proportion of each habitat type in the polygon. Lewis distinguished 3 levels of heterogeneity by his classification 14 method; for example, the label S3S5 was a 50:50 ratio of the 2 habitat types S3 and S5, S3/S5 was a 65:35 ratio, and S3//S5 was an 85:15 ratio. I took cover values from the major habitat type (the first type listed in the label). Interspersion The interspersion of forage and cover appears to be critical to elk (Thomas et al. 1979; Skovlin 1982; Witmer et al. 1985; Brunt 1990). While the best forage grows in areas with little or no canopy cover (natural openings, clear cuts, and young forests), the most valuable cover occurs when the canopy closes (mature forests or old growth). To take advantage of the best forage and the best cover, elk would need to live along a boundary between different aged forests (Skovlin 1982). Elk will use a good forage area if it is close (< 140 m) to cover, while forage located far (> 300 m) from cover is almost never used (Brunt 1991). Interspersing different types or ages of forests increases the number and types of these boundaries, thus elk use would become more extensive in such areas. Interspersion values are determined for the model using the distance to high quality cover (0.99) or to forage (> 0.50) areas (Table 4). I created the areas in each category by placing 140, 250 and 300 m buffers around high quality cover and forage areas in the GIS. Aspect/Elevation During winter, the depth and persistence of snow affects the winter suitability values because it buries forage and increases locomotion costs at a time when food is 15 Table 4. Interspersion modifier values from Brunt (1991). DISTANCE FROM COVER OR PREFERRED FORAGE AREAS (m) MODIFIER VALUE 0* 1.0 <140 1.0 141 -249 0.6 250 - 300 0.4 >300 0.1/0.01** * Site qualifies as cover or a preferred forage area **0.1 for cover; 0.01 for food limited and energy demands are high (Parker et al. 1984, Brunt 1991). Aspect and elevation combine to give a good indication of snow conditions. Snow tends to be deeper and more persistent on north slopes at high elevations than on south slopes at lower elevations. Using the GIS, I created elevation and aspect raster layers with 20 x 20 m pixel resolution from contours and used these to assign modifier values (Table 5). Table 5. Aspect/elevation modifier values used in the mild and severe winter models from Brunt (1991). ELEVATION (m) ASPECT 290-70° (NORTH) 71-110° (EAST) 110-250° (SOUTH) 250-290° (WEST) FLAT 0-350 0.6 0.8 10 0.8 1.0 351-550 0.4 0.6 0.8 0.6 0.8 551-1050 0.2 0.4 0.6 0.4 0.6 >1050 0 0 0 0 0 16 Habitat Suitability Calculations Using the same (20 x 20 m) pixels as the slope and aspect calculations, I created the following raster layers: SFOR = summer forage value (modified for logging) WFOR = winter forage value (modified for logging) TCOV = thermal cover value (modified for logging) SCOV = security cover value (modified for logging) SUMDIST = interspersion value: distance to summer forage of value > 0.5 WINDIST = interspersion value: distance to winter forage of value > 0.5 TDIST = interspersion value: distance to thermal/snow interception cover SDIST = interspersion value: distance to security cover ASP_ELEV = aspect/elevation modifier value Brunt's (1991) model provides an HSI valuefor each pixel within a habitat patch. Each patch (i.e. an area which contains a homogeneous combination of mapped variables) is rated on a scale from 0.0 to 1.0 based on its estimated value for elk. I used Brunt's (1991) algorithms (below) to calculate the suitability values for each pixel, which I then put into a new raster layer for each seasonal model. The 3 seasonal models (Brunt 1991) are: Summer Habitat Suitability = exp(0.5//?(1-exp(0.7/n(1-SFOR) + 0.15/n(1-TCOV) + (0.15//7(1-SCOV)))) + 0.25//7(SUMDIST) + 0.125/n(TDIST) + 0.125/A7(SDIST)) Mild Winter Habitat Suitability = exp(0.5/A7(1-exp(0.7/n(1-WFOR) + ((0.2/n(1-TCOV)) + (0.1 //7(1 -SCOV))))) + 0.25/A?(WINDIST) + 0.125/n(TDIST) + 0.125/n(SDIST)) x ASP_ELEV 0 2 5 Severe Winter Habitat Suitability = exp(0.5/n(1-exp(0.5/n(1-WFOR) + ((0.4/n(1-TCOV)) + (0.1//7(1-SCOV))))) + 0.20//?(WINDIST) + 0.20/n(TDIST) + 0.1/n(SDIST)) x A S P _ E L E V 1 0 Figure 2 illustrates how each of the input variables contribute to the summer suitability model (winter models are the same except for the final multiplication of the aspect/elevation modifier). 17 The raster maps of habitat suitability values consisted of pixels, each containing a suitability value (between 0.0 and 1.0). These values were used to calculate mean suitability values of areas, or to represent the suitability value at an elk location which fell within a pixel. For some tests, suitability values were grouped into 1 of 5 classes: 0.00 - 0.20 (Very Low) 0.21 - 0.40 (Low) 0.41 - 0.60 (Medium) 0.61 - 0.80 (High) 0.81-1.00 (Very High) Using the original continuous data, or choosing a larger number of classes, would have made the results very difficult to interpret. As the number of categories increases, the frequency of locations within each category decreases and it becomes more difficult to resolve differences in the animals' behaviour. However, using too few classes would allow a very large difference in quality within the same class. An odd number of classes makes interpretation clearer by providing a "middle" category, and I chose a compromise of 5 categories for comparative purposes. Validation of Brunt's Models To validate Brunt's models, I investigated elk use and the predicted HSI values of the study areas. Assuming that the study area is large relative to the home range, a model is validated if predicted high suitability habitats are used by elk proportionally more often than they occur in the study area. Because the winter of 1992-93 was mild, I have no data to test the severe winter model, so I attempted to validate only the summer and mild winter models. I made the assumption that elk will use areas of higher quality whenever possible; therefore, I expected elk to spend more time in 18 Thermal Cover Security Cover Logging t Logging Distance to G o o d The rma l Cover Distance to G o o d Security Cove r Suitability Value Figure 2: Diagram of input variables for Brunt's (1991) summer suitability model, thickness of the lines relates to the relative input of the variable. The 19 habitats with higher predicted HSI values than in those with lower ones. Because the 5 suitability classes are not necessarily available in the same amount, I would not always expect increasing use from the lowest to the highest class. However, the difference between the proportion of available habitat in each class and the proportion of time spent in each class should increase from negative to positive as the suitability moves from the lower classes to the higher classes (Figure 3). Suitability Class Figure 3: Expected difference between the proportion of habitats used by elk and that of the expected use if elk are using habitat classes in proportion to their availability. In an attempt to validate Brunt's seasonal models, I employed 3 methods of testing use vs. availability at Johnson's (1980) level 2 selection (locations vs. study area, home ranges vs. study area and home ranges vs. random circles) and 1 method at Johnson's (1980) level 3 selection (locations vs. home ranges). 20 Use vs. Availability - Level 2 Selection For each season and each individual, I compared the suitability values of elk use (both locations and home ranges) to what was available in the study area. Habitat availability was estimated to be the total proportion of each habitat suitability class in the study area (calculated with the GIS). 1) Locations vs. Study Area For the elk locations, a chi-square test was used to test the null hypothesis: proportional use of suitability classes by elk does not differ from the proportions available in the study area. Expected locations were calculated as the percentage of the study area within each class, multiplied by the total number of elk locations per individual per season. Some classes were grouped together to ensure that no expected frequency was < 1.0, and no more than 20% of the expected frequencies were < 5.0 (Cochran 1954). 2) Home Ranges vs. Study Area The chi-square test cannot be performed with home ranges because they are measured in units of area, not frequency data which the test requires. Therefore, to compare home ranges to the study area, I used only graphs to illustrate the habitat classes used by elk and those available in the study area. 3) Home Ranges vs. Random Circles The previous test estimated availability from the proportions that each habitat type (i.e. suitability class) made up of the entire study area. Another method is to 21 estimate availability from random dots on a map of the study area (Marcum and Loftsgaarden 1980). However, neither of these approaches takes into account the spatial relationships among the habitats. Both assume an animal is not restricted from moving anywhere within the study area to gain resources. Not only is the total amount (area) of habitat important for survival and reproduction, but so is the dispersion, both absolute (throughout the study area) and relative (with respect to other classes of habitat), of the valuable habitats. Suitable habitat for non-migratory elk must meet their year-round needs. It should also be continuous, or in patches which are relatively close so that they benefit (through a net gain in energy or decreased chance of predation) by using them. The farther apart valuable habitat patches are, the less useful they are to an animal because of energetic costs of locomotion and predation risks while travelling among them. To allow a more realistic assessment of habitat availability for elk, I placed circles of equal area to each elk's 95% home range, randomly throughout the study area to represent a random selection of seasonal ranges. A circle was chosen as the shape to place around randomly selected coordinates on the map, eliminating a source of bias due to orientation. I compared an observed elk's seasonal range with a random sample of 50 equal-sized circles within the study area. For each individual's seasonal range (summer 1992, winter, summer 1993), and for each random circle, I calculated a mean HSI value. I also calculated a grand mean and standard error of all circle means, and ranked the seasonal range means along with the circle means. From the rankings, I obtained a probability estimate of potential home ranges (i.e. circles) which have lower means than the one chosen by the elk. 22 Assuming high quality habitat patches are smaller than the home range and are heterogeneously distributed (Wilson et al., in prep.), I would expect this probability to be high (i.e., > 80%) if the model is valid. This procedure is similar to a Monte Carlo simulation, except that only 50 home ranges were simulated (1000 is suggested as a minimum; Manly 1991, Wilson et al., in prep.). While I would have preferred to do 1000 simulations for each individual/season, it was not feasible due to computer time constraints. I compared runs of 25, 50, 75, and 100 circles, and only the 25 circles differed; therefore, I chose to use 50 as a minimum estimate. Use within the Home Range - Level 3 selection While the previous tests examined Johnson's (1980) level 2 selection (choice of a home range from within a larger available area), it is useful also to examine level 3 -selection of habitats within the home range (Aebischer et al. 1993, Carroll et al. 1995). To do this, I compared HSI values at locations used by elk to those available in the seasonal range. For each individual in each season, I performed a chi-square test with the observed value being the number of elk locations in each suitability class, and the expected being the proportion of each suitability class in the seasonal range multiplied by the total number of locations. This tested the null hypothesis: proportional use of suitability classes by elk does not differ from the proportions available in the seasonal range. As before, I expected the actual sites used by elk to have higher suitability values than seasonal ranges because the latter may contain areas not actually used by elk. RESULTS 23 Population Estimates Estimates of the total population of the Benson valley have been as high as 80 animals in 1991 (D. Janz, pers. comm.). However, aerial counts made on 14 January 1993 indicated at least 41 elk, found singly or in groups of 2, 8, 13, or 15, were present in the study area. Estimates of the population in the Nahwitti study area were 13 in 1990, 19 in 1991 (Rick Davidge, pers. comm.), 25 in 1992, and 34 in 1993 (my estimates based on maximum age-sex counts). General Habitats and Movements of Elk In general, elk were found in riparian areas and wetlands, as well as in clearcuts, young regenerating stands, and old growth on valley bottoms and lower slopes. There was no seasonal migration by collared animals, although logging truck drivers in the Benson reported seeing elk in higher elevation areas adjacent to the Benson valley only in summer. While snow was present in winter, the collared elk used clearcut or relatively open slopes at slightly higher elevations than they used throughout the rest of the year. The collared elk were found mostly in groups, that ranged in size from 2 to 34 individuals. There was some evidence that females moved away from the group to areas of little use to give birth in spring. I identified one calving spot in a small brushy draw on a clearcut sidehill. 24 Telemetry Accuracy To estimate accuracy of the locations I used only sightings of collared elk which occurred immediately after triangulation was complete. From 19 such sightings in the Benson, the location precision was 0.43 ha, and in the Nahwitti, the precision from 5 sightings was 0.28 ha. Seasonal Ranges While the sizes of the 2 study areas were similar, the seasonal ranges in the Nahwitti were larger in general than those in the Benson, especially 3 of the 4 summer ranges (Table 6). Table 6: Sizes (ha) of study areas and seasonal 95% ranges used by collared elk. Benson Season Elk #241 Elk #249 Elk #849 Summer 1992 288 242 1482 Total = 14504 Winter 1992/93 259 253 663 Summer 1993 259 305 N/A Nahwitti Season Elk #441 Elk #461 Summer 1992 3236 2380 Total = 15259 Winter 1992/93 1419 1732 Summer 1993 961 4147 Seasonal Models Compared to the Benson, the Nahwitti contained very little of the low suitability classes (tan and brown in Figures 4 and 5). The areas of high suitability (green areas) 25 in the Benson and Nahwitti are aggregated mostly on the valley bottoms near the rivers. In the Benson, the class covering the largest area is the 0.21 - 0.40 class in the summer model, and the 0.00 - 0.20 class in the mild winter model. In the Nahwitti, the 0.41 - 0.60 class covers the largest area for both seasonal models (Table 7). In both study areas, the predicted suitability of the study area, in general, decreased from summer to mild winter. The severe winter model was not tested; however, it is interesting to note the decrease in higher suitability areas from the mild to severe winter maps (Figures 4 and 5) since it is likely the amount of high quality severe winter habitat which limits a population in the long term. Table 7: Availability of suitability classes for the 2 study areas as determined by the HSI summer and mild winter models. Suitability Class Nahwitti Benson Summer Mild Winter Summer Mild Winter Very Low 2410(16.6)* 3117 (21.5) 3195 (20.9) 8350 (54.7) Low 1499(10.3) 1856(12.8) 8184 (53.6) 3643 (23.8) Medium 7000 (48.3) 7369 (50.8) 586 (3.8) 1210 (7.9) High 2706(18.7) 1743(12.0) 1815(11.9) 1180 (7.7) Very High 785 (5.4) 315 (2.2) 1130 (7.4) 528 (3.5) Water 104 (0.7) 104 (0.7) 349 (2.3) 349 (2.3) Total 14504 15259 *area in hectares (and percentage of total area). T I CO c —* CD X*. c f < CD Q)CO c 3 3 CD 0) CT CD CD —\ CD fi W CL = 3 $ CD CD ~ * T 0) cn 3 — CL O. (/> 0) < <" CD S Q CD - i CD 9 S3 -> CD 3 o rs d-§ ® CO CO Q. •< 03 "3 CD CD —h —1 o 3 on < o ex 93 LZ 28 Tests of the Model Level 2 selection 1) Locations vs. Study Area For all elk in all seasons in both the Benson and Nahwitti, the distribution of locations over the 5 suitability classes was significantly different (a = 0.05) from what was available in the study area (Table 8). In all cases, the elk tended to use a greater proportion of the higher suitability classes (High and Very High) and a lesser proportion of the lower suitability classes (Very Low and Low) relative to what were available in the study area. Table 8: Results of chi-square tests for locations vs. study area. Proportions and total number of observed locations are included. Expected locations = total locations X proportions of study area in HSI class (from Table 7). Elk ID Season V Low Low Medium High VHigh Total x2 df Benson 241 Sum92 0.03 0.02 0.23 0.47 0.26 66 198 4 Sum93 0.12 0.14 0.14 0.51 0.09 43 45 2 Winter 0.00 0.00 0.26 0.50 0.24 42 173 2 249 Sum92 0.02 0.08 0.23 0.43 0.25 65 96 2 Sum93 0.00 0.00 0.32 0.41 0.27 41 132 2 Winter 0.14 0.12 0.16 0.47 0.12 43 83 2 849 Sum92 0.02 0.32 0.06 0.22 0.38 65 70 2 Winter 0.07 0.21 0.07 0.29 0.36 14 12 1 Nahwitti 441 Sum92 0.09 0.05 0.29 0.42 0.15 55 31 4 Sum93 0.00 0.09 0.26 0.49 0.17 35 33 2 Winter 0.00 0.08 0.49 0.36 0.08 39 33 3 461 Sum92 0.04 0.06 0.25 0.47 0.19 53 51 3 Sum93 0.03 0.06 0.31 0.34 0.25 32 22 2 Winter 0.00 0.05 0.54 0.33 0.08 39 30 3 29 2) Home Ranges vs. Study Area The seasonal ranges in the Benson and the Nahwitti appear to be centred around areas where the predominant colours are green and yellow (the top three suitability classes), whereas they tend to exclude the bottom two classes - tan and brown (Figures 6 and 7). While not as dramatic as with location data, in most cases the elk chose less of the lower quality habitat, and more of the higher quality habitat in both the Benson (Figure 8) and the Nahwitti (Figure 9), relative to what were available. Winter ranges in the Benson included greater proportions of the lower suitability classes than did summer ranges. However, in the Nahwitti, the distribution of suitability class proportions changed very little from summer to winter, and the most frequently used HSI class was medium (0.41 - 0.60). 3) Home Ranges vs. Random Circles In the Benson, all seasonal ranges had mean suitability values > 2 standard errors above the grand mean of all circles (Table 9). All summer ranges were ranked either first or second among the random circles, meaning the probability of finding an equal sized area that has a mean HSI value < that of the summer range was < 0.98. The probability of finding such an area for winter ranges was between 0.74 and 0.88. In the Nahwitti the probability of finding a random circle with a mean HSI value < that of the seasonal range was decreased to between 0.06 and 0.80. Only the 1993 summer ranges had mean suitability values that were higher than the grand mean of the circles, and both were > 2 standard errors from their respective means (Table 9). These results do not show elk seasonal ranges to be different from 3 0 31 CO Q) T3 -4—» CO o CD CD C CO cn 75 c o (/) (0 0) C O Cj c o •c o Q_ O 241, Summer 1992 0.6 T 0.3 4-II VL L M H VH 849, Summer 1992 0.6 0.3 4-I. r -p i , r i • •-• i VL L M H VH 241, Summer 1993 0.6 -r 0.3 4-I VL L M H VH 241, Winter 0.6 -r 0.3 +• I ' 1 1 I *\ , f i l l , r-IJl VL L M H VH 849, Winter 0.6 -r 0.3 + I I ix l i I law | I — W | VL L M H VH 249, Summer 1992 0.6 -r . - 3 - 1 - . i . J l . r i 1, rl, VL L M H VH 249, Summer 1993 0.6 T 0.3 El VL L M H VH 249, Winter 0.6 T 0.3 4-I ' 1 1 I VL L M H VH • Study Area H Seasonal Range 32 Suitability Class Figure 8: Used (seasonal range) and available (study area) areas in each HSI class for the Benson study area. Numbers in graph titles are elk ID's. VL=very low; L=low; M=medium; H=high; VH=very high. 441, Summer 1992 461, Summer 1993 33 CD CD 5K TJ ZJ -«—< CO 0.6 T 0.3 4-f t , rn 1 1 1 1 1 VL L M H VH 0.6 -r 0.3 4-n VL L M H VH O CD cn c CD C£ ro c o to CD CD CO >4— o C o •c o C L o 441, Summer 1993 0.6 T 0.3 +• i • • VL L M H VH 461, Summer 1993 0.6 -r 0.3 + I 1 '"• I i 1 1 1 1 VL L M H VH 441, Winter 461, Winter 0.6 -r 0.3 + VL L M H VH 0.6 -r 0.3 + I i I~I | i I-I | i i"i | i—R | L M H VH VL Suitability Class • Study Area [5] Seasonal Ranges Figure 9: Used (seasonal range) and available (study area) areas in each HSI class for the Nahwitti study area. Numbers in graph titles are elk ID's. VL=very low; L=low; M=medium; H=high; VH=very high. 34 what could be chosen at random from the study area. However, the Nahwitti study area has a higher average suitability than does the Benson. Therefore, grand means of random circles were approximately 0.2 higher in the Nahwitti, which moves them into a higher suitability class. The variation around the grand means was also much lower in the Nahwitti as shown by the standard errors of the means (Table 9). Table 9: Comparison of mean HSI for seasonal ranges and grand mean of all corresponding circles. (P = probability of finding circles with a higher mean HSI than the seasonal range used by elk.) Elk ID# Season Grand Mean SE* (Circles) Mean HSI of P (Circles) Seasonal Range Benson 241 Sum92 0.32 0.013 0.69 0.00 Sum93 0.33 0.017 0.65 0.02 Winter 0.29 0.016 0.44 0.12 249 Sum92 0.31 0.015 0.67 0.00 Sum93 0.33 0.014 0.66 0.00 Winter 0.28 0.017 0.43 0.16 849 Sum92 0.35 0.007 0.52 0.00 Winter 0.29 0.013 0.36 0.26 Nahwitti 441 Sum92 0.53 0.004 0.48 0.94 Sum93 0.52 0.011 0.60 0.20 Winter 0.54 0.008 0.48 0.88 461 Sum92 0.53 0.006 0.52 0.64 Sum93 0.52 0.003 0.54 0.32 Winter 0.52 0.008 0.47 0.80 * standard error of the grand mean of all circle means. 35 Level 3 selection Locations vs. Seasonal Ranges The locations show that even within a seasonal range, elk rarely used lower quality habitat. For example, in Figure 10a, a patch of low quality habitat (tan) juts into the seasonal range from the left, and the locations show that this area was generally unused by the elk. Figures 10b and 11 also show that locations were in higher quality habitats within the seasonal range. In both the Benson and Nahwitti, 11 of 14 chi-square tests (Table 10) showed the proportions of elk locations in each HSI class to be significantly different from those of the corresponding seasonal range (the exceptions were summer 1992 for elk 241 and 249 in the Benson, and summer 1993 for elk 441 in the Nahwitti). In general, there was a greater proportion of the higher suitability classes, and a lesser proportion of the lower suitability classes used by the elk than what were proportionally available in the seasonal ranges. Table 10: Results of chi-square tests of locations vs. seasonal ranges for each animal/season. Observed locations are in Table 8. Expected locations = total locations X proportions of range in HSI class from Figures 8 and 9. Benson Nahwitti Elk ID Season 2 X df Elk ID Season 2 X df 241 Sum92 6.1* 3 441 S '92 21.4 4 Sum93 9.9 3 S '93 4.8* 4 Winter 40.8 4 Win 12.8 4 249 Sum92 2.7* 4 461 S'92 25.0 4 Sum93 9.8 3 S '93 7.4 2 Winter 14.1 4 Win 15.2 4 849 Sum92 16.9 4 Winter 4.2 1 * No significant difference (a = 0.05). Suitability of Habitat for Elk 1 IVfirylnw 1 1 High ~f~ Elk location 1 1 Low | ;| Very High — 9 5 % seasonal range 1 1 Medium H Water boundary Figure 11: Illustrative examples of elk use within seasonal ranges in (a) summer and (b) mild winter (Nahwittti). DISCUSSION 38 Sample Size While the sample size (n = 5 collared elk) seems small, a) I am not inferring anything about elk, in general, from them but using them to test the model's predictive ability, and b) they represented the behaviour of a larger number of elk (12-32) that probably had very similar home ranges. Roosevelt elk live in groups, and through sightings and aerial censuses, I estimated that the whole population in the Nahwitti (except adult males) and 1/3 of the population in the Benson had seasonal ranges that were similar to the collared elk. Independence Researchers often assume that independence is lost when one individual is sampled more than once (Hurlbert 1984) or, in home range calculations, if the time interval between locations is too short (Swihart and Slade 1985; Reynolds and Laundre 1990). In my study, I used each individual as a separate test of the model, so repeated sampling from each individual was necessary. For the chi-square test, locations are assumed to be independent. One view of independence is that the animal has sufficient time between consecutive locations to traverse its home range (Swihart and Slade 1985). I felt that my sampling schedule (locations were generally 2-5 days apart) allowed this, and I recorded elk moving from one end of the home range to the other between consecutive locations. Furthermore, in the chi-square tests, I considered the number of suitability classes (not locations) as the sample size. Hence, increasing the 39 number of locations does not increase the degrees of freedom, but it does increase the accuracy of the estimate of habitat use. While I believe my locations were independent, independence was not essential for the home range calculations. Autocorrelated data collected at short time intervals can still be appropriate for obtaining valid home range estimates, and at the same time maximize the information available about habitat selection (Reynolds and Laundre 1990). By compiling the set of locations, or calculating a home range, I was trying to estimate the total area used by that individual over the duration of the season, which can also be thought of as a trajectory through time and space (Aebischer et al. 1993). In the ideal situation, the animal would be tracked continuously so that its exact route is known, as would the time spent at each point on its trajectory. Radio-locations taken from the ground sample this trajectory. As long as locations are spaced either randomly or relatively regularly in time, the sample can be considered a good estimate of habitats used. Telemetry Error The ad hoc method (Nams and Boutin 1991) of triangulation that I used seemed most appropriate even though no statistical method of assessing error was used. Factors which increase error include increased distance from the animal, topographic or vegetative barriers, atmospheric conditions, and the time between bearings (Harris etal. 1990), as well as intersection angle of the bearings (Saltz 1994). In both areas, the error around each triangulated point was small compared to the average patch size (Benson = 44 ha, Nahwitti = 54 ha) of the most coarsely mapped input variable - the habitat map. Furthermore, the accuracy of the locations was similar to the accuracy of 40 the map layers in the GIS, and I checked each location which was near a boundary in the habitat or forest cover maps to ensure it fell in the proper polygon. For these reasons, I believe the location error was not significant. Seasonal Ranges The average size of seasonal ranges in the Benson area (468 hectares) fell within the range of sizes found in other elk studies: 637 ha for migratory elk on southern Vancouver Island (Brunt 1991) and 290 ha for resident elk in California (Franklin et al. 1975). However, the seasonal ranges of the Nahwitti elk were much larger on average (2,313 ha) than in the Benson, and closer in size to the average range of resident elk in both Sovka's (1993; 3,000 ha) and Brunt's etal.'s (1989; 1,710 ha) studies, and to those reported for Rocky Mountain elk (Cervus elaphus nelsoni; 1,880 and 3,020 ha; Pederson et al. 1980). Such a large difference in seasonal range size between the Benson and Nahwitti could indicate that the Benson has higher quality elk habitat, because home range size tends to increase with decreasing food density (McNab 1963; McCorquodale etal. 1989) and forage is the most influential factor in Brunt's (1991) models. The general habitats selected by all collared elk, including riparian areas, wetlands, clearcuts, young regenerating stands, and old growth on valley bottoms and lower slopes, were similar to those used by elk in previous studies (Schwartz and Mitchell 1945, Pederson etal. 1980, Collins and Urness 1983, Irwin and Peek 1983, Witmerand deCalesta 1983, Brunt etal. 1989, Brunt 1990, Brunt 1991, McCorquodale 1993, Sovka 1993). These habitats tend to provide preferred forage for elk, such as 41 elderberry, salmonberry, huckleberry, thimbleberry, fireweed, forbs and grasses (Troyer 1960, Franklin ef al. 1975, Janz 1983, Brunt 1990, Sovka 1993). Model Validation "The validation process in some way compares model predictions to observations of species' responses (or to prior knowledge), determines the strong and weak points of the model according to specified criteria, and guides further empirical studies or model development and revision" (Marcot et al. 1983:317). This statement identifies each of the important steps in validating an HSI model. Brunt (1991) first used prior knowledge of elk habitat use to develop the seasonal models, then he and Sovka (1993) compared model predictions to elk locations and home ranges to determine that the models have some predictive capability. My results also validate the models. Each study provided information which could be used in the future to revise or simplify the models. Elk locations should be more accurate as indicators of selection than are seasonal ranges because the seasonal ranges are extrapolations from the locations, thus there were probably areas within the seasonal ranges where elk were never found. Therefore, using seasonal ranges rather than locations to compare elk use to availability in the study area is probably more conservative. The results of all tests of the model applied to the Benson, and all but the home ranges vs. random circles test in the Nahwitti, validated the models by showing that the elk were using more of the high quality and less of the low quality habitat relative to what was available. In the Nahwitti, the mean HSI values of the home ranges were generally no better than those of similar sized areas selected at random from within the study area. Not all locations were found in high quality habitat, but I expected to find 42 some locations in lower quality areas if the elk were forced there by predators or human harassment, or if they were travelling between high quality patches. Thus, one would not expect a perfect fit to the model, and working with statistical significance levels of 95% is overly conservative for wildlife managers when 75-80% would suffice for large scale planning (Hurley 1986). Wiser management decisions can be made by using models, even if they are imperfect (Chalk 1986). There are several possible explanations for finding no difference between the seasonal ranges and the study area (represented by circles) in the Nahwitti. First, the adaptive kernel method of calculating home ranges includes areas peripheral to the locations which may be of low quality and not used by the elk at all; therefore, the estimated home range may be of lower quality than the true home range. This seemed to be more a problem in the Nahwitti, and was especially evident in the southern end of some seasonal ranges where the river runs east-west. The shape of the "kernel" in KERNELHR is determined by the mean squared error (MSE) in the X and Y directions. The program chooses a smoothing parameter which minimizes the MSE, and when the locations span a greater distance in one direction, the kernels, and thus the home range, are extended in that direction (i.e. the kernels become elongated instead of circular; Worton 1989). In the Nahwitti the seasonal ranges were extended in the north-south direction (Figures 7 and 11). This appears realistic in the part of the valley that runs north-south, but not when the valley runs east-west. This problem might be overcome by choosing a smoothing constant such that the area around each location was circular instead of allowing KERNELHR to choose the optimal one. 43 A second alternative would be to use the minimum convex polygon (MCP) instead of the adaptive kernel method. I tried this method with the 1992 summer locations of elk #441 which showed the worst result in the tests of seasonal ranges vs. random circles. The adaptive kernel summer range had a higher HSI value than only 6% of the corresponding circles, whereas the MCP summer range was higher than 58% of the circles (Figure 12). 0.7 • I" 0 . 6 • | 0 .5 "5 0 .4 CO & ° 3 E 0 . 2 i a i . CO 0 X = 0.53, SE = 0.004 0.48 Adaptive Kernel Range or Random Circle 0 7 -, 0.6 -S 0 .5 -3 0 .4 -CO 0 .3 -4> E 0.2 -E 3 0.1 -CO 0 J 0.53 v X= 0.51, SE = 0.009 MCP Range or Random Circle Figure 12: Mean HSI values of elk #441's summer 1992 adaptive kernel and Minimum Convex Polygon ranges (black) and corresponding random circles (white). The third explanation is that the elk's seasonal ranges were relatively large compared to the total Nahwitti study area (the largest filled 29% of the study area). In this case, any circle of equal size to the elk range would contain part of the seasonal range because of the latter's shape, size and central placement in the study area. Fourth, due to budget constraints, the habitat mapping was not done to the finest detail possible in the Nahwitti. The Nahwitti study area in general had higher HSI values (mean summer HSI = 0.49, mild winter = 0.42) than did the Benson (summer = 0.34, mild winter = 0.26). The large expanses of the Nahwitti plateau which run along each 44 side of the study area have a large amount of "medium" suitability in summer, but they were not used by collared elk. My feeling is that the plateau may have been given too high an HSI value because it was considered one broad mosaic instead of many small patches of different habitat. Last, although this "circle method" was used to incorporate some of the spatial components of the habitat suitability map into the test, averaging suitability values of the seasonal ranges and the circles may have resulted in a failure to identify high quality habitat. For example, a homogeneous area of suitability value 0.45 may not be as valuable for elk as would an equal sized area with half being valued at 0.05 and half at 0.85, yet both would have the same mean. Presumably there is some point at which an area changes from "sufficiently high suitability to satisfy the needs of elk" to "insufficient". In my example, if that threshold was 0.50, the homogeneous area could not support elk at all, whereas half of the heterogeneous area could. The placement of the study area boundaries may affect the results of a use-availability study for Johnson's (1980) level 2 selection, especially when the pattern of habitat patches is aggregated (Porter and Church 1987). Ideally, the study area should extend as far as it would be possible for elk to travel - in this case, perhaps all of Vancouver Island. In reality, due to their gregariousness, they are not likely to travel outside of the range of the population into which they are born. An area outside this is either not available to them, or is available but not selected. I believe the boundaries of my study area in the Benson were realistic because much of the boundary was at the height of land, a natural barrier. However, in the Nahwitti I would have preferred (a 45 posteriori) a larger study area because the seasonal ranges covered a larger portion of the study area, and the rolling topography did not provide any obvious barriers. Uncertainty in the extent of the true available area can be reduced by considering 2 different levels of selection (Aebischer et al. 1993; Carrol et al. 1995). The study area could be very different from what elk perceive as the area available to them. Reducing the available area to the extent of the home range, increases the probability that elk are familiar with the area. An unused area within the home range may be avoided, whereas an unused area within the study area (but outside the home range) may be unavailable. On the other hand, elk have already selected the home range, presumably because of its high quality, so using this as the available area would result in a more conservative test than one using the study area. Therefore, testing at both levels seems appropriate. In both study areas, there were large areas of high HSI value that were not used by the collared elk, and some areas of low HSI value that were used. The latter case suggests that some factors are missing from the model (see "Problems with the models"). The former has many possible explanations. The HSI does not include any effects of competitors, predators, parasites, exploitation, or harrassment by humans (Schamberger and O'Neill 1986). In the Benson, there were other groups of elk separate from the collared elk. This, along with the presence of wolves (Canis lupus), cougars (Felis concolor), hunters, and both commercial and recreational traffic could have affected which areas were used by collared elk. Therefore, use of an area by elk does not always indicate that the area is of high HSI value, nor does non-use indicate 4 6 low HSI values (especially since one cannot be certain that an area is not used by some elk). Animals of different age and sex may have different movements and patterns of range use (Harris et al. 1990), so conclusions cannot be drawn from the collared animals about the entire elk population. Only adult females were used to test the model, but elk are gregarious, and age-sex class counts of other individuals observed with the collared elk lead me to believe that other adult females and juveniles had very similar home ranges to the collared elk. Adult males were observed with the female-young groups during the rut (autumn), and occasionally during the rest of the year. Problems with the Models If a model's predictions are accurate, some confidence is placed in its predictive ability, but no conclusions can be drawn about its veracity. In fact, predictions which are clearly wrong can lead us to a better understanding of the system being modeled (Bunnell 1973) because they can help to identify gaps in our knowledge. While my data generally seem to fit the model, the results are not absolutely clear, and this has given me insights into possible problems with either the model itself or with my method of testing the model. I believe significant improvements to the models could be made in how cover influences the HSI values. Polygons were given cover values of either 0 or 0.99, indicating that cover was either absent or present. I felt that in some cases there should have been intermediate values for security cover. For example, in a large cutblock with undulating topography, the cover would have been rated "0" because the trees were not tall enough to hide the elk, yet the elk could be completely hidden 47 behind a log, a shrub, or a small rise. Combinations of undulating topography (which could be identified on aerial photographs) and height of shrubs (which would be related to the date of logging) could have been used to improve the models by providing intermediate values for security cover. Brunt (1991) discussed the importance of both thermal and snow interception cover, yet his model combined the two because his study area did not include any second growth old enough to qualify as thermal cover, but too young to qualify as snow interception cover. In the Benson, this type of forest stand was present and probably should have been distinguished in the model. I suggest that this modification be made if the model is used again. Furthermore, a problem arose with cover values in stands that were mosaics of 2 different habitat types because I gave these polygons the cover value of the predominant habitat type; however, an intermediate value may have been more accurate. To compound the security cover problem, darkness may provide adequate security in clearcuts at night (Beyer and Haufler 1994). This would explain why "pit-lamping" is so successful for poachers, and why I was able to get within 25 m of elk groups in a clearcut at night, whereas during the daylight this was rarely possible. The ability to see long distances may also gives elk a sense of security. While snow was present in the Benson, collared elk stayed on a large clear-cut side hill with very little vegetative cover. Schwartz and Mitchell (1945) also found that elk used logged areas in winter, and in fact most elk and elk tracks seen during an aerial census of northern Vancouver Island at this time were on clearcut or relatively open side slopes of valleys. Wolves are known predators of elk on Vancouver Island (Scott and Shackleton 1980) and I found their sign in the Benson in winter. A small patch of 48 bushes (i.e. ~ 0.1 ha) in that same clear-cut side hill was also chosen as a calving spot by one of my collared elk. Perhaps in both cases, the elk needed extra time to flee from predators (because of the presence of snow or young), and having a clear vantage point to see any threat well in advance, might have been more advantageous than being in dense cover where a predator could surprise them. On the other hand, open clearcuts in winter could be the only place where food is found in sufficient quantity. There, full light allows maximum summer growth of shrubs and small conifers which provide most of the winter diet (Brunt 1990), and unless directly threatened by a predator, elk may not move through snow to safer cover because of energetic constraints. Furthermore, winter temperatures on these slopes (east and west facing) could be higher than on the valley floor due to exposure to sunlight on the slopes and to cold air pooling on the valley floor. During the time when snow was present, elk #241 and #249 and other members of the group (N = 5 to 13) chose the side slopes, possibly for one of the above reasons. The average HSI value of their locations at this time was 0.19 and the minimum 0.06. Their choice of low suitability slopes over nearby areas of significantly higher suitability leads me to conclude that some factor or combination of factors (discussed above) is either misclassified or missing from the mild winter suitability model. Management Implications To have confidence in applying the models to forest management, data for severe winter should be used to test the severe winter model. Knight (1970) found that Rocky Mountain elk used open areas in mild winters, but used timbered areas in severe winters. In any winter, the lack of high quality forage and increased energetic costs of 49 thermal regulation and movement through snow probably create a net loss of energy for elk (Janz 1983), and a severe winter would make this problem more extreme. The amount of high quality, severe winter habitat may be the limiting factor to the long term survival of a population. In this case, management decisions for an area should maximize the high quality habitat delineated by the severe winter model. Due to recent mild winters on Vancouver Island, however, the severe winter model has not been tested (this study, Sovka 1993, Brunt 1991). While Brunt's (1991) models appear to predict which habitats are suitable for elk, the HSI values would have to be updated along with the forest cover maps as stands are harvested and succession proceeds. The understory-based habitat map, which is essential to the HSI models for elk, is not normally available to forest managers, and is expensive to create. Thus, the models may not be practical at this time in their present form. However, I believe that there is still valuable information to be gained from them. Forest managers should be aware of any old growth areas which have high HSI values, especially in severe winters, because logging would substantially reduce these values by removing thermal and snow interception cover. Also, if they are aware of how each variable in the model influences the HSI value, then they will have a better idea of which areas will be suitable for elk, even without running (or updating) the models. Monitoring wildlife populations is the key to adaptive resource management (Salwasser 1985), and a healthy, stable elk population should indicate that the area provides good elk habitat. Also, recording the location of elk sign and sightings can help delineate preferred habitats. The stand prescription controls the vegetative 50 substrates upon which wildlife species can find food and cover resources (Salwasser 1985), so forest managers have the ultimate responsibility to ensure that habitats for all species in their areas are maintained, not just for popular animals like elk, deer, bears, and spotted owls. 51 LITERATURE CITED Ackerman, B. B., F. A. Leban, M. D. Samuel and E. 0. Garton. 1990. User's manual for Program Home Range: second edition. For., Wildl. and Range Exp. Stn. Tech. Rep. 15 Contrib. No. 259. University of Idaho, Moscow. Aebischer, N. J., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of habitat use from animal radio-tracking data. Ecology 74: 1313-1325. Berry, K. B. 1986. Introduction: development, testing, and application of wildlife-habitat models. Pp. 3-4 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. Beyer, D. E. Jr. and J. B. Haufler. 1994. Diurnal versus 24-hour sampling of habitat use. J. Wildl. Manage. 58:178-180. Brunt, K. R. 1990. Ecology of Roosevelt elk. Ch. 3 in J. B. Nyberg and D. W. Janz eds. Deer and elk habitats in coastal forests of southern British Columbia. B. C. Min. For., Victoria, B.C. Brunt, K. R. 1991. Testing models of the suitability of Roosevelt elk seasonal ranges. M.Sc. thesis, Univ. of Victoria, Victoria. 157 pp. Brunt, K. R., D. Q. Becker and J. A. Youds. 1989. Vancouver Island Roosevelt elk intensive forestry interactions phase 1 job completion report. B. C. Min. Env. and Min. For. IWIFR-33 Wildl. Bull. No. B-51. Bunnell, F. L. 1973. Theological ecology or models and the real world. For. Chron. 49:167-172. Bunnell, F. L. 1989. Alchemy and uncertainty: what good are models? Gen. Tech. Rep. PNW-GTR-232. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station. 27 pp. Burt, W. H. 1943. Territoriality and home range concepts as applied to mammals. J. Mammal. 24: 346-352. Carroll, J. P., R. D. Crawford and J. W. Schulz. 1995. Gray partridge winter home range and use of habitat in North Dakota. J. Wildl. Manage. 59: 98-103. Chalk, D. E. 1986. Summary: development, testing and application of wildlife-habitat models - the researcher's viewpoint, pp. 155-156 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. 52 Clark, J. D., J. E. Dunn and K. G. Smith. 1993. A multivariate model of female black bear habitat use for geographic information system. J. Wildl. Manage. 57: 519-526. Cochran, W. G. 1954. Some methods for strengthening the common chi-square tests. Biometrics 10:417-451. Collins, W. B. and P. J. Urness. 1983. Feeding behaviour and habitat selection of mule deer and elk on northern Utah summer range. J. Wildl. Manage. 47: 646-663. Dixon, K. R. and J. A. Chapman. 1980. Harmonic mean measure of animal activity areas. Ecology 61: 1040-1044. Edge, W. D., and C. L. Marcum. 1989. Determining elk distribution with pellet-group and telemetry techniques. J. Wildl. Manage. 53: 621-624. Edge, W. D., C. L. Marcum and S. L. Olsen-Edge. 1987. Summer habitat selection by elk in western Montana: a multivariate approach. J. Wildl. Manage. 51: 844-851. Eng, M. and J. Schieck. 1992. Evaluating an expert system model: habitat use by black-tailed deer. B.C. Ministry of Forests unpublished report. Franklin, W. L, A. S. Mossman and M. Dole. 1975. Social organization and home range of Roosevelt elk. J. Mammal. 56: 102-118. Hanley, T. A. 1984. Habitat patches and their selection by wapiti and black-tailed deer in a coastal montane coniferous foest. J. Appl. Ecol. 21: 423-436. Harris, S., W. J. Cresswell, P. G. Forde, W. J. Trewhella, T. Woollard, and S. Wray. 1990. Home-range analysis using radio-tracking data - a review of problems and techniques particulartly as applied to the study of mammals. Mammal Rev. 20: 97-123. Hurlbert, S. H. 1984. Pseudoreplication an the design of ecological field experiments. Ecol. Monogr. 54: 187-211. Hurley, J. F. 1986. Summary: development, testing and application of wildlife-habitat models - the manager's viewpoint. Pp. 151-153 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. Irwin, L. L and J. M. Peek. 1983. Elk, Cervus elaphus, foraging related to forest management and succession in Idaho. Can. Field Nat. 97: 443-447. 53 Janz, D. W. 1983. Seasonal composition and quality of Roosevelt elk diets on Vancouver Island. M.Sc. thesis. Univ. B.C., Vancouver, B.C. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61: 65-71. Knight, R. R. 1970. The Sun River elk herd. Wildl. Monogr. 23. 66 pp. Lancia, R. A., D. A. Adams, and E. M. Lunk. 1986. Temporal and spatial aspects of species-habitat models. Pp. 65-69 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. Lancia, R. A., S. D. Miller, D. A. Adams, and D. W. Hazel. 1982. Validating habitat quality assessment: an example. Trans. N. Am. Wildl. Nat. Res. Conf. 47: 96-110. Larkin, R. P. and D. Halkin. 1994. Wildlife software: a review of software packages for estimating animal home ranges. Wildl. Soc, Bull. 22: 274-287. Laymon, S. A. and R. H. Barrett. 1986. Developing and testing habitat-capability models: pitfalls and recommendations. Pp. 87-91 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. Laymon, S. A. and J. A. Reid. 1986. Effects of grid-cell size on tests of a spotted owl HSI model. Pp. 93-96 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. Manly, B. F. J. 1991. Randomization and Monte Carlo methods in biology. Chapman and Hall, London, U.K. 281 pp. Marcot, B. G., M. G. Raphael, and K. H. Berry. 1983. Monitoring wildlife habitat and validation of wildlife-habitat relationships models. Trans. N. Am. Wildl. Nat. Res. Conf. 48: 315-329. Marcum, C. L. and D. O. Loftsgaarden. 1980. A nonmapping technique for studying habitat preferences. J. Wildl. Manage. 44: 963-968. Marcum, C. L. and M. D. Scott. 1985. Influences of weather on elk use of spring-summer habitat. J. Wildl. Manage. 49: 73-76. McCorquodale, S. M. 1993. Winter foraging behaviour of elk in the scrub-steppe of Washington. J. Wildl. Manage. 57: 881-890. 54 McCorquodale, S. M., K. J. Raedeke, and R. D. Taber. 1989. Home ranges of elk in an arid environment. Northwest Sci. 63: 29-34. McNab, B. K. 1963. Bioenergetics and the determination of home range size. Am. Nat. 97: 133-140. Merrill, E. H. 1991. Thermal constraints on use of cover types and activity time of elk. Appl. Anim. Behav. Sci. 29: 251-267. Nams, V. O. and S. Boutin. 1991. What is wrong with error polygons? J. Wildl. Manage. 55: 172-176. Parker, K. L, C. T. Robbins, and T. A. Hanley. 1984. Energy expenditures for locomotion by mule deer and elk. J. Wildl. Manage. 48: 474-488. Pederson, R. J., A. W. Adams and J. M. Skovlin. 1980. Elk habitat use in an unlogged and logged forest environment. Oregon Dept. of Fish and Wildlife, Research and Development Section, Portland, Oregon. 121pp. Porter, W. F. and K. E. Church. 1987. Effects of environmental pattern on habitat preference analysis. J. Wildl. Manage. 51: 681-685. Reynolds, T. D. and J. W. Laundre. 1990. Time intervals for estimating pronghorn and coyote home ranges and daily movements. J. Wildl. Manage. 54: 316-322. Saltz, D. 1994. Reporting error measures in radio location by triangulation: a review. J. Wildl. Manage. 58: 181-184. Salwasser, H. 1985. Integrating wildlife into the managed forest. For. Chron. 61:146-149. Schamberger, M. L. and L. J. O'Neill. 1986. Concepts and constraints of habitat-model testing. Pp. 5-10 in J. Verner, M.L. Morrison, and C.J. Ralph, eds. Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. U. of Wisconsin Press, Madison, Wisconsin. Schmutz, J. A. and G. C. White. 1990. Error in telemetry studies: effects of animal movement on triangulation. J. Wildl. Manage. 54: 506-510. Schoen, J. W., R. W. Flynn, L. H. Suring, K. Titus and L. R. Beier. 1992. Habitat capability model for brown bear in southeast Alaska. U.S. Dept. Agric. For. Serv, Alaska. 37 pp. Schooley, R. L. 1994. Annual variation in habitat selection: patterns concealed by pooled data. J. Wildl. Manage. 58: 367-374. 55 Schwartz, J. E. and G. E. Mitchell. 1945. The Roosevelt elk on the Olympic Peninsula, Washington. J. Wildl. Manage. 9: 295-319. Scott, B. M. V. and D. M. Shackleton. 1980. Food habits of two Vancouver Island wolf packs: a preliminary study. Can. J. Zool. 58: 1203-1207. Seaman, D. E. and R. A. Powell. 1991. Kernel home range estimation program. North Carolina State University, Raleigh, NC. Skovlin, J. M. 1982. Habitat requirements and evaluations. Pp. 369-413 in J. W. Thomas and D. E. Toweill, eds. Elk of North America: ecology and management. Stackpole, Harrisburg, PA. Sovka, D.G. 1993. Home range behaviour of Roosevelt elk in Strathcona Park. M.Sc. thesis, Univ. of British Columbia, Vancouver. 117 pp. Springer, J. T. 1979. Some sources of bias and sampling error in radio triangulation. J. Wildl. Manage. 43: 926-935. Swihart, R. K. and N. A. Slade. 1985. Testing for independence of observations in animal movements. Ecology 66: 1176-1184. Thomas, J. W., H. Black Jr., R. J. Scherzinger, and R. J. Pederson. 1979. Deer and Elk. Pp. 104-127 in J. W.Thomas, ed. Wildlife habitats in managed forests -the Blue Mountains of Oregon and Washington. U.S. Dept. Agric. For. Serv. Agric. Handbk. No. 553. Washington, D.C.. Troyer, W. A. 1960. The Roosevelt elk on Afognak Island, Alaska. J. Wildl. Manage. 24:15-21. Wilson, S. F., K. L. Campbell, and D. M. Shackleton. Methods for measuring home range selection within a study area. In preparation. Witmer, G. W. and D. S. deCalesta. 1983. Habitat use by female Roosevelt elk in the Oregon Coast Range. J. Wildl. Manage. 47: 933-939. Witmer, G. W., M. Wisdom, E. P. Harshman, R. J. Anderson, C. Carey, M. P. Kuttel, I. D. Luman, J. A. Rochelle, R. W. Scharpf, and D. Smithey. 1985. Deer and elk. Pp. 231-258 in E. R. Brown, ed. Management of wildlife and fish habitats in forests of western Oregon and Washington. U. S. Dept. Agric. Publ. No. R6-F&WL-192-1985. Worton, B. J. 1987. A review of models of home range for animal movement. Ecological Modelling 38: 277-298. . 1989. Kernel methods for estimating the utilization distribution in home-range studies. Ecology 70: 164-168. 56 Appendix 1: Understory-based habitat types. Habitat Label Habitat Type Trees Understory Hygro-tope* Tropho-tope** CP Pine-redcedar-cypress woodland pine, redcedar, yellow cedar, western hemlock salal, deer fern, goldthread 5-6 B S1CH salal-moss, cedar-hemlock phase open redcedar-western hemlock (amabilis fir) salal, (Vaccinium), moss 4-5 C S1HA salal-moss, . hemlock-amabilis fir phase western hemlock, amabilis fir moss, (alaskan blueberry, red huckleberry, salmonberry, deer/sword/ spiny wood fern, 3-leaved foam flower) 4-5 C S1C colluvial phase of S1 western hemlock, amabilis fir moss, Vaccinium 4-5 C S2 salal-folisol over rock open, scrubby redcedar & western hemlock salal, very sparse herb layer 2-3 B S2F same as S2, Douglas fir phase same, includes Douglas Fir same as S2, also lichen and bare rock 2-3 B S3 alluvial swordfern-foamflower Sitka spruce, western hemlock, amabilis fir Hiaher terraces and microsites: Red huckleberry, alaskan blueberry, sword/deer/spiny wood fern, 3-leaved + cut-leaved foamflower Low terraces: salmonberry. elderberry, devil's club. ferns(oak, lady, maidenhair), twistedstalk, false bugbane, bedstraw, hedge-nettle, coast boykinia, pink fawn lily 5 D S3B beach phase of S3 see S3 higher terraces 4 D S4 alluvial alder-spruce-herb red alder (Sitka spruce) salmonberry, elderberry, stink currant, youth-on-age.sword fern, hedge-nettle, (cow parsnip, lady fern, bleedingheart, bedstraw, purple sweetcicely, Pacific oenanthe, 3-leaved + cut-leaved foamflower, grasses) 5 E S5 wet alluvial spruce-redcedar-skunk cabbage Sitka spruce, redcedar, (western hemlock) same as S3, but more lady fern, also skunk cabbage, (slough sedge, nodding trisetum) 6 E S6 redcedar-skunk cabbage swamp forest redcedar (spike-topped) poor western hemlock and Sitka spruce, all on raised microsites Raised Microsites: salal, (Vaccinium, Pacific menziesia), deer fern, bunchberry, false Solomon's seal, twayblades Depressional Microsites: skunk cabbage, goldthread 7 D S7 shore pine -redcedar, sedge - skunk cabbage shore pine, redcedar skunk cabbage, sedge 8 D S8 shore pine bog shore pine Labrador tea, sphagnum moss . 7 B 57 S9 open bog nil sedges, sphagnum moss 8 B S10 redcedar, western hemlock salal, moss 4 B S11 sweetgale-hardhack-sedge fens sparse small red cedar sweet gale, hardhack, slough sedge, (slender rein orchid, great burnet, shooting star) 8 D S12 limestone holly fern same as S2 salal, ferns:(holly, maidenhair, sword, spiny wood), green & maidenhair spleenwort, polypody, twistedstalk, 3- & cut-leaved foamflower 3-4 D-E S12F Douglas fir phase of S12 redcedar, Douglas fir, western hemlock same as S12 3-4 D-E S13 colluvial salmonberry-swordfern western hemlock, amabilis fir, (redcedar, Sitka spruce) Vaccinium, menziesia, (salmonberry, red elderberry), devil's club, swordfern, 3- & cut-leaved foamflower, diverse herbs 4-5 D S13L limestone phase of S13 same as s13 same as S13, also thimbleberry 4-5 D M1 montane Vaccinium moss western hemlock, amabilis fir, (yellow cedar in climax) moss, alaskan blueberry, oval-leaved blueberry, (bunchberry. 5-leaved bramble, twistedstalk. wintergreen, clintonia, twayblade, false hellabore, spleenwort-leaved goldthread.) 4 C M1C colluvial phase of M1 same as M1 same as M1 4 C M2 montane salal yellow cedar, western hemlock, suppressed mountain hemlock mostly salal, (red huckleberry, alaskan blueberry, oval-leaved blueberry, menziesia),bunchberry, deer fern, 5-leaved bramble, northern twinflower, twayblades, twistedstalk, clubmoss 3 B M3 montane oakfern same as M1, but old growth is larger and taller like M1. also (Devil's club), oak fern, red baneberrv. 3-leaved foamflower, yellow wood violet 5 D M3C colluvial phase of M3 same as M3 same as M3 5 D M4 wet montane skunk cabbage -marsh marigold western hemlock, yellow cedar, small mountain hemlock, all on raised microsites Hiah sites: Vaccinium. salal. red mountain heather, deer fern, bunchberry, twistedstalk, 5-leaved bramble, wintergreen, twayblades. Low sites: skunk cabbaae. marsh marigold, deer cabbage, false hellabore, goldthread. 6 D M5 montane Devil's club western hemlock, amabilis fir, yellow cedar Vaccinium. menziesia. Devil's club, salmonberry. oak and deer fern, twistedstalks, clintonia, false hellabore, coast boykinia, bunchberry, twayblades, clubmoss, 5-leaved bramble, 3-leaved foamflower 5 D MH1 closed mountain hemlock forest closed canopy of mountain hemlock & yellow cedar, shrubby amabilis fir oval leaved blueberry, (red mountain heather), sparse 5-leaved bramble, bunchberry, twayblades, twistedstalk, goldthread, with false hellabore & marsh marigold in wet areas 4 C MH2 rocky mountain hemlock forest same as MH1 same as MH1 3 B 58 p mountain hemlock parkland stunted mountain hemlock & yellow cedar in clumps openings dominated by heather or sedge oval-leaved & alaskan blueberry, black crowberry, red mountain heather, cassiope, copperbush, bunchberry, goldthread various A snow avalanche track Sitka alder, or stunted mountain hemlock & yellow cedar lush herbs and ferns 4-5 D T talus - - - -NV non vegetated - - - -AT alpine tundra - heathers, lichens various W wetland (montane / sub-alpine) (yellow cedar) sphagnum, sedges 8 D R bare rock - •• - - -* Soil moisture regime: 0 (very dry) to 8 (very wet). ** Soil nutrient regime: A (very poor) to E (very rich). 59 Appendix 2: Common and scientific names of plant species listed. Trees: alder, red balsam fir cedar, western red cedar, yellow Douglas fir hemlock, mountain hemlock, western pine, shore spruce, Sitka Alnus rubra Abies amabilis Thuja plicata Chamaecyparis nootkatensis Pseudotsuga menziesii Tsuga mertensiana Tsuga heterophylla Pinus contorta Picea sitchensis Shrubs: blueberry, alaskan blueberry, oval-leaved copperbush crowberry, black currant, stink devil's club elderberry, red hardhack heather, red mountain heather, white mountain huckleberry, red Labrador tea Oregon grape, dull Pacific menziesia Pacific ninebark salal salmonberry sweet gale thimbleberry willow Herbs: baneberry, red bedstraw, sweet-scented bleedingheart, Pacific bracken bramble, 5-leaved bugbane, false bunchberry burnet, great clintonia clubmoss Vaccinium alaskaense Vaccinium ovalifolium Cladothamnus pyroliflorus Empetrum nigrum Ribes bracteosum Oplopanax horridus Sambucus racemosa Spiraea douglasii Phyllodoce emeptriformis. Cassiope mertensiana Vaccinium parvifolium Ledum groenlandicum Mahonia nervosa Menziesia ferruginea Physocarpus capitatus Gauitheria shallon Rubus spectabilis Myrica gale Rubus parviflorus Salix spp. Actaea rubra arguta Galium triflorum Dicentra formosa Pteridium aquilinum Rubus pedatus Trautvetteria caroliniensis Cornus unalaschkensis Sanguisorba officinalis Clintonia uniflora Huperzia selago coast boikinia Boikinia occidentalis cow parsnip Heracleum sphondyllium false hellabore Veratrum viride eschscholtzii fern, deer Blechnum spicant fern, holly Polystichum lonchitis fern, lady Athyhum filix-femina fern, maidenhair Adiantum pedatum aleuticum fern, oak Gymnocarpium dryopteris fern, spiny wood Dryopteris assimilis fern, sword Polystichum munitum fireweed Epilobium angustifolium foamflower, 3-leaved Tiarella trifoliata foamflower, cut-leaved Tiarella laciniata goldthread Coptis aspleniifolia hedge-nettle Stachys cooleyae lilly, pink fawn Erythronium revolutum marigold, marsh Caltha leptosepala biflora orchid, slender rein Platanthera stricta Pacific oenanthe Oenanthe sarmentosa polypody Polypodium hesperium sedges Carex, spp. shooting star Dodecatheon jeffreyi skunk cabbage Lysichiton americanum slough sedge Carex obnupta Solomon's seal, false Maianthemum dilatatum spleenwort, green Asplenium viride spleenwort, maidenhair Asplenium trichomanes sweetcicely, purple Osmorhiza purpurea trisetum, nodding Trisetum cernuum twayblades Listera spp. twinflower Linnaea borealis twistedstalk Streptopus spp. yellow wood violet Viola glabella youth-on-age Tolmiea menziesii wintergreen Pyrola asarifolia 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items