Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A quantitative evaluation of porcupine-habitat relationships in the Kalum Valley, B. C. Lawson, Andrea L 1991

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

Item Metadata


831-ubc_1992_spring_lawson_andrea.pdf [ 14.22MB ]
JSON: 831-1.0086459.json
JSON-LD: 831-1.0086459-ld.json
RDF/XML (Pretty): 831-1.0086459-rdf.xml
RDF/JSON: 831-1.0086459-rdf.json
Turtle: 831-1.0086459-turtle.txt
N-Triples: 831-1.0086459-rdf-ntriples.txt
Original Record: 831-1.0086459-source.json
Full Text

Full Text

A QUANTITATIVE EVALUATION OF PORCUPINE-HABITATRELATIONSHIPS IN THE KALUM VALLEY, B.C.byANDREA L. LAWSONB.Sc., The University of Western Ontario, 1987A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESDepartment of BotanyWe accept this thesis as conformingTHE UNIVERSITY OF BRITISH COLUMBIADecember 1991© Andrea L. Lawson, 1991In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.Department ofThe University of British ColumbiaVancouver, CanadaDate DE-6 (2/88)iiABSTRACTThe implications that the interactions between animals and thecommunities in which they exist have to wildlife management arejust being realized. The assessment of species-habitatinteractions necessitates the consideration of many variables andthe adaptation of multivariate statistics to ecology has madepossible the multidimensional consideration of habitats. Thepurpose of this thesis is to evaluate the usefulness of amultivariate approach to an applied management problem: porcupine-habitat interactions in the Kalum Valley B.C. In recent years analarming amount of damage caused by the winter feeding ofporcupines has been occurring on the north coast of B.C. There isconcern that silvicultural practices, such as thinning of croptrees, are predisposing stands to attack. The three specificobjectives addressed in this thesis are: 1) To test the hypothesisthat thinned stands incur more damage than unthinned stands 2) Toinvestigate the process by which porcupines are selecting habitatsand individual trees 3) To determine the variable or combinationof variables that best predicts damage.Four 100 ha. blocks of forest were selected in the Kalum Valley.Two stands had been spaced and two were unmanaged. Within thesestands 25 sampling plots were randomly selected. Three sets ofiiivariables were recorded in each plot: recent and past porcupinedamage to trees, % cover of all species of vegetation, % cover ofvegetation strata.The results indicate that thinned stands do not incur more damagethan unthinned stands. In fact, unthinned stands incur more newdamage than do spaced areas, indicating that porcupinesdemonstrate choice at the level of large blocks of forest perhapson the basis of stand properties such as density or basal area oftrees. Hemlock was almost exclusively attacked with damage peakingin the 20.1-20.5 diameter class. Within stands, damage is notrelated to density, basal area of trees or any site features at thelevel of the plot except the number of stumps. Damage is relatedto the cover of a few herbs. This result is probably related tothe greater amounts of light reaching the forest floor in damagedareas resulting from the dead tops of the hemlock trees.Thus,^damage is predictable from the individual treecharacteristics of species and diameter class of trees. Trees inunthinned areas appear to be more likely to incur damage.ivTABLE OF CONTENTSPAGEABSTRACT^ iiLIST OF TABLES^ viLIST OF FIGURES viiAKNOWLEDGEMENTS^ ixCHAPTER 1^INTRODUCTION^ 1^1.1^Herbivore-Habitat Interactions; theporcupine problem in B.C.^ 11.2^General Porcupine Ecology 51.3^Habitat and Feeding Preferences^101.4^Relevant Studies in the Pacific Northwest^141.5^Objectives^ 16CHAPTER 2 METHODS 172.1^Study Site^ 172.2^Stand Structure and Damage Assessment^252.3^Habitat Description^ 292.3.i Vegetation Description^ 292.3.ii Site Description 302.5^Data Analysis^ 352.5.i^Stand Structure^ 382.5.ii Damage Data 392.5.iii Site and Vegetation Data^402.5.iv Relating Damage to Site andVegetation Data^ 43VTABLE OF CONTENTS continued^ PAGECHAPTER 3 RESULTS^ 45^3.1^Stand Structure 453.2^Incidence of Damage^ 473.3^Damage Index^ 573.4^Vegetation and Site Variables^ 633.5^Relating Damage to Site and Vegetation^69CHAPTER 4^DISCUSSION^ 784.1^Comparability of Stands^ 784.2^Effects of Thinning 784.3^Process of Selection by Porcupines^794.3.i Evidence for a hierarchicalprocess of selection^ 804.3.ii Evidence for selective feeding^834.4^Recommendations^ 854.5 A Comment on the Use of Multivariate Methodsin Wildlife Management.^ 86LITERATURE CITED^ 88viLIST OF TABLESTable 2.1.Table 2.2.Table 2.3.Table 3.1.Table 3.2.PAGEDescription of the four study areas.^22Index used for percent cover of plant species 30Variables measured in the site description.^31Summary of the number and species of treessampled and those incurring damage.^45Estimates of density and basal area for alltree species and for hemlocks only.^46The incidence of damage to western hemlock.^55The results of regressions of damage on densi-ty and basal area of all tree species and ofhemlocks only.^ 62Table 3.3.Table 3.4.Table 3.5. The treatment means of the site, species andvegetation stata variables, and the resultsof the MANOVA testing for differences betweengrids.^ 65Table 3.6. The results of the canonical correlationanalysis of the site and vegetationmatrices.^ 67Table 3.7. The correlations (loadings) between the siteand vegetation variables and their first twocanonical discriminant axes. 68Table 3.8.Table 3.9.The results of a canonical redundancy analy-sis for the first two canonical variables ofthe CCA of site and vegetation variables. 70The results of regressions of damage on setsof habitat variables.^ 72Table 3.10. The canonical loadings of all variables onthe discriminant axis from the canonicaldiscriminant analysis with respect to the'high' or 'low' damage categories.^76vi iLIST OF FIGURESPAGEFigure 1.1^Photograph of a porcupine.^ 6Figure 1.2^Photograph of porcupine damage.^8Figure 1.3^Photograph of a 'spike top'. 9Figure 2.1^Map of the north coast of B.C. showingthe Kalum Valley.^ 18Figure 2.2^Photograph of the general appearenceof forests in the Kalum Valley.^19Figure 2.3^The biogeoclimatic zones of northernB.C.^ 20FIgure 2.4^Map showing the location of the fourstudy grids.^ 21Figure 2.5a Photograph of the appearance of aspaced forest area.^ 23Figure 2.5b Photograph of the appearance of anunspaced forest area. 24Figure 2.6^Photograph showing the use of therelescope.^ 26Figure 2.7^Photograph showing damage assessmentof a hemlock tree.^ 28Figure 2.8^Photograph showing a sampling plot.^36Figure 2.9^Photograph showing a soil pit.^36Figure 3.1a Graph of the mean and standard error ofthe density (stems/ha.) of all trees ineach of the four grids. 48Figure 3.1b Graph of the mean and standard error ofthe density (stems/ha.) of hemlocks ineach of the grids.^ 48Figure 3.2a Graph of the mean and standard error ofthe basal area (m 2 /ha.) of all trees ineach of the grids. 49viiiLIST OF FIGURES continuedFigure 3.2b Graph of the mean and standard error ofthe basal area (m 2/ha.) of hemlocks ineach of the grids.^ 49^Figure 3.3^Graph of the mean and standard error ofthe diameter at breast height of hemlocksin each of the grids.^ 50Figure 3.4^Graph of the mean and standard error ofthe number of wounds per hemlock in eachof the grids.^ 51Figure 3.5^Bar graph of the number of old and newwounds per hemlock with respect to treediameter class.^ 53Figure 3.6^Graph of a quadratic regression of thenumber of new wounds per hemlock withrespect to tree diameter class.^54Figure 3.7a Bar graph of the number of old woundsper hemlock in each third of the treewith respect to tree diameter class. 56Figure 3.7b Bar graph of the number of new woundsper hemlock in each third of the treewith respect to tree diameter class. 56Figure 3.8^Graph of the variables 'number of woundsper hemlock' and 'circumference girdledper hemlock' showing the first PC-axis.^58Figure 3.9a Graph of the mean and standard error ofthe damage summary index, calculated onthe basis of old wounds, in each grid. 60Figure 3.9b Graph of the mean and standard error ofthe damage summary index, calculated onthe basis of new wounds, in each grid. 60Figure 3.10^Graph of the variables 'number of woundsper hemlock' and 'circumference girdledper hemlock' showing the division forthe categorical damage variable.^73Figure 3.11^Box plots of the categorical damagevariable against the discriminantfunction.^ 74LXACKNOWLEDGEMENTSI would like to thank John Krebs without whose effort andexpertise my own research would not have gotten off the ground.I would also like to thank Martin Geertsema for his assisstancewith the soil descriptions. Funding and logisitical support forthis research was provided by Mammal Pest Management, The B.C.Ministry of Forests, G. Lawson, J. Millar—Kane and Dr. G.Bradfield. I would like to thank Dr. Alan Slemon and Dr. RogerGreen of The University of Western Ontario for their suggestionsregarding the data analysis. Thanks also to Dr. G. Bradfieldand Dr. D. Schluter for their helpful comments on themanuscript.A special thanks to my parents, George and Grace Lawson.1CHAPTER 1 INTRODUCTION1.1 HERBIVORE-HABITAT INTERACTIONS: THE PORCUPINE PROBLEM INB.C.A special issue of Bioscience published in December 1988 con-tained a series of articles that reflected the emerging realiza-tion that the interactions between animals and their environmenthave long-term implications for ecosystem dynamics. The sig-nificant ecological roles that animals may play in communitiesgo beyond the immediate requirements for food and habitat. Inmany cases they are responsible for biogeochemical, successionaland landscape alterations that may persist for centuries (Naiman1988). Changes occurring at the community level have implica-tions on different spatial scales which reverberate throughoutall trophic levels, sometimes causing unpredictable alterationsto the dynamics of the system (Naiman 1988; Naiman et al. 1988;Pastor et al. 1987; Pastor et al. 1988). The substantial andlong-lasting consequences of these interactions to wildlifemanagement are just now being realized.There are a number of herbivorous mammals in North America whosewinter habits of feeding on the foliage and vascular tissues ofconiferous trees are of concern to the management of forestcrops. In the province of British Columbia an alarming amount ofcaused by the winter feeding of porcupines has been observed in2recent years. Damage has been reported over a wide range ofbiogeoclimatic subzones and environmental conditions althoughoutbreaks seem to be occurring mainly in second growth westernhemlock (Tsuga heterophylla Raf.(Sarg.)) and Douglas fir(Psuedotsuga menziesii (Mirbel) Franco) (Dodge and Cannutt 1969;Hooven 1971; Sullivan et al. 1986; Sullivan and Cheng 1989).The Prince George and Prince Rupert regions sustain the majorityof the damage in the province.It is possible that the recent expansion of porcupine popula-tions in the northern half of the province is the result oflogging and silvicultural practices in these areas. Suchpractices provide a mosaic of forest types, increasing thesuitable habitat for porcupines who seem to prefer earlysuccessional stages and second-growth stands over mature, old-growth forest (Golley 1956; Van Duesen and Myers 1962).Population outbreaks reported in California (Lawrence 1957;Yocom 1971; c.f. Sullivan and Cheng 1989) have occurred instands associated with recent logging, similar to the situationnow occurring in B.C. and the Alaska Panhandle (Sullivan et al.1986; Sullivan and Cheng 1989). Porcupines appear to preferfeeding on vigorous stems which are usually the dominant andcodominant trees in natural and managed stands (Sullivan et al.1986; Sullivan and Cheng 1989). This damage may have severeimpacts on the forest industry, leading to a state where someareas do not have enough merchantable timber, or lengthening the3rotation time to reach merchantable volume. This would impacton the annual allowable cut (Sullivan and Cheng 1989). There isalso evidence that silvicultural activities such as thinning andfertilization may increase the susceptibility of trees to attack(Sullivan and Cheng 1989).Most fundamental to wildlife management is an understanding ofthe interactions between an animal and its environment (San-derson et al. 1979). The demonstrated complexity and multi-dimensional nature of these interactions necessitates theconsideration of many variables when assessing species-habitatrelationships and highlights the inappropriateness of onedimensional approaches (i.e. measurements of response to onresource such as food, protective cover etc.). The developmentof multidimensional treatments of habitats was concurrent withthe adaptation of multivariate statistical methods to ecology.This has made possible the application of multidimensionalthinking to field studies of species environment relationships(Carey 1981), as discussed by Green (1971, 1974). Many ecolo-gical problems involve numerous variables and numerous in-dividuals or samples. Community data are multivariate becauseeach sample site is described by the abundances of a number ofspecies and because numerous environmental factors affectcommunities. Multivariate analysis of community data cannotreplace experimental manipulation, but neither can experimen-tation replace multivariate analysis (Pimentel 1979). Resear-4chers in community ecology and wildlife management are in-creasingly using multivariate statistical procedures to identifymeaningful habitat variables.It was the purpose of this thesis to evaluate the utility of amultivariate approach in the consideration of porcupine-habitatrelationships in the north coast region of British Columbia.Practically, it was hoped that the identification of variablesimportant in determining the susceptibility of particular foreststands to porcupine attack would allow for recommendations to bemade regarding damage reduction measures. Such informationwould also, perhaps, lend some insight into the generalrelationships between herbivores, their food resources, andother aspects of their habitats.This chapter includes a brief introduction to the ecology of theporcupine followed by a review of the relevant research to dateregarding the habitat selection and feeding preferences of theseanimals. This information was necessary in the selection ofpertinent factors to be included in this study. Finally thespecific objectives of this thesis are stated.Chapter two, Methods, describes the study site, and variablesmeasured and the methods of data analysis. Chapter three,Results, presents the results of the analyses, and chapter fourfollows with a Discussion of these results.51.2 GENERAL PORCUPINE ECOLOGYThe porcupine (Erithizon dorsatum Allen) (Figure 1.1) is asolitary animal, occurring in a number of forest types acrossCanada, southeast Alaska and the northern and western UnitedStates (Bannan 1974). Porcupine densities are generally low,but numbers in an area tend to fluctuate greatly depending uponthe season and the availability of food. There are no reliableestimates of porcupine densities for the north coast of BritishColumbia. Nevertheless, this animal, which lives only 5 to 10years and produces only one offspring per year, has demonstrateda surprisingly ability to attain high densities (Sullivan andCheng 1989).Porcupines are strictly herbivorous, feeding on grasses andherbs duing the summer months and moving up to 10 km in theautumn to search for food and cover or denning sites (Sullivanet al. 1986; Sullivan and Cheng 1989). These seasonal migra-tions could be due to the requirement for den sites, shelter orfood. The switch in habitat is accompanied by a switch in diet.During the winter porcupines feed on the vascular tissues, thephloem and cambium, of coniferous trees and hardwoods. Winterhome ranges are reported to be between 0.1 and 12.1 ha andprobably vary according to the availability and quality of food,protective cover etc. In northern coastal climates, most winterfeeding occurs near den sites and often on the same trees in6Figure^1.1.^The^porcupine,^Erithizon^dorsatum.7successive years (Sullivan et al. 1986; Sullivan and Cheng1989).Porcupine feeding during the winter months causes broad promi-nent, horizontal and diagonal incisor marks on exposed sapwood(Figures 1.2a and b), usually concentrated in the upper parts ofthe tree. Complete girdling of the tree results in death of thestem above the point of injury, leaving the tree with a charac-teristic 'spike top' (Figure 1.3). Partial girdling can greatlyaffect the vigour of the tree for up to ten years as has beenshown in the case of ponderosa pine (Storm and Halverson, 1965).The physiological stress of lost vascular tissues and increasedvulnerability to attack from insects and diseases probablyaccounts for the reduced vigour of semi-girdled trees.Porcupine damage is significant in many forested areas of NorthAmerica. Damage is common in the coniferous-hardwood forests ofthe northeast and north-central United States (Curtis 1941,1944; Rudolph 1949; Cook and Hamilton 1957; Krefting et al.1962) and in the dry western forests (Taylor 1935; Curtis andWilson 1953; Lawrence 1957; Van Deusen and Meyers 1962; Stormand Halverson 1967; Hooven 1971). Damage has also been reportedin the western hemlock - Sitka spruce (Picea sitchensis (Bong.)forests of southeast Alaska (Meehan 1974; Ruth and Harris 1979).In Canada it has been reported that porcupine damage hasoccurred to limber pine (Pinus flexilis James) and Douglas-fir8Figure 1.2. The results of porcupine feeding on hemlock.Figure 1.3. The results of porcupine feeding on hemlock; ce+°P ' •910in southern Alberta (Gill and Cordes 1972; Harder 1979), and towhite spruce (Picea glauca (Moench) Voss), balsam fir (Abiesbalsamea (L.) Mill.) and other species in the Maritimes (Reeks1942; Radvanyi 1953; Speer and Dilworth 1978), and juvenilelodgepole pine (Pines contorta Dougl.) in central BritishColumbia (Sullivan and Sullivan 1982a).Surveys have indicated that porcupine damage in the northernhalf of B.C. is presently increasing due to expanding popula-tions of these animals (Sullivan and Cheng 1989). The avail-ability of suitable winter habitats is probably most crucial tothe survival of porcupines. Bark is not very nutritious and thequality of this food source will depend upon the age, speciesand condition of the tree being utilized. Porcupines have beenfound dead from starvation during the winter, yet with a gutfull of bark. Thus, there may only be one point in the year, a'critical period' (Lack, 1954), when porcupine populations arelimited by a shortage of food.1.3 HABITAT SELECTION AND FEEDING PREFERENCESThe selectivity of porcupine feeding behaviour is an importantconsideration before recommendations can be put forward fordamage reduction methods. Harder (1980) points out that theselection of individual trees by porcupines could occur byeither of two processes: a) porcupines may seek out an11appropriate community and then, having found one, search for anappropriate patch within this and, finally, decide upon anattractive tree; b) alternatively, they may simply search fora particular species of tree.There exist published accounts which characterize these animalsas selective feeders (process b). Harder (1980) comments,however, that it is difficult to assess the accuracy of manyreports for three reasons: data were not provided regarding theavailability of tree species; preference rankings weresubjective (Gabrielson 1928; Taylor 1935; Curtis and Kozicky1944; Shapiro 1949; c.f. Harder 1980); the data presented werenot in support of conclusions (Krefting et al. 1962; Gill andCordes 1972; c.f. Sullivan 1986a). Porcupines do, however,demonstrate preferences on the basis of tree species in someareas (Curtis 1941; Rudolph 1948; Brander 1973). One study byTennison and Oring (1985), which addressed the interaction ofporcupine damage and a number of inter- and intra-communityvariables, found that used versus unused areas differed withrespect to tree species composition. In a study by Roze (1984)individuals had narrower diets than the population as a whole,each animal specializing, at least temporarily, on one or twospecies.Some studies have indicated that habitat selection proceedshierarchically (process a).^Landscapes are characterized by12spatial gradients of habitat variables such as elevation, slope,soil moisture and nutrient status. The prevalence of porcupinesin a particular forest area may be determined by the physicalfeatures of the area as much as, or more, than the tree typesavailable (Taylor 1935; Van Duesen and Meyers 1962). Speer andDilworth (1978) measured porcupine use and a number of habitatvariables in mixed coniferous/hardwood forests in New Brunswickand concluded that moist areas were preferred over more mesichabitats. In Harder's (1980) study in Alberta there were markedcommunity preferences for low density stands in leeward com-munities.Whether or not porcupines are proceeding in a hierarchical way,it is important to consider how individual trees within a standare selected. In Speer and Dilworth's (1978) study, everyspecies of tree was attacked which comprised more than 5% of thestand, with the exception of red maple (Acer rubrum); it wasattacked at even lower densities. Eastern hemlock (Tsugacanadensis), present in small numbers, was not utilized althoughother studies have found it to be the preferred food where itwas common. Thus, porcupine winter food may vary at the scaleof the stand such that feeding may occur on any speciescomprising a substantial part of the stand. Roze (1984),working in New York state, also found that the primary food ofporcupines was the most abundant food. Obviously, this wouldincrease foraging efficiency since less time would be spent13searching for specific food types (MacArthur and Pianka 1966;c.f. Roze 1984).Harder (1980) concluded that porcupines demonstrated a lack ofpreference between the coniferous species comprising thecommunities he studied. He argued that in the preferred standstrees were larger, taller, and generally younger, and suggestedthat if porcupines select on the basis of size and/or vigour,the collective population response would be manifest inintercommunity preferences similar to those observed. The sizeof trees fed on by porcupines ranges from 3 cm dbh to almost 90cm dbh but the bulk of feeding occurs on pole-sized trees from7.6 to 38.1 cm dbh. A predilection may exist in some cases fora particular size class of tree regardless of geographicallocation or species of tree. Perhaps, as proposed by Curtis andWilson (1953), porcupines are most adept at climbing ormaintaining feeding positions in this size of tree. Severalstudies have concluded that relatively large, fast growing treesare preferred (Rudolph 1949; Curtis and Wilson 1953; Krefting etal. 1962; Spencer 1964; c.f. Harder 1979). Vigorous trees withlarge open crowns produce a large annual increment of phloem andprovide more foliage than crowded trees (Bannan 1955; Grillosand Smith 1959) and provide more large branches allowing greateraccess to food. Porcupines may be identifying vigorous trees bytheir obvious physical features (Harder 1979).14The implications to the silvicultural industry of such selectivefeeding are that practices such as thinning may predispose thetrees to attack due to their increased vigour and accessibility.Porcupine preference for crop trees in thinned stands has beenreported by Van Deusen and Meyers (1962), and Dodge and Canutt(1969) in the Pacific northwestern U.S., by Eglitis and Hennon(1987), in Alaska, and by damage surveys in the Kalum Valley ofB.C. by Sullivan and Cheng (1989).1.4 RELEVANT STUDIES IN THE PACIFIC NORTHWESTSelectivity in feeding on the basis of size and vigour has beenreported in the Khuzeymateen Inlet on the north coast of BritishColumbia (Sullivan et al. 1986). Studies have concluded that insecond growth stands, feeding damage, measured as the averagenumber of wounds per tree based on crown class, was the highestfor dominant and codominant trees which were usually larger thanthose in intermediate or suppressed crown classes. This patternwas also recorded for the percentage of trees completelygirdled. Damage peaked (83.3%) among second growth hemlock inthe 27.5-32.4 cm dbh class. Sitka spruce sustained less damagethan hemlock, while amabilis fir and western redcedar were notattacked (Sullivan et al. 1986).These conclusions are supported by studies of porcupine feedingdamage in Alaska (Eglitis and Hennon 1986, 1987) and elsewhere15(Curtis 1941; Rudolph 1949; Curtis and Wilson 1953; Krefting etal. 1962; Van Deusen and Meyers 1962). Comparable damage pat-terns have been observed in studies of red squirrel feeding onjuvenile lodgepole pine (Sullivan and Sullivan 1982a and b;Sullivan and Vyse 1987). Spencer (1964), Storm and Halverson(1967), and Harder (1979) have also reported a preference byporcupines to feed on vigorous stems for various species ofconifers in the interior of western North America (Sullivan andCheng 1989).In the Kalum Valley of B.C. (Prince Rupert District) the annualrate of attack has increased from 0.6% in 1986 to 1.8% in 1987for western hemlock, which is the major species in managedstands, whereas damage to Sitka spruce has remained constant.Lodgepole pine is also attacked were it grows (Sullivan andCheng 1989).Severe damage has been incurred by both western hemlock andSitka spruce in managed stands on Mitkof Island in Alaska andthe degree of attack may depend on their relative abundances ina given managed stand (Eglitis and Hennon 1986, 1987). Sullivanand Cheng (1989) comment that this severe example may representthe future situation for the Kalum Valley.161.5 OBJECTIVESThe overall objective of this thesis was to evaluate therelationships existing between porcupine damage due to thewinter barking of trees and relevant vegetational and environ-mental variables in managed and unmanaged stands. Threespecific questions were addressed:1.) Do thinned areas sustain more damage than unthinnedareas?2.) Are porcupines choosing on the basis of a communitytype (a suite of properties such as manifested in thevegetation, soil nutrient status, etc.), or on thebasis of the species, size or vigour of individualtrees?3.)^Which variables or combination of variables best predict the amount of damage that an area will incur?Answers to these questiions are necessary before considering theappropriateness and efficacy of damage reduction measures. Itis hoped that the study would also contribute to the under-standing of relationships between herbivores and the communitiesin which they exist.17CHAPTER 2 METHODS2.1 STUDY SITEThe Kalum Valley District (Figure 2.1) was selected as anespecially appropriate area within the Prince Rupert District inwhich to carry out the study. This is an extensive area ofrelatively young (12-15 years) managed stands. Silviculturalwork has been put into approximately 9,086 ha and another 48,560ha have been reforested at a total investment of 68.6 milliondollars. These stands are now entering the susceptible stage(15-35 years) for intensive porcupine attack. Damage to westernhemlock has increased three fold over the last two years. Anadditional 32,000 ha is proposed for stand-tending over the nextten years (Sullivan and Cheng 1989). Figure 2.2 illustrates thegeneral appearance of the forests in the area of the studysites. These forests are within the Coastal Western HemlockZone, Northern Drier Maritime Subzone (CWHf) (Figure 2.3).Four blocks of forest in the Kalum Valley, 100 ha each, wereselected on the basis of walk-throughs and forest cover maps.They were located on the north side of the Kitsumkalum River,north of Terrace, B.C. (Skeena Cellulose TFL #1) as shown inFigure 2.4. These blocks were similar in species composition,topography, elevation, height class of trees and years sincelogging (Table 2.1). All four areas had regenerated naturally4550u^ 1033,5--:.^• ,^• 'rea"v 2459'yy^. '^01770 -^kCs ‘-'2653ICOW, .d01101HoganSO5080,immie Mt 'Rimer^BirnieTERRACEFirllayso59.rgetern —Gee ,,471.00.5 -31'^50'^40'^30'^20'^10'`51i.phens , • •MiroV 4f A ,PrescottT • e)^Island^•Larrerer, ;,,ickingNORTH ORCNlRwatisrnif^ 2450'^,''•4 • .1.•i'4^te• RP-let-ANON\Sa^'PON lLl?50 ^1 rriote Is^ :^tak^r.^ Lucy 4 ,,o DibyTree Nob ,P.Islan0,004Archibald^ Kinakan9iclleYRachel lap^IsQ^ Lelu IMetI ugwellfSIMPS [Ag1altielt340. reeM:e vatc •rr-a ct i a8PRINCE RUPERTAKen,n,ady2:150IrS orsey4t7335°I''^385'f,:3650',a^38.50t m°4 01,3"'n.P.CP. •^01.40 (-.7",(\\,_a544130°40,50)0a •500450,1,50'^40'e.11250 -0004650'55503055517.• 6350 - (555850 020'5050•-•565050we Lek.' craS^ .,__---.■-...,-,--"--72 ,--'^ , —-^-1--,)^7 --r"^5750,- :-F ." e,.,!  ^.)...,, ^,.....)--- -,..7i;^5623:--^2350 `. "" \\.\ \ s.:50 ...0., ^i^ N' .---. 7..,,,-„f--__:_ 45p^P^s ^\^1^L•P--,.-- — --,0_, Ø*1. /^,÷^\ ....^ 6t,'^Ll DVIsii., ,A ...P•"—", 11.93^"or 4:,.. ^\ '^*-ls,^'.^i___ .6250 1.Kama0150Nllimal M1—$coL00-601` ^6450--:— ' •rake^Pi^650=76* L110'^129° SO'^40'^30'^20'10'Figure 2.1. Location of the Kalum Valley in north coastal B.C.(yellow)Trentlm Pt Z . 2 .^T-414 e-^&Jr:671_^r) pear-a,ANCE%Port Edward. 41cm—AAIFigure 2.3. The biogeoclimatic zones of the north coast of B.C.Green represents the coastal western hemlock zone.. .';otteSandipt21Figure 2.4. The location of the four grids on the north sideof the Kitsumkalum River.22after deforestation and two of the areas had been thinned(spaced several years previous to this study) in an effort toreduce the effects of competition between crop trees. Figure2.5a presents an example of the typical appearance of an areawhich has been spaced compared to Figure 2.5b, an unspaced area.Table 2.1. Description of the four study areas.Grid Location Species Class Logged SpacedA 128 44'E 34'N Hemlock/Balsam 1960 1984B 128 45'E 38'N Hemlock/Balsam 1956/58 1980/84C 128 43'E 36'N Hemlock/Balsam 1958 notD 128 50'E 50'N Hemlock/Balsam 1957/60 notUsing hip string and a compass, one 100 ha grid was laid out andflagged in each forest block, the exact shape of the grid beingdependent on the topography. Stations were flagged every 100m,the intersection of grid lines being possible plot locations.Twenty-five sampling plots were then randomly selected on eachgrid. Twenty-one intersections on grid C had to be excludedfrom the random draw because they were discovered to be unspacedeven though they were classified as spaced on the forest covermaps. Two plots were excluded on D and 5 on B because they werein swamps. Two more on B were excluded as they fell in gravel22Figure 2.5 a. The general appearance of a thinned forest area.Figure 2.5 b.^The general appearance of an unspaced forestarea.2425pits. Plots that landed directly on a skid road or on an oldlanding area (ground severely compacted) were offset by 15m atright angles to the road.2.2 STAND STRUCTURE AND DAMAGE ASSESSMENTSAll damage data were collected in May, June and July of 1989.Data was recorded on rainproof Cruise Tally Sheets (Province ofB.C., MOF). Trees greater than 10 cm dbh were selected using arelescope (Bitterlich, 1985) (Figure 2.6). The number of bandsused in the relescope viewfinder (determining basal area fortrees in to be included) was held constant in each area in orderto achieve an average number of trees/plot of between 10 and 12.Thus, the size of the sample plots for assessing damage and treespecific variables varied dependent on tree density, but wasconsistently larger than 50m 2, the area of the vegetation andsite description plots (see below). Only live trees wereincluded.Each tree was tagged and the following variables recorded:species; dbh (diameter at 1.3 m above point of germination) tonearest 0.1 cm; crown class, ranked from 1 to 3 (1=dominant;2=codominant; 3=intermediate/suppressed); and tree height. Theheight of trees up to 12 m was measured using a height pole.Trees taller than 12 m were measured using a clinometer andmeasuring tape. In this case the height was calculated4■111■11aIFT25Figure 2.6. The use of a relescope to select sample trees.27according to the formulaHeight = (B+T)(cos(tan -IB)SD))+correctionwhere B is the angle of the tree bottom, as seen throughthe clinometer, T is the angle of the tree top, SD isthe distance from the clinometer to the tree at breastheight in m, and Correction is the height person at bottomof tree, since clinometer measurements were taken to topof the head.My original intention was to calculate the growth form of eachtree as the ratio between its height and dbh. However, so manyof the damaged trees had dead tops, reflecting the extent ofgirdling, that this was not possible.Basal area/ha. and number of stems/ha were calculated at thegrid level and at the level of the plots. These variables wereconsidered to be important indicators of stand structure andpossible factors in porcupine choice. They are properties whichmay be manifest at the level of the whole stand or a smallercommunity level, such as the plot.Porcupine damage data were obtained by climbing nearly all thetrees in the plots (Figure 2.7) unless they could be assessedfrom the ground. For each tree I recorded the number of new(1988-89 winter) and old (prior to 1988-89) damage wounds andtheir position on the stem, in thirds (lower, middle, upper).2'6Figure 2.7. Climbing a hemlock to assess damage.29I also recorded the circumference of stem girdled (classes: 1-25, 24-49, 50-75, 76-99, and 100%) of the most severe wound ineach third of the tree. The total number of wounds and thetotal wound circumference was computed as the sum of the valuesfor each third of the tree.2.3 HABITAT DESCRIPTION2.3.i VegetationAll vegetation and site description data were taken in a circleof 3.99 m radius (50m 2 ) positioned at the center of the damageplots (Figure 2.8). Data were recorded on standard, rainproof,ecological classification reconnaissance forms (Province ofB.C., MOF).The percent ground cover of all species of herbs and shrubsoccurring within the plots was recorded and coded as shown inTable 2.2.Comparison charts were used in the field and trial plots weredone to establish consistency.Distribution codes were also recorded for each species but werenot included in the analysis.30Table 2.2. Percent cover codes for plant species data. AfterWalmsley et al. (1980)1 - present outside of plot2 - 0-1%3 - 1-5%4 - 5-25%5 - 25-50%6 - 50-75%7 - 75-100%8 - 100%Plants were identified using Hitchcock and Cronquist (1973),Coupe et al. (1982), Lyons (1952).Strata coverage, the percent cover of trees, shrubs, herbs,mosses and lichens was also recorded as per Klinka et al.(1981). These variables were treated as a separate data matrixfor many of the analyses.2.3.ii Site DescriptionThe variables measured for the complete site description arepresented in Table 2.3.^The methods of description follow31those given in 'Describing Ecosystems in the Field' (Walmsley etal. 1980), unless otherwise stated. The gradients described forthe variables were tailored to the particular biogeoclimaticsubzone in which the study was carried out.A soil pit, usually about 2 feet deep, was dug at each plot(Figure 2.9) and a complete soil description was recorded. Thesoil descriptions used to designate the Humus Form Class werealso used in the evaluation of the variables Ecological NutrientRegime, Moisture Regime and Drainage.Table 2.3.^Description of variables recorded for the sitedecription (after Walmsley et al., 1980).SLOPE - in degrees, measured using clinometer.ASPECT - the direction perpendicular to maximum slope,measured using compass.SITE POSITION - Upper, Mid or Lower, receiving SlopeSURFACE SHAPE - Categories apply within Site Positionand refer to the surface profile of the site, concave,convex or flat.32Table 2.3 continuedMICROTOPOGRAPHY - categories refer to the variability of thesurface of a site:1. Smooth^- no mounds2. Microrounded^- mounds less than 0.3m high3. Slightly Mounded - mounds 0.3m to lm high, over 7mapart4. Moderately Mounded - mounds 0.3m to lm high, 3m to 7mapart5. Strongly Mounded^^- mounds 0.3m to lm high, im to 3mapart6. Severely Mounded - mounds 0.3m to im high, 0.3 to 1mapart7. Extremely Mounded - mounds more than lm high and 3mapart8. Ultra Mounded^- mounds more than im high and lessthan 3m apart.ECOLOGICAL MOISTURE REGIME (Hygrotope) - Categories wererelative to the macroclimatic conditions of thebiogeoclimatic subzone; regime signifies the actual amountof moisture available for plant growth, and integrates manyinterrelated environmental and biotic variables (Walmsleyet al. 1980). The field assessment was completed by33evaluating a combination of soil properties, physical sitefactors and indicator plants (Klinka et al. 1989).0. Very Xeric1. Xeric2. Subxeric3. Submesic4. Mesic5. Subhygric6. Hygric7. SubhydricECOLOGICAL NUTRIENT REGIME (Trophotope) - Scale was ap-propriate to climatic conditions of the biogeoclimatic zone;regime signifies on a relative scale the nutrient supplyavailable for plant growth (Walmsley et al. 1980).Integrates many environmental and biotic parameters which,in combination, determine the avialable nutrients.Trophotope was evaluated by a qualitative examination ofsoils and indicator species (Klinka et al. 1989).1. Oligotrophic (Very Poor)2. Submesotrophic (Poor)3. Mesotrophic (Medium)4. Permesotrophic (Rich)5. Eutrophic (Very Rich)34Table 2.3 continuedSOIL DRAINAGE - Seven classes assessed from topography,position, vegetation and soil characteristics as perWalmsley et al. (1980).1. Very Rapidly Drained2. Rapidly Drained3. Well Drained4. Moderately Well Drained5. Imperfectly Drained6. Poorly Drained7. Very Poorly DrainedHUMUS FORM CLASS - Determined from the soil profiles as perKlinka et al. (1981). Humus fell into 2 orders: Mors andModers, and was thus coded according to the 14 groups ofhumus taxa within these orders. The procedure is tooelaborate to describe here. For details see Klinka et al.1981. Humus Form Class was considered to be a continuousvariable since it was coded in such a way that it providesfor segregation along the soil moisture gradient and, moreloosely, the soil nutrient gradient. The Mor orderencompasses the least biologically active humus forms of thetwo orders, with fungi being dominant in the upper profile.35Moders, however, have soil fauna dominant in the upper soilhorizons and provide generally more available nutrients thanMors (Klinka et al. 1981). Soil analysis was performed, orconfirmed, by Martin Geertsema, a pedologist with the MOF,Smithers, B.C.SURFACE SUBSTRATE - The percent cover of the followingvariables was recorded: Humus, Dead Wood, Bedrock, Rocks(includes rocks >7.5 cm in diameter, which may be covered byan organic layer <2 cm deep, moss or lichen), ExposedMineral Soil and Standing Water.2.4 DATA ANALYSISMultivariate analysis examines numerous variables simul-taneously, summarizing the data and revealing its structure.One of the main purposes of this research was to exploremultivariate methods as they apply to a wildlife managementproblem. A number of exploratory procedures were examined andseveral were chosen as particularily useful.There was an overlap between the Objectives and in the datarequired to address them and, thus, in the methods of analysis.First, though, it was necessary to summarize and to simplify36-41W1017:,%:4010-Te, - •Figure 2.8. A typical plot.Figure 2.9. An example or a soil pit__37some of the data and to compare the areas on the basis ofdamage, site and vegetation variables. Analyses were performedusing BMDP, SAS and the Systat package software.The general approach used in analysing the data was to doparametric tests because of the greater power they afford. Thisassumes that the variables have a normal distribution which maynot be the case considering the nature of the variablesinvolved. However, sample sizes were such (80-100) that mosttests would be robust. Nonparametric tests were used to confirmanalyses wherever possible and results were unaltered.In biological situations, as means increase so often do vari-ances. Significance tests for differences between variables areextremely powerful, and may pick up even minor differences(Pimentel 1979). Log transformations often tended to renderdata more normally distributed and to equalize variances butresults were generally unchanged. Also, I was not interested inthe prediction of, for example, log(damage). Hence, I presentresults based on the untransformed variables. I did not detectany appreciable non-linear relationships between variables, andso I did not correct for it by transforming data.2.4.i Stand StructureNested analysis of variance procedures were used to test for38differences between treatments (thinned and unthinned) and gridswithin treatments in the density (stems/ha.) and the basal area(m2/ha.) of trees. These estimates were calculated (as perBitterlich 1985) for trees of all species in the plots, and thenfor hemlock only as it was the only species utilized byporcupines. The ANOVA model predicts the dependent variable foreach grid by a sample mean. The difference between the actualand the predicted response is the residual error. It is the sumof squares of residual errors that is minimized by the procedurein the fitting of parameters. Because grids are nested withintreatments the total sum of squares (SS) can be partitioned intoa treatment SS, a between grids within treatment SS , or nestedfactor, and a within grid SS. No interaction can be obtainedbecause grids are not completely crossed with treatment, but arenested under treatment. One-tailed tests were used here sincethe a priori expectations were for lower values in the treatmentareas.The estimates of stems/ha. and basal area (m 2 /ha.) were derivedfor each individual plot for use in regression against thedamage index. For this purpose, plots with no hemlock treeswere excluded from the analyses. In these plots damage was,obviously, undefined.Differences between treatments, between grids within treatmentsand between all four grids in the size structure of the trees39(mean dbh) were assessed using nested analysis of variance.Mean dbh was also calculated at the plot level for use inregression against damage.2.4.ii Damage DataIt was necessary to summarize the damage to trees in order tocompute a measure of damage on a 'per plot' basis. Two aspectsof damage had been measured, the number of wounds per hemlockand the circumference of the most severe wound in each third ofthe tree. For each plot I computed the 'number of wounds perhemlock' (WPH) as the total number of wounds on hemlocks dividedby the number of hemlocks present. I also calculated for eachplot the 'circumference girdled per hemlock per plot' (CPH) asthe sum of the circumferences of wounds on all hemlocks presentdivided by the number of hemlocks.The variables WPH and CPH were highly correlated and so asummary measure of damage per plot was derived that retained asmuch information as possible but that was free from redundancy.Principal components analysis (PCA) was performed on the cor-relation matrix of the untransformed damage variables WPH andCPH. In a PCA the original variables are transformed to vari-ables that have zero intercorrelations. The transformationrotates the original axes but maintains the original relation-ship among data points (Pimentel 1979). The new PC-axes are40linear combinations of the original variables with coefficientsequal to the eigenvectors of the correlation matrix. The firstprincipal component was used as the new damage variable,referred to as 'PC1D total', 'PC1D old' and 'PC1D new',depending on whether it was calculated on the basis of total,old or new wounds.This summary variable, PC1D, was used in an ANOVA to test fordifferences between grids in the amount of damage incurred.It was the intention to test the degree to which damage could bepredicted from tree specific variables (species, dbh, crownclass and growth form) using analysis of variance and regressionprocedures. However, because of the resulting stand structure,and the fact that so many of the damaged trees had dead tops, itwas not possible to evaluate crown class.2.4.iii Site and Vegetation DataOf the site and vegetation variables measured, I selected thosethat occurred in at least 25% of the plots as these wereconsidered to be most useful to this applied management problem.Rarer plants or site characteristics, though they may sometimescorrelate highly with damage in the few plots in which theyoccurred, would not be useful for the future practicalprediction of threatened areas because of their rarity.41A multivariate analysis of variance (MANOVA) would have been thepreferred method for detecting the effect of treatment on thetotal set of site and vegetation variables, but this was notpossible due to a lack of degrees of freedom (so many variablesand only two treatments). Therefore, univariate ANOVAs wereperformed for each of the variables separately. The probabilitylevel (0) for significance tests was set to 0.05 divided by thenumber of univariate ANOVAs performed (Bonferroni standardprocedure). There were a resulting 11 variables in both thesite and vegetation submatrices after those variables occurringwith a frequency of less than 25% were dropped. Thus, P=0.004.A nested MANOVA was used to test for differences between gridswithin treatments for sets of site, species and vegetationstrata variables separately. The methodology of MANOVA is thesame as that of ANOVA, however, since multiple measurements havebeen made in each plot, the method is multivariate.Finally, I attempted to explore the relationship betweenvegetation data the site data using canonical correlationanalysis (CCA). The strategy of CCA is to search simultaneouslyfor linear combinations of each set of variables (the canonicalvectors) such that the correlation between the sets is maximized(Cooley and Lohnes 1971; Gittens 1985). The variables with thegreatest contribution to the canonical vectors were determinedby their standardized correlations with the canonical variates.42A redundancy analysis was included in the CCA. Although thesquared correlation coefficient is a measure of the overlapbetween canonical variates it may not be a very good indicatorof the importance of the linear combination of variables, sincethese may not account for much of the variation in the originaldata sets (Gittens 1985). The redundancy associated with acanonical variate is a useful index of the predictive orexplanatory power of each canonical variate in relation to thevariation in the opposite set of variables. Redundancy is theproportion of total variance in one domain (set of variables)that is predictable from a linear composite of the other domain,given the availability of the second domain (Gittens 1985).This quantity is arrived at if the variance extracted by the kthcanonical variate is multiplied by the squared canonicalcorrelation coefficient which expresses the proportion of thevariance of one of a pair of canonical variates shared by theother.2.4.iv Relating Damage to Site and Vegetation DataThis part of the data analysis most directly addresses thesecond and third objectives stated in the Introduction: Objec-tive 2) By what process are porcupines selecting habitats andtrees and 3) What variable or combination of variables bestpredicts the amount of damage an area will incur. Similarmethods of analysis were employed to address these objectives43because of the obvious degree of overlap.I initially used CCA in an attempt to relate the damage vari-ables WPH and CPH to site or vegetation variables, because thisapproach was considered useful for detecting associations.However, this was abandoned because the canonical vector ofdamage variables proved not to be an index of damage, but anuninterpretable ratio of WPH and CPH. Hence, regression usingthe damage index PC1D was then used as the alternative.For the purposes of regression, PC1D was considered to be thedependant variable. Stepwise regression was considered but notemployed because it was found that the set of variables includedby the stepwise selection procedure was highly sensitive to theorder in which they were added. Variables added last to themodel were not themselves related to damage, but were simplyselected relative to the other variables already included.Multiple regression was used instead. Regressions were perform-ed within thinned and control grids as treatment had an effecton the damage index and on the stand structure.Another method explored because of the possible applications towildlife management was Canonical Discriminant Analysis (CDA).Given a classification variable and several quantitativevariables, CDA derives canonical variables (linear combinationsof the original quantitative variables) that summarize variation44between classes. On the basis of a graph of WPH against CPH theplots were coded as low or high damage areas. CDA was used totest if these two categories of total damage could bediscriminated on the basis of the complete subset of site,species and vegetation strata variables within the controlgrids.45CHAPTER 3 RESULTS3.1 STAND STRUCTUREA summary of the number, species of trees sampled, and treesincurring damage, is presented in Table 3.1. Of 884 treessampled, almost all were western hemlock or amabilis fir. Thecalculated mean and standard errors for the estimates ofstems/ha. and mean basal area/ha. for each grid are shown inTable 3.2.Table 3.1. Summary of the damage data. Total number of treesexamined, and numbers showing damage (in parentheses), in thefour grids. There were 25 plots per grid. Plots are variablein size, and No./Plot does not indicate tree density.SPECIESTotalGrid No.Trees Hemlock Fir Spruce RedcedarAverageNo./PlotA 259 75 (26) 180 (2) 4 0 10.2B 195 76 (35) 116 (3) 2 1 7.7C 224 125 (45) 99 (3) 0 0 9.0D 213 100 (40) 113 (1) 0 0 8.5Total 891 376 (146) 508 (9) 6 1 9.046Table 3.2. The estimated mean number of stems/ha., and basalarea (m2) /ha. with standard errors, for each grid. Means aregiven for trees of all the species counted and for hemlocksonly.STEMS/HA.^ BASAL AREAAll Trees Hemlock Only All Trees Hemlock OnlyGrid Mean S.E. Mean S.E. Mean S.E. Mean S.E.A 429.5 33.4 153.2 30.1 15.9 1.2 4.8 0.8B 384.3 43.3 179.5 36.5 23.5 2.2 8.7 1.6C 1100.1 110.2 677.1 92.4 27.4 2.6 15.4 2.6D 1946.1 271.2 1019.7 182.4 43.1 4.5 20.5 4.5The estimated total density of trees, and density of hemlocksalone, was lower on the experimental (thinned) grids (A and B)than on the control grids (C and D) (Table 3.2, Figure 3.1).This effect of treatment was statistically significant forhemlock density (nested ANOVA, F=15.76, df=1,2, P=0.029), andnearly so for the density of all trees (F=6.94, df=1,2,P=0.060). Thus, thinning did appear to reduce tree density.The trend was the same for basal area of trees (Table 3.2,47Figure 3.2) although it was significant for only hemlocks alone(F=12.36, df=1,2, P=0.036), but not for all trees (F=3.19,df=1,2, P=0.11).The variable dbh (tree breast height diameter) was consideredto be an important character of stand structure. Only hemlockswere considered for this analysis as they were almost exclu-sively attacked by porcupines (Table 3.1). A nested ANOVAcomparing treatments in the mean of this variable revealed nosignificant difference (F=5.86, df=1,2, P=0.136) although thetrend is towards larger trees in thinned grids (Figure 3.3).There was a highly significant difference between grids withintreatments (F=10.24, df=2,372, P<0.001). Tukey pairwise com-parisons indicated that all grids differed significantly fromone another in the mean dbh of hemlocks with the exception ofgrid C from D. Tests repeated on log transformed data gaveidentical results.3.2 INCIDENCE OF DAMAGEDamage was confined almost exclusively to western hemlock (Table3.1). Thirty-two percent of the hemlocks from the first gridwere damaged, 42% from grid B, 33% from grid C and 37% from gridD. Damage to amabilis fir was incidental and western red cedarand Sitka spruce were not attacked at all. This comparisonshows that incidence of damage was not noticibly greater in4+1a•A B C DCd2000Cl)0a)1000ED01500CdN0Z 1000a)c(i)5008EI0GridFigure 3.1. The mean estimated density (number of stems/ha.) andstandard error for all tree species (a) and for hemlocks only (b)in each of the grids (n=25/grid).25DC \j 20E1co-6031005(DE0B CGrid50773cNiE 40(Da'stea30JDscc•2 20780F-1049Figure 3.2. The mean estimated basal area (m 2/ha.) and standarderror of all tree species (a) and of hemlocks only (b) in each ofthe grids (n.25).270E2.5y0 2400E46 21L.OEces18aia)1550IA^B^C^DGridFigure 3.3. The mean and standard error of the diameter atbreast height of hemlocks in each _of the grids. All grids aresignificantly (P<0.001) different from one another with theexception of C and D (Tukey pairwise comparisons).- •0A^B^C^DGrid51Figure 3.4. The mean and standard errors of the number of woundsper hemlock in each of the grids.52thinned grids (A and B) than in unthinned grids.The incidence of western hemlock attack by porcupines during the1988-89 winter and prior to this period is shown in Table 3.3.The total incidence of damage was 39%, of which 15% were treeswith new wounds.The number of wounds per hemlock was considered as a measure ofporcupine feeding intensity. The number of wounds per hemlockwas also not greater in thinned grids (Figure 3.4; nested ANOVA,F=0.004, df=1,2, P=0.480). However, there was substantialvariation in the mean number of wounds per hemlock between gridswithin each treatment (F=3.65, df=2,83, P=0.03).The average number of wounds per hemlock with respect to treediameter class during and prior to the winter of 1988-1989 isillustrated in Figure 3.5. The number of old wounds accumu-lates steadily as the trees grow larger until the largest(presumably oldest) diameter classes, in which severely damagedtrees have probably died. A quadratic regression on the meannumber of new wounds against ordered tree diameter classindicated a significant trend (tree diameter class P=0.02, treediameter class 2 P=0.01) with damage peaking in the 25.1 cm to30.0 cm tree diameter class as shown in Figure 3.6.VF537654321s10 10.1- 15.1- 20.1- 25.1- 30.1- 35.1- >4015.0 20.0 25.0 30.0 35.0 40.0Hemlock diameter (cm)Figure 3.5. The mean number of wounds per hemlock with respectto tree diameter class. The solid bars are new (1988/89 winter)wounds and the hatched bars are old (prior to 1988/89 winter)damage.1^I^I5401.5NcoDc1.0c(i>00.5a)20.02.01Q1- 15.1- 20.1- 26.1-150 200 260 300 30.1- sal- >40350 4Q0Hemlock diameter (cm)Figure 3.6. Quadratic regression of the mean number of new(1988/89 winter) wounds per hemlock with respect to tree diameterclass. R2 =0.76; for DBH, P=0.016, for size 2' P=0.012.55Table 3.3. Number of western hemlock trees with porcupinedamage inflicted during the 1988-89 winter season ('New Dam-age'), and during previous seasons ('Prior Damage').Prior Damage^None^Number Damaged TotalNone^230^98^328New DamageNumber^22^26^48DamagedTotal^252^124^376The average number of wounds per hemlock, new and prior to the1988-89 winter, in each third of the trees with respect to treediameter class is shown in Figure 3.7a and b respectively. Ingeneral, the pattern of new and old wounds with respect to theposition in the tree appears similar. The results of threepairwise t-tests indicated significantly more wounds in the1.00.80E_cQ0. 10.1- 15.1- 20.1- 25.1- 30.1- 35.1- >4015.0 20.0 25.0 30.0 35.0 40.0Hemlock diameter (cm)Figure 3.7. The mean number of a) new (1988/89 winter) woundsper hemlock and b) old (prior to 1988/89 winter) wounds per hemlockwith respect to tree diameter class.^Position on tree:lertopthird, lir middle third, C,74: lower third.57middle third of the stems than in the upper or lower thirds(P<0.01 in each case, DF=375). Fewer wounds were recorded inthe upper stems in cases where the trees were stripped almostcompletely because individual wounds could not be distinguished.It was not possible to evaluate the effect of crown class owingto the even-aged structure of these second-growth stands.Nearly all of the trees (296) fell into the codominant crownclass; only 14 were classified dominant. Although 65 trees wereclassified as intermediate/suppressed, they were represented bythe smallest size classes only, (<=20.0 cm dbh).3.3 DAMAGE INDEXA plot of the two variables 'wounds per hemlock' (WPH) and'circumference girdled per hemlock' (CPH) (Figure 3.8) indicatesthat the error in these measurements increases as the variablesincrease.The variables WPH and CPH were highly correlated (86%), meaningthat trees with the most wounds also had the most severe wounds.Principal components analysis was employed to create a summaryvariable of these two measures, free of redundancy yet retainingas much information as possible. The first PC-axis of thedamage variables accounted for 96% of the variation in the data(Figure 3.8). The variable contributions to this axis wereMean circumf. wounds per hemlockFigure 3.8. The mean number of wounds per hemlock versus themean circumference of wounds per hemlock calculated on a plotbasis.^The line shows the first PC-axis from a principlecomponents analysis of these two variables. The broken line is thedeviation of the observation from the PC-axis.5859relatively equal (component loadings, WPH= 0.63 and CPH=0.51).This PC1-axis was considered as a new variable (as in methods)and will be referred to hereafter as 'PC1D total', since it wascalculated on the basis of old and new wounds. The damagesummary indices calculated on the basis of new wounds and oldwounds separately were also calculated and are referred to inthe text as 'PC1D new' and 'PC1D old' respectively.There was no significant treatment effect for the total damagesummary index, 'PC1D total', (F=0.05, df=1,2, P=0.901) nor werethere any differences between grids nested within treatments(F=1.61, df=3,82, P=0.192) for this variable. There was nodifference between treatments in "PC1D old', the damage summaryindex calculated on the basis of old wounds (F=0.005, df=1,2,P=0.952; Figure 3.9a), as expected, since 'old' damage wasinflicted before the treatments were imposed, though there wasa significant difference between grids nested within treatments(F=3.34, df=2,83, P=0.040). However, there was significantlymore new damage in unthinned areas than in thinned areas(F=25.84, df=1,2, P=0.037; Figure 3.9 b) while there was nosignificant difference between grids within treatments (F=0.186,df=2,83, P=0.831). With regards to Objective 1, the hypothesisthat thinned areas incur more damage than do unthinned areas canbe rejected. In fact, the opposite would seem to be true. Withregard to Objective 2, it is possible that porcupines arechoosing at the scale of large blocks of forest perhaps on thea) old wounds060510.0A B C D1.8GridFigure 3.9. The mean and standard error for the damage summaryindex in each of the four grids, a) calculated on the basis of oldwounds ('PC1D old') and b) calculated on the basis of new wounds('PC1D new').61basis of properties such as stand density or the basal area oftrees. It is not likely that they are responding to thediameter classes of trees. Although porcupines demonstrate apredilection for a particular size of trees they are notattacking more trees in the thinned areas were the trend, thoughnot significant, is towards the preferred size classes of trees.Regression of damage, 'PC1D total', 'PC1D new' and PC1D old',calculated at the level of the plot on the stand structurevariables of density and basal area were performed for treatmentgrids and control grids separately. The results indicated nosignificant relationships using either the total number of treesor only hemlocks (Table 3.4). Porcupines are not, therefore,choosing at the level of the plot on the basis of theseproperties.62Table 3.4. Results of regressions of the damage index, calcu-lated on the basis of new, old and total wounds, and the density(stems/ha.) and basal area (m 2/ha.)of all tree species and ofhemlocks only. (For control grids, total damage, N=46, df=1,44;old damage and new damage, N=47, df=1,47. For treatment grids,new damage, N=40, df=1,38).NEW DAMAGECONTROL GRIDS TREATMENT GRIDSF-Ratio P F-Ratio PTotal Density 0.296 0.589 0.408 0.527Hemlock Density 0.691 0.410 2.181 0.148Total Basal Area 0.360 0.551 0.014 0.905Hemlock Basal Area 1.768 0.190 0.194 0.662OLD DAMAGETotal Density 1.767 0.190Hemlock Density 1.350 0.251Total Basal Area 0.821 0.370Hemlock Basal Area 0.004 0.953TOTAL DAMAGETotal Density 2.358 0.132Hemlock Density 1.045 0.312Total Basal Area 0.936 0.339Hemlock Basal Area 0.768 0.386633.4 VEGETATION AND SITE VARIABLESThe plant species present in at least 25% of the plots, andtherefore retained for analysis (see Methods) were: Athyriumfilix-femina, Clintonia unifoliata, Cornus canadensis,Dryopteris assimilis, Epilobium angustifolium, Oplopanaxhorridus, Rubus pedatus, Rubus spectabilis, Streptopusstreptopoides, Tiarella unifoliata, and Vaccinium alaskense.The site variables occurring in at least 25% of the plots were:Slope, Aspect, Surface Shape, Moisture Regime, Nutrient Regime,Number of Stumps, % Cover Humus, % Cover Dead Wood, Microtopo-graphy, Drainage, and Humus Form.Analysis of variance on the vegetation, site and vegetation stratavariables indicated that the only variables significantly differentbetween treatments were % Cover Humus (F=36.31, df=1,2, P=0.003)and % Cover Dead Wood (F=21.15, df=1,2, P=0.004). These resultsare probably reflect the amount of slash (downed wood) lying on theground in the thinned areas. The mean % Cover of Humus in thinnedareas was 55.76 and in unthinned areas was 71.35, since the humusin unthinned areas was not covered in slash. The mean % Cover DeadWood was, of course, higher in thinned areas, 42.00 as compared tounthinned areas, 26.76.A nested MANOVA was used to test for the effect of grids within64treatments for each of the site, vegetation species and vegetationstrata matrices. Multivariate test statistics of Wilks' Lambda,Pillai Trace and Hotelling Lawley Trace revealed no significanteffect for the site variables (Table 3.5). However, there was ahighly significant difference between grids within treatments forthe species variables as well as for the vegetation stratavariables (Table 3.5). There was a significant degree of dif-ference in the % cover of the herbs Cornus canadensis, Epilobiumangustifolium and the shrub Vaccinium alaskense between grids.This was also the case for the % cover of trees, shrubs and mosses.Canonical correlation analysis was used to relate the sitevariables and the vegetation variables. The overall CCA was sign-ificant as indicated by the multivariate tests of Wilks' Lambda(F=1.67, df=121,518, P<0.001), Hotelling-Lawley Trace (F=1.74,df=121,648, P<0.001) and Pillai trace (F=1.53, df=121,814, P<0.001). The first two canonical correlations were sig-nificant and accounted for 55% of the variation on the data (Table3.6). The adjusted canonical correlation coefficients of the firsttwo canonical correlations were 0.649 and 0.544 respectively (Table3.6). These values would indicate that the goodness of fit of oneset of canonical variates to the other is reasonably good, but notstrong.The site variable contributing most strongly to the determinationof the first canonical variable (Table 3.7) was Microtopography (0-.64). The species Tiarella unifoliata and Vaccinium65Table 3.5. The differences between the means of variables withinthe spaced and unspaced grids and the results of nested MANOVAstesting for differences between grids within treatments for thehabitat variables. Significance, P<0.05, is indicated by theastrix (*), and P<0.001 by the double astrix (**).MEANSSpaced^Unspaced^Univariate FSlope^ 0.46 5.63 0.71Aspect 59.75 44.35 4.34 *Surface Shape^0.36 0.35 2.36Mosture Regime 0.13 0.07 0.14Nutrient Regime^0.31 0.32 1.34Number of Stumps^0.05 0.74 1.16Humus^ 4.48 2.75 0.23Dead Wood 4.30 0.50 0.37Microtopography^0.07 1.09 2.94Drainage^ 0.35 0.49 1.31Humus Form 0.41 0.57 0.28Multivariate Test Statistic F-Statistic DFWilks' Lambda 1.36 22, 144Pillai Trace 1.35 22, 146Hotelling-Lawley Trace 1.36 22, 142Table 3.5 continuedAthyrium felix-femina 0.98 0.36 0.91Clintonia unifoliata 0.34 0.80 4.00Cornus canadensis 0.22 1.30 3.74^*Dryopteris assimilis 0.75 0.33 1.86Epilobium angustifolium 0.72 0.36 3.80^*Oplopanax horridus 0.44 0.01 0.40Rubus pedatus 0.27 0.71 1.88Rubus spectabilis 0.56 0.14 0.62Streptopus streptopoides 0.60 0.28 2.11Tiarella unifoliata 0.06 0.67 2.46Vaccinium alaskense 1.15 1.43 8.11^**Multivariate Test Statistic F-Statistic DFWilks' LambdaPillai TraceHotelling-Lawley Trace3.173.093.24^22,^144^* *^22, ^146^* *22,^142^**% Cover Trees 24.31 12.59 7.04 *% Cover Shrubs 14.20 5.41 3.14 *% Cover Herbs 1.90 7.24 1.17% Cover Mosses 13.30 10.43 3.94 *Multivariate Test Statistic F-Statistic DFWilks' Lambda 3.60 8, 158 **Pillai trace 3.53 8, 160 **Hotelling-Lawley Trace 3.67 8, 156 **6667alaskense were similarly the major contributors to the firstcanonical axis of the species variables, having correlations of0.44 and -0.56 respectively. Moisture Regime, Nutrient Regime,Drainage and Humus Form Class were most highly correlated withtheir second canonical variable. Athyrium felix-femina (0.72),Dryopteris assimilis (0.73) and Rubus spectabilis (0.67) were mosthighly correlated with their second canonical vector.The squared canonical correlation coefficients are also presentedin Table 3.6. but although this statistic is a measure of theoverlap between canonical variates it may not be a very goodTable 3.6 Results of the CCA using site and vegetation variables.Adjusted^Squared^Eigenvalue Proportion Pr>FCanonical^CanonicalCorrelation Correlation1 0.649 0.550 1.221 0.361 0.0002 0.544 0.431 0.759 0.224 0.0233 0.461 0.339 0.513 0.151 0.2534 0.423 0.269 0.367 0.109 0.6635 0.248 0.179 0.218 0.065 0.92768Table 3.7 Canonical correlations of the site variables on thefirst 2 canonical axes.Canonical Loadings 1 2Slope 0.376 -0.192Aspect -0.209 0.385Surface Shape 0.309 -0.082Moisture Regime 0.235 0.536Nutrient Regime 0.310 0.780Stumps -0.046 0.094% Cover Humus 0.108 -0.014% Cover Dead Wood -0.092 0.030Microtopography -0.639 0.422Drainage Class -0.220 0.585Humus Form Class 0.178 0.718Athyrium felix-femina 0.213 0.725Clintonia unifoliata 0.293 0.152Cornus canadensis -0.105 0.438Dryopteris assimilis -0.052 0.735Epilobium angustifolium -0.340 0.338Oplopanax horridus -0.049 0.252Rubus pedatus -0.138 0.295Rubus spectabilis 0.246 0.671Streptopus streptopoides 0.195 0.546Tiarella unifoliata 0.440 0.319Vaccinium alaskense -0.562 0.26869indicator of the importance of the linear combination of variablescomprising the variate to the original data set. Therefore acanonical redundancy analysis has been included (Table 3.8).The total redundancy value for the site variables and their firsttwo canonical variables is 0.13, indicating that the actual overlapbetween the two batteries of variables on the relevant canonicalaxis, as seen from the perspective of the species variables addedto an already available set of site variables, is not great. Theredundancy value is identical for the species variables. The totalredundancy value for the site variables and the species variablesand their opposite canonical variables is 0.06. Thus it could beconcluded that neither set of variables is an excellent predictorof the other set, though there is some relationship between thetwo, not surprisingly.3.5 RELATING DAMAGE TO SITE AND VEGETATION VARIABLESCCA was considered an appropriate procedure to address Objec-tives 2 and 3, by relating the damage matrix of variables WPH andCPH to the matrices of habitat variables. It was not carried far,however, because it was immediately evident that the firstcanonical vector of the damage matrix, which included the twodamage variables WPH and CPH, was not in fact a measure of damage.The canonical coefficients standardized by sample standarddeviations on the first canonical vector were 1.81 for WPH and -0.97 for CPH, an unusual negative relationship. This70TABLE 3.8 The results of the canonical redundancy analysis for thesite and species variables. Variances are standardized.Redundancy of the Site^Redundancy of theVariables and the^Species VariablesCanonical Variables of^and the CanonicalVariable ofTheir Own The Opposite Their own The OppositeR2^Set^Set^Set^SetVariate1^0.55^0.05^0.03^0.04^0.022^0.43^0.08^0.03^0.09^0.04Total^0.13^0.06^0.13^0.06procedure was abandoned for the purpose of relating damage tovegetation and site variables and regression was used instead.The results of the multiple regression of the damage indices, 'PC1Dtotal', 'PC1D new' and 'PC1D old' on the subsets of habitatvariables are presented in Table 3.9. Regression was not performedon the habitat variables of the treatment grids using 'PC1D old'or "PC1D total' as some of the old damage may have been incurredbefore the thinning took place. Only the regression of 'PC1D71total' was significant (R 2=0.49). The plant species with the mostsignificant partial correlation were Athyrium felix-femina(P=0.093) and Rubus pedatus (P=0.077). The regression coefficientsfor these two species were both positive, indicating that theyincrease in % cover as damage increases. This regression was alsoperformed using analysis of covariance and the results were unal-tered.Although it is impossible to determine cause and effect from thesedata, it might be expected that areas of high damage would beassociated with higher % covers of certain herbs since highlydamaged trees often had spike tops allowing greater amounts oflight to penetrate to the forest floor.One further analysis used to explore the relationship betweendamage and habitat variables was canonical discriminant analysis.Recall Figure 3.8, which plotted WPH against CPH. On the basis ofthe dispersion of these variables, plots were coded as havingincurred 'low' or 'high' damage, a value of 3 being chosen as thearbitrary cutoff point for the variables WPH and CPH (Figure 3.10).This categorical variable for damage, calculated on the basis ofboth new and old damage, was used in a CDA for relating damage tothe habitat variables in a 2-group canonical discriminant analysis.This analysis was performed only for the control grids and used thematrix which included all of the habitat variables. The multi-variate statistics of Wilks' Lambda, Pillai Trace and Hotelling-72Lawley Trace, all of which tested the hypothesis thatthe group means are equal, were significant (df=26,19, P=0.019).Table 3.9. The F-ratios and P values from the results of multipleregressions of the damage index and subsets of the habitatvariables.CONTROL GRIDS^TREATMENT GRIDSF-Ratio^P^F-RatioNEW DAMAGEVEGETATION 0.39 0.95 0.58 0.82STRATA 1.37 0.27 0.60 0.67SITE 0.85 0.60 0.98 0.50OLD DAMAGEVEGETATION 1.18 0.34STRATA 0.36 0.84SITE 1.02 0.45TOTAL DAMAGEVEGETATION 2.94 0.01STRATA 0.47 0.76SITE 1.52 0.17Figure 3.11 shows box-plots for the low and high damage categoriesagainst the discriminant function. The data for the 'high'category is obviously skewed as indicated by the asymmetry of theinterquartiles, indicating very little overlap in the discrimin-1^2 3 4 5 61 0000Ea)Q. 6DC40Caia) 2 0OMBhigh00^000 ^00Mean circumf, wounds per hemlockFigure 3.10. The mean number of wounds per hemlock versus themean circumference of wounds per hemlock calculated on a plotbasis. The shaded area designates plots that were classified ashaving 'low' damage and those outside of the shaded area weredesignated as having 'high' damage.734 74004^0--E°CCSCE -217_0SOC)ROM-4-6low^highDamage categoryFigure 3.11. Box plots of the damage class variable, 'PC1Dtotal' against the discriminant function. From the CDA of allhabitat variables (P<0.001).75ation.The variables significant in the discrimination between areas oflow and high damage were Stumps, % Cover Herbs, % Cover Athyriumfelix-femina, Rubus pedatus, Streptopus streptopoides and Tia-rella unifoliata (Table 3.10). As indicated by the canonicalloadings (Table 3.10), the species variables Athyrium filix-femina (0.202) and Rubus pedatus (0.257) made the greatestcontribution to the discrimination. The mean of Athyrium filix-femina in areas of low damage was 0.50 (standard error=0.19) andin areas of high damage was 1.83 (standard error=0.49). Rubuspedatus had a mean % cover in areas of low damage of 0.70(standard error=0.18) and in areas of high damage of 2.33(standard error=0.48). The site variable Stumps contributedheavily to the discrimination (0.205) with a mean in areas oflow damage of 1.35 (standard error=0.30) and in areas of highdamage of 3.50 (standard error=0.79), perhaps indicative of moredenning opportunities.76Table 3.10 The results of the CDA using the categorical damagevariable:^F-values,^P^values^and canonical^loadings^for thehabitat variables on the discriminant axis.F -value P Canonical(df=1,44) LoadingSlope 0.013 0.909 -0.009Aspect 0.563 0.457 -0.060Surface Shape 1.117 0.296 0.085Moisture Regime 0.080 0.779 -0.023Nutrient Regime 0.929 0.340 0.077Number of Stumps 6.526 0.014 0.205% Humus 2.899 0.096 -0.137% Dead Wood 3.998 0.052 0.161Microtopography 1.679 0.202 -0.104Drainage 0.255 0.616 -0.041Humus Form Class 1.740 0.194 0.106Athyrium filix-femina 6.295 0.016 0.202Clintonia unifoliata 0.713 0.403 0.068Cornus canadensis 1.527 0.223 -0.099Dryopteris assimilis 1.588 0.214 0.101Epilobium angustifolium 2.966 0.092 0.138Oplopanax horridus 0.004 0.952 0.005Rubus pedatus 10.253 0.003 0.257Table 3.10 continuedRubus spectabilis 0.314 0.578 0.045Streptopus streptopoides 4.828 0.033 0.177Tiarella unifoliata 6.060 0.018 0.189Vaccinium alaskense 0.022 0.883 0.012% Cover Trees 1.175 0.284 -0.087% Cover Shrubs 1.164 0.286 0.087% Cover Herbs 4.131 0.048 0.163% Cover Moss 1.741 0.194 0.1067778CHAPTER 4 DISCUSSION4.1 COMPARABILITY OF STANDSThe four study areas (grids) in which this research was carriedout had been chosen, using forest cover maps and walk-throughs,to be comparable in general site features. This was confirmedby statistical analysis. However, there was a significantdifference between grids within treatments in both the speciesvariables and the vegetation strata variables, as discussed inthe Results. It was expected that there would be variationbetween grids in the habitat data that could be used as a basisfor comparison with porcupine usage.4.2 THE EFFECT OF THINNING ON DAMAGERegarding Objective 1, thinned stands did not sustain moredamage than did unmamaged stands. On the contrary, unthinnedareas sustained significantly more new damage than did thinnedstands. As thinned stands differed from unthinned stands in thedensity and basal area of the preferred food species, westernhemlock, it appears that porcupines are exercising choice on thebasis of these properties. This was somewhat surprising asthinned areas had more trees in the preferred diameter class,though the trend was not significant.^The result is alsocontrary to earlier studies.^Harder (1980) found that79porcupines preferred low density stands.It is quite possible that the porcupines were not responding tohemlock density at all, but to the amount of slash left on theground after thinning. This slash was dense and often piledhigh and may have hampered movements by the porcupines.Porcupines are not agile animals (personal observation). Of 15adult skeletons examined, Roze (1984) found that 9 showedevidence of healed fractures, probably incurred from fallingfrom trees. In thinned areas, slash is sometimes 0.6 m deep andof less than optimum diameter for porcupines to walk on.However, for some of the winter (December to March) the slash inthe area of the study would be well covered in snow.On the basis of this research the hypothesis that thinned areassustain more damage than unthinned areas can be rejected.However, it is possible that the vigour of trees in thinnedareas may in the future surpass that of unthinned stands, makingthese stands more attractive to porcupines and predisposing themto attack.4.3 PROCESS OF SELECTION BY PORCUPINESThe process by which porcupines are selecting trees as food,whether porcupines actually select food and habitat or merelyrely upon chance, is central to understanding the observed80patterns of use (Objective 2), especially when consideringrecommendations for damage reduction measures.4.3.i Evidence for a Hierarchical Process of SelectionPorcupines seem to be choosing at the level of large blocks offorest, as thinned stands are incurring less damage than un-thinned areas. Other studies have documented the response ofwildlife to stand structure (Sullivan and Vyse 1987; Hodorff etal. 1988).Studies that have also presented evidence for choice by por-cupines on the basis of large forest units includeHarder's (1980) study, mentioned above. Harder suggested thatthe occurrence of intercommunity preferences for low densitystands resulted from the greater abundance of large, vigoroustrees in low density stands, In an earlier study Harder (1979)found that stands sustained a degree of damage that depended ontheir species composition. Tenneson and Oring (1985) also foundthat areas used heavily for winter feeding differed in the treespecies composition from less used areas. Stand density wasuniform in their study and its effects on damage could not betested. The stands in my own study were similar in speciescomposition and so the effect of this variable on stand-leveldamage could not be tested.The results of this research, however, indicate little evidence81for porcupines choosing areas of feeding in successively smallerunits. One level lower, that of the plots, could be consideredas a smaller community type or association of habitat variables.Regression analysis revealed no significant relationship betweenthe amount of damage in a plot (calculated on the basis oftotal, new or old wounds) and the structural properties of treedensity and total tree basal area.It is possible that porcupines were choosing in a hierarchicalway within a stand at a smaller unit size characterized by somecombination of site or vegetation factors. Regression analysisindicated that the herbs Athyrium filix-femina and Rubus pedatuswere increasingly abundant as the damage index variablecalculated on the basis of total wounds increased. Canonicaldiscriminant analysis confirmed these results. It is difficultto disentangle cause and effect here, but the reason for agreater cover of these species is probably greater levels oflight reaching the forest floor in highly damaged areas. Manyheavily damaged trees had dead tops, creating openings in thecanopy through which a greater amount of light could pass.However, although these two species typically occur in semi-open, coniferous forests, they are both considered to be shadetolerant (Klinka et al. 1989) and are probably not character-istic gap species. The fact that the site variables werecorrelated, though not strongly, with the vegetation variablesand yet damage was significantly related to one set (vegetation)82but not the other (site), would support the conclusion thatthese two plants were responding to some damage-relatedcondition.The only significant relationship between the amount of damageat the plot level and the site variables, as revealed by theCDA, involved the number of stumps. Porcupines typically usestumps as denning sites and insular cover is important to theseanimals in the winter (Roze 1987). Porcupines demonstrate ahigh level of den fidelity which probably results in repeateddamage to nearby trees (Roze 1987).In Harder's (1980) study, communities (on the scale of approx-imately 30 ha in size) within major blocks of forest differedfrom one another in repeated use of individual trees, withrepeated use being highest on leeward slopes. This pattern issuggestive of selective behaviour at a forest unit within thelevel of the whole stand.At a smaller community level (30 m radius plots), Speer andDilworth (1978) found that porcupines used winter areas thatwere moister and closer to standing water though they felt thatporcupines were not in fact using the water for drinking as itwas frozen for most of the winter. However, in the presentstudy, attempts to relate damage to community level factors ofsites and vegetation through regression analyses indicated no83trend strong enough to be of use in prediction.4.3.ii Evidence for Selective Feeding on the Basis ofIndividual Tree CharacteristicsThe results of the damage assessments indicated that westernhemlock was the only species attacked, though the sites werecomprised almost exclusively of this species and amabilis fir.species in previousnorth coast (SullivanPorcupines have beenpresent but it wouldWestern hemlock was also the preferredstudies of porcupine barking damage on theet al. 1986; Sullivan and Cheng, 1989).known to feed on balsam fir where it wasseem not to be abeen carried out.preferred food in any area where studies haveSpeer and Dilworth (1978) and Radvyani (1952),working in central New Brunswick, found that porcupinespreferred eastern larch (Larix laricina) and red spruce (Picearubens) over balsam fir. Curtis (1944) and Dodge (1982) havereported that feeding in their study areas was more common onspruce than on balsam.The total incidence of damage to western hemlock was high at 41%and the fact that 10% of trees had been damaged in the winterprevious to the study is alarming. With hemlocks representing43% of the trees sampled and an annual rate of attack of 10%,within which 15% had never been damaged before, it is not likelythat these areas will ever be worth harvesting. This will be84true even if only a small fraction of the attacked trees areseverely reduced in vigour or possess spike tops. Such a rateof attack is certainly precedented. An earlier study in theKalum Valley reported that the annual rate of attack hadincreased from 0.6% in 1986 to 1.8% in 1987 for western hemlock(Sullivan and Cheng 1989).The intensity of attack in the 1988-89 winter, as indicated bythe mean number of new wounds, peaked significantly in the 25.1to 30.0 cm tree diameter class. The fact that the totalincidence of damage (new and prior) was greatest in the sizeclass above this (30.1 to 35.0 cm dbh) is reasonable as thetrees were probably attacked when they were in the preferredsize class of 25.1 to 30.0 but have put on growth since the timeof attack. In a study by Sullivan et al. (1986) in the Khut-zeymateen Inlet, damage peaked at 83.3% among second-growthhemlock in the 27.5-32.4 cm dbh class.Such preferences for particular size classes of trees would seemto be typical of porcupines. Tenneson (1983) found that theamount of damage to trees resulting from the winter bark removalby porcupines increased with increasing dbh classes, preferredtrees being over 15 cm dbh. Harder (1979) found a predilectionfor a particular size class irrespective of geographic locationor tree species. This is in agreement with the suggestion byCurtis and Wilson (1953) that porcupines may be most adept at85climbing trees with a dbh between 15 and 25 cm. It is likelyeasier for porcupines to maintain a feeding position in trees ofa particular diameter, and this could explain their preference.It is also possible that trees of the preferred diameter classesprovide superior quality of food. Quality of a food item is acomplex property that involves not only its energy and nutrientcontent but also the digestion and assimilation efficiency ofthe consumer (Longhurst et al. 1968, Harder 1979). Porcupinesmay prefer to feed on trees producing a large annual incrementof phloem yet without a tough bark.Thus, in this study there is evidence that porcupines arechoosing feeding areas through a hierarchical process, evalua-ting large stands of forest but then within these stands aresimply choosing on the basis of individual tree species anddiameters.4.4 RECOMMENDATIONSThe third objective of this study, the determination of vari-ables most useful in predicting damage, pertained most directlyto the ability to put forward recommendations for forestmanagement. However, the only variables highly related to86damage are tree species and dbh. This information could be usedto space, or replant, in favour of less preferred species or toleave patches of sacrifice trees of the preferred species,western hemlock. It does not seem possible to evaluate areas interms of susceptibility to attack on the basis of site orvegetation characteristics.At the present rates of damage, however, it is expected that thedamage situation will become so severe as to have broaderimplications for the whole ecosystem. If this study wereconsidered to represent a kind of baseline data there would bea future opportunity to determine the ecosystem effects of thisherbivore by a continued monitoring of the variables measured inthis study.4.5 A COMMENT ON MULTIVARIATE METHODS IN WILDLIFE MANAGEMENTThere has been some concern that multivariate statisticalanalyses have the potential to fabricate relationships notinherent in the data. At one extreme Green (1971) has excusedthe use of some mutivariate methods in violation of assumptionsif the results make some ecological sense. However, Rexstad etal. (1988) questioned the use of the multivariate techniques instudies of wildlife habitat stating that "Sophisticated multi-variate techniques cannot be expected to replace careful, apriori thinking and design;". Although this statement is87reasonable, their own efforts to discredit certain mutivariatemethods, including PCA and CCA, by using a set of meaninglessdata, were unsupported by their analyses (Taylor 1990). Thisissue would be worth pursuing.Williams (1983) differentiates between exploratory and confir-matory analyses. The methods used in this thesis were intendedas an exploration of the data with the aim to better understandthe relationship between a destructive herbivore and itshabitat. Surely animals respond to many variables in theirenvironment simultaneously. Though one must be thoughtful whenusing multivariate techniques, and remain aware of the limita-tions, they assist in a realistic consideration of habitats.88LITERATURE CITEDAustin, M.P. 1968. An ordination study of a chalk grasslandcommunity. J. Ecol. 56: 739-757.Bannan, A.W.F. 1974. The Mammals of Canada. University ofToronto Press, Toronto, Ontario.Begon, M., J.L. Harper and C.R. Townsend. 1986. Ecology.Sinauer Associates Inc. Publishers. Mass.Bitterlich, W. 1985. Relescope: Relative Measurements inForestry. CAB International, Oxford.Botkin, D.B., J.M. Melillo, and L.S.-Y. Wu.^1981.^Howecosystem processes are linked to large mammal populationdynamics. In: Dynamics of Large Mammal Populations. C.F.Fowler and T.D. Smith (eds.). John Wiley and Sons, N.Y. pp.213-225Brander, R.B. 1973. Life history notes on the porcupine in ahardwood hemlock forest in upper Michigan. Mich. Acad. 5:425-433.Carey, A.B.^1981. Multivariate analysis of niche, habitatand ecotope. In: The use of multivariate statistics instudies of wildlife habitat. USDA For. Ser. Gen. Tec. Rep.RM-87. pp. 104-113.Cargill, S.M., and R.L. Jeffries.^1984.^The effects ofgrazing by lesser snowgeese on the vegetation of a subarcticsalt marsh. J. Appl. Ecol. 21: 669-686.Cook, D.B., and W.J. Hamilton. 1957. The forest, the fisherand the porcupine. J. For. 55: 719-722.Coupe, R., C.A. Ray, A. Comeau, M.V. Ketcheson, and R.M.Annas. 1982. A Guide to Some Common Plants of the SkeenaArea, British Columbia. Province of British Columbia Ministryof Forests Publication H28-81065.Curtis, J.D.^1941.^The silvicultural significance of theporcupine. J. For. 39: 583-594.Curtis, J.D. 1944. Appraisal of porcupine damage. J. Wildl.Manag. 8: 88-91.Curtis, J.D. and E.L. Kozicky.^1944.^Observations on theeastern porcupine. J. Mamm. 25: 137-146.Curtis, J.D. and A.K. Wilson.^1953.^Porcupine feeding onponderosa pine in central Idaho. J. For. 51: 339-341.89Dodge, W.E.^1982. Porcupine.^In: Wild Mammals of NorthAmerica. Biology, management and economics. J.A. Chapman andG.E. Feldhamer (eds.).^John Hopkins University Press,Baltimore. pp 355-366.Dodge, W.E. and P.R. Cannutt. 1969. A review of the statusof the porcupine (Erithizon dorsatum) in western Oregon.^Bur.Sport Fisheries and Wildl., Portland, Oreg. 25 pp.Eglitis, A. and P.E. Hennon. 1986. Porcupine damage in Sitkaspruce and western hemlock in southeast Alaska. USDA ForestService, Pest Management Report. Juneau. 21pp.Eglitis, A. and P.E. Hennon. 1987. Porcupine damage in Sitkaspruce and wester hemlock in southeast Alaska. Proc. of Symp.Animal Damage Management in Pacific Northwest Forests.Spokane, Wash. pp 79.Gabrielson, I.N. 1928. Notes on the habits and behaviour ofthe porcupine in Oregon. J. Mamm. 9: 269-289.Gagnon, D. and G.E. Bradfield.^1987. Gradient analysis ofwest central Vancouver Island forests. Can. J. Bot. 65:^822-833.Gauch, H.G.^1982.^Multivariate Analysis in CommunityEcology. Cambridge University Press.Gill, D., and L.D. Cordes. 1972. Winter habitat preferenceof porcupines in the southern Alberta foothills. Can.^Field-Nat. 86:349-355.Gittins, R.^1979.^Ecological applications of canonicalanalysis. In: L. Orloci, C.R. Rao, and W.M. Stiteler^(eds.).Multivariate Methods in Ecological Work.^pp.309-^535.Burtonsville, Md. :International Cooperative.Gittins, R. 1985. Canonical Analysis. Springer-Verlag, NewYork.Golley, F.B.^1956. Distribution of the porcupine in upperMichigan. J. Mammal. 38: 526-527.Green, R.H. 1971. A multivariate statistical approach to theHutchinsonian niche: bivalve molluscs of central Canada.Ecology 52: 543-556.Green, R.H.^1974.^Multivariate niche analysis withtemporally varying environment factors. Ecology 55: 73-83.Grillos, S.J., and F.H. Smith. 1959. The secondary phloem ofDouglas-fir. For. Sci. 5: 377-388.90Harder, L.D. 1979. Winter feeding by porcupines in montaneforests of southwest Alberta. Can. Field-Nat. 93: 405-410.Harder, L.D.^1980.^Winter use of montane forests byporcupine in southwestern Alberta: preferences, densityeffects and temporal changes. Can. J. Zoo. 58: 13-19.Hatton, J.C., and N.O.E. Smart.^1984.^The effect of longterm exclusion of large herbivores on soil nutrient status inMurchison Falls National Park, Uganda. Afr. J. Ecol. 22: 23-30.Hitchcock, C.L. and A. Cronquist. 1973. Flora of the PacificNorthwest. Univ. Washington Press, Seattle, Washington..Hodorff, R.E., C.H. Sieg, and R.L. Linder.^1988. Wildliferesponse to stand structure of deciduous woodlands. J. Wildl.Manage. 52(4): 667-673.Hooven, E.F.^1971.^The porcupine in Oregon.^Its lifehistory and control. Oreg. State Univ. For. Res. Lab., Res.Pap. No. 10.Klinka, K., R.N. Green, R.L. Trowbridge, and L.E. Lowe. 1981.Taxonomic classification of humus forms in ecosystems ofBritish Columbia. Province of British Columbia, Ministry ofForests, Land Management Report, ISSN 0702-9861; no. 8.Klinka, K., V.J. Krajina, A. Ceska, and A.M. Scagel. 1989.Indicator Plants of Coastal British Columbia. University ofBritish Columbia Press. 288 pp.Krefting, L.W., J.H. Stoeckler, R. Bradle , and W.D.Fitzwater. 1962. Porcupine-timber relationships in the LakeStates. J. For. 60: 325-330.Lack, D. 1954. The Natural Regulation of Animal Numbers.Oxford University Press, Oxford.Longhurst, W.N., H.K. Oh, M.B. Jones, and R.E. Kepner. 1968.A basis for the palatability of deer forage plants.Transactions of the North American Wildlife and NaturalResouces Conference 33: 181-189.Lyons, C.P.^1952.^Trees, Shrubs and Flowers to Know inBritish Columbia. J.M. Dent and Sons Limited. Vancouver.Mattson, W.J., and N.D. Addy. 1975. Phytophagous insects asregulators of forest primary production.^Science 190: 515-522.91Meehan, W.R. 1974. The forest ecosystem of southeast Alaska.Wildlife habitats. U.S. Dep. Agri. For. Ser. Gen. Tech. Rep.PNW-16.Naiman, R.J. 1988. Animal influences on ecosystem dynamics.Bioscience 38: 750-752.Naiman, R.J., C.A. Johnston, and J.C. Kelley.^1988.Alteration of North American Streams by beaver. Biosciences38: 753-762.Pastor, J., R.J. Naiman, and B. Dewey. 1987. A hypothesis ofthe effects of moose and beaver foreging on soil carbon andnitrogen cycles, Isle Royal. Alces 23: 107-124.Pastor, J., R.J. Naiman, B. Dewey, and P. McInnes.^1988.Moose, microbes, and the boreal forest. Bioscience 38: 770-777.Pimentel, R.A. 1979. Morphometrics. Kendall/Hunt PublishingCompany, USA.Radvyani, A. 1952. Report on the porcupine and its damage.Canada Deptartment of Research and Development, ForestryBranch, Maritime District, Acadia Experiment Station,Fredricton, N.B.Reeks, W.A.^1942. Notes on the Canadian porcupine in theMaritime provinces. Forestry Chronical 18: 182-187.Rexstad, E.A., D.D. Miller, C.H. Flather, E.M. Anderson, J.W.Hupp, and D.R. Anderson. 1988. Questionable multivariatestatistical inference in wildlife habitat and communitystudies. J. Wildl. Manage. 52: 794-798.Rizenhoover, K.L., and S.A. Mass. The influence of moose onthe composition and structure of Isle Royal forests. Can. J.For. Res. 17: 357-366.Roze, U. 1984. Winter foraging by individual porcupines.Can. J. Zool. 62: 2425-2428.Roze, U. 1987. Denning and winter range of the porcupine.Can. J. Zool. 65: 981-986.Rudolph, P.O.^1949.^Porcupine preference in pineplantations. J. For. 47: 207-209.Ruth, R.H., and A.S Harris.^1979.^Management of westernhemlock - Sitka spruce forests for timber production. U.S.Dep. Agric. For. Serv. Gen. Tech. Rep. PNW-88.92Sanderson, G.C., E.D. Ables, R.D. Sparrows, J.R. Groeb, L.D.Harris, and A.N. Moen. 1979. Research needs in wildlife.Transactions North American Wildlife and Natural ResourcesConference 44: 166-175.Shapiro, J.^1949.^Ecology and life history notes on theporcupine in the eastern Adirondacks. J. Mamm. 30: 247- 257.Speer, R.J., and T.G. Dilworth. 1978. Porcupine winter foodsand utlization in central New Brunswick. Can. Field Nat. 92:271-274.Storm, G.L., and C.H. Halverson. 1967. Effect of injury byporcupines on radial growth of ponderosa pine. J. For. 65:740-743.Sullivan, T.P. and C. Cheng.^1989.^Impact of porcupinefeeding damage on intensive forest management in the PrinceRupert Forest Region: Problem analysis. Province of BritishColumbia, Ministry of Forests, ISSN 0835 0752.Sullivan, T.P., and D.S. Sullivan. 1982a. Barking damage bysnowshoe hares and red squirrels in lodgepole pine stands incentral British Columbia. Can. J. For. Res. 12: 443-448.^ . 1982b. Influence offertilization on feeding attacks to lodgepole pine bysnowshoe hares and red squirrels. For. Chron. 58: 263-266.Sullivan, T.P., and A. Vyse.^1987. Impact of red squirrelfeeding damage on spaced stands of lodgepole pine in theCariboo Region of British Columbia. Can J. For. Res. 17:^666-674.Sullivan, T.P., W.T. Jackson, J. Pojar, and A. Banner. 1986.Impact of feeding damage by the porcupine on western hemlock-Sitka spruce forests of north-coastal British Columbia. Can.J. For. Res. 16: 642-647.Taylor, J.^1990.^Questionable statistical inference inwildlife habitat and community studies: a comment. J Wildl.Manage. 54(1): 186-189.Taylor, W.P.^1935.^Ecology and life of the porcupine(Erethizon epixanthum) as related to the forests of Arizonaand Southwestern and the Southwestern United States. Univ.Ariz., Biol. Bull. 3: 1-177.Tenneson, C.^1983.^Winter ecology of the porcupine,Erethizon dorsatum. M.S. Thesis, Univ. North Dakota, GrandForks. 62pp.93Tenneson, C., and L.W. Oring. 1985. Winter food preferencesof porcupines. J. Wildl. Manag. 49:28-33.Van Deusen, J.L., and C.A. Myers. 1962. Porcupine damage inimmature stands of ponderosa pine in the Black Hills. J. For.60: 811-813.Walmsley, M., G. Utzig, T. Vold, D. Moon, and J. vanBarneveld.^1980.^Describing Ecosystems in the Field.Province of British Columbia, Ministry of the Environment,Ministry of Forests, RAB Technical Paper 2.Williams, B.K.^1983.^Some observations on the use ofdiscriminant analysis in ecology. Ecology 64: 1283-1291.Williams, W.T. (ed.). 1976. Pattern Analysis in AgriculturalScience. Elsevier Press, New York.


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